From vision to a data strategy you can run like a scorecard
Derive the data strategy from the business strategy, one level at a time, and run it with measures, targets, owners and a quarterly rhythm. Six worked industry examples, eight strategy tracks from metadata to decision automation, calculators and 22 decks to download.
Each level is derived from the one above and tested by the one below.
Chapter 1 of 7
The cascade
1.1The idea
A data strategy is the business strategy, seen through its information
A data strategy is not a separate strategy. It is the part of the business strategy that decides which information the organisation must master to win, and how it will measure, fund and govern that.The SCIKIQ view.
It isIt is not
Derived from the business strategyA separate plan written by the data team
Objectives traced to the strategy mapA list of platforms and projects
Measured on a balanced scorecardMeasured by reports built and tools deployed
A deliberate offence and defence balanceGovernance or analytics, whichever is louder
Funded through stage gates on valueFunded once, up front, by department
Reviewed monthly, refreshed yearlyPresented once and filed
1.2The strategy cascade
Nine levels from vision to the operating rhythm
Each level is derived from the one above and tested by the one below. Walk it with a worked example: pick an industry, then step down the cascade.
Exhibit 1Cascade explorer, with a worked example at every level
Worked example
Steps down all nine levels for the selected industry, with what was decided and the so what at each.
Level 1 of 9
Vision: Where we are going
The destination, ten years out: what the organisation wants to become and be known for.
Good looks like
Ambitious, specific enough to rule things out, and memorable enough that people repeat it unprompted.
The data leader's question
Which decisions will this future depend on that we cannot make with today's data?
The trap
A vision so generic it could belong to any competitor, so nothing below it has to change.
In the worked example:
Level 2 of 9
Mission: Why we exist
Who the organisation serves, what it does for them and how, today.
Good looks like
Names the customer and the value delivered; holds steady while strategies change.
The data leader's question
Who are the customers of our data, and what do they need it to do?
The trap
Confusing the mission with a list of products or a slogan.
In the worked example:
Level 3 of 9
Strategic choices: Where to play, how to win
The integrated set of choices: winning aspiration, where to play, how to win, the capabilities that make it possible, and what we will not do.
Good looks like
Choices that exclude options. Each how-to-win choice is backed by a capability someone can build.
The data leader's question
Which must-have capabilities are data capabilities, and which choices are impossible without them?
The trap
A list of goals with no choices: growing everywhere, for everyone, on every dimension.
In the worked example:
Level 4 of 9
Strategic themes: The pillars of the strategy
Three to five themes that group the strategic objectives into streams of work leaders can own.
Good looks like
Few themes, each with an executive owner and a clear intent; together they cover the strategy.
The data leader's question
Which themes share the same data, so that one investment serves several of them?
The trap
Themes that mirror the org chart instead of the strategy.
In the worked example:
Level 5 of 9
Strategy map: Cause and effect on one page
Strategic objectives arranged in four perspectives (financial, customer, internal process, learning and growth) and linked by cause and effect.
Good looks like
Fifteen to twenty-five objectives, each a verb and a noun, with links a sceptical executive would accept.
The data leader's question
Which process and learning objectives fail without trusted information? That is where data earns its place.
The trap
Objectives with no links: a list dressed up as a map.
In the worked example:
Level 6 of 9
Data objectives: What data must achieve
Data strategy objectives, each one traced to the strategy-map objectives it enables, balanced between offence and defence.
Good looks like
Every data objective names the business objectives it serves; none exists for its own sake.
The data leader's question
If we stopped this data objective, which business objective would fail?
The trap
Data objectives written as technology projects: migrate, implement, upgrade.
In the worked example:
Level 7 of 9
Data scorecard: Measures, targets, initiatives
A balanced scorecard for the data strategy: objectives across value, stakeholder, data operations and capability, each with lead and lag measures, a baseline, a target, an initiative and an owner.
Good looks like
One or two measures per objective, a mix of leading and lagging, with baselines measured rather than guessed.
The data leader's question
Which leading measure will tell us in one quarter that the lagging one will move in a year?
The trap
Measuring activity (reports built, tools deployed) instead of outcomes.
In the worked example:
Level 8 of 9
Portfolio: Use cases, data products, enablers
The prioritised use cases, the reusable data products they draw on, and the foundations they depend on, funded through stage gates.
Good looks like
Use cases ranked on value and feasibility; data products built once and used many times.
The data leader's question
Which data product unlocks the most use cases, and is it funded as shared infrastructure?
The trap
Funding each use case as a one-off project, rebuilding the same data every time.
In the worked example:
Level 9 of 9
Roadmap and rhythm: How the strategy stays alive
The three-horizon roadmap, the first hundred days and the management rhythm: monthly scorecard reviews, quarterly strategy reviews and an annual refresh.
Good looks like
Reviews that test the assumptions and move money, not status meetings.
The data leader's question
What would have to be true for us to stop, start or re-fund an initiative this quarter?
The trap
A roadmap that is presented once and never reviewed.
In the worked example:
Worked examples are illustrative companies; baselines and targets are illustrative.
Seven strategic themes and the data plays behind them
Most strategies are built from a handful of recurring themes. For each: the business objectives, the data plays that serve them, the data products they need, a leading and lagging measure, and how it looks in six industries.
Exhibit 2Seven themes: objectives to data plays to data products
Profitable growth
Offence
Business objectives
Grow share of wallet in priority segments
Price for value, not volume
Win in chosen channels
Data plays
Customer 360 and propensity models
Price and margin analytics
Pipeline and opportunity intelligence
Data products
Customer 360Product and price masterSales pipeline
LeadingShare of opportunities with a propensity score in the CRM
LaggingRevenue growth in target segments
How it looks by industry
Industry
Business objective
Data play
Measure
Manufacturing
Grow aftermarket revenue from the installed base
Join installed-base and condition data to predict replacements and price service contracts
Recurring service revenue as a share of sales
Banking
Grow income from existing customers
Time relevant offers from life-event signals in a single customer view
Primary customers holding two or more products
Retail
Grow share of wallet among loyalty members
Personalise offers and recommendations from consented purchase history
Spend per active loyalty member, rolling 12 months
Healthcare
Grow planned-care activity within existing capacity
Forecast demand by specialty and fill theatre lists from a live capacity view
Planned-care cases per theatre session
Telecom
Grow broadband take-up among mobile customers
Target bundle offers using fibre footprint, household and usage data
Mobile customers also taking broadband
Insurance
Grow profitable small commercial premium through brokers
Score broker submissions against appetite and quote standard risks straight through
Small commercial premium written within target loss ratio
Customer experience and loyalty
Offence
Business objectives
Resolve issues first time
Personalise every interaction
Retain the most valuable customers
Data plays
Journey analytics
Next best action
Service and complaint insight
Data products
Customer 360Interaction historyService cases
LeadingShare of interactions handled with full customer context
LaggingRetention of high-value customers
How it looks by industry
Industry
Business objective
Data play
Measure
Manufacturing
Be the supplier customers plan production around
Commit promise dates from one shared demand and capacity view
On time in full against customer request date
Banking
Keep valuable relationships from drifting away
Flag attrition signals in balances and transactions and route calls to relationship managers
Attrition among top-value relationships
Retail
Recognise the same customer in every channel
Link loyalty, web, app and store transactions to one consented profile
Transactions linked to an identified, consented customer
Healthcare
Cut waits and wasted clinic slots for patients
Predict no-show risk and rebook freed slots from the live clinic view
Clinic slots lost to late cancellation or no-show
Telecom
Cut churn among high-value customers
Score churn risk weekly from billing, usage and network experience data
Monthly postpaid churn
Insurance
Settle genuine claims faster
Triage claims at first notice and route simple claims straight through
Simple claims settled within ten working days
Operational excellence
Balanced
Business objectives
Deliver on time, in full
Lower cost to serve
Release working capital
Data plays
Demand sensing and inventory optimisation
Process mining
Predictive maintenance
Data products
Demand and inventoryOrder-to-cash eventsAsset health
LeadingShare of plans built on integrated, daily data
LaggingCost to serve and on-time-in-full
How it looks by industry
Industry
Business objective
Data play
Measure
Manufacturing
Build quality in at the line
Link process, inspection and supplier-lot data to every serial number
First-pass yield on critical lines
Banking
Lower the cost of routine servicing
Mine servicing events and cost data to choose the next journeys to automate
Cost-to-income ratio
Retail
Plan and fulfil from one stock pool
Combine store, warehouse and in-transit stock into one live position
Stock cover in weeks at target availability
Healthcare
Discharge patients as soon as they are ready
Show live bed status and discharge blockers across wards in one view
Ready patients still in hospital after 24 hours
Telecom
Fix network faults before customers call
Match network alarms to affected customers and contact them proactively
Fault-related calls per thousand customers
Insurance
Cut claims leakage
Score claims for overpayment risk and target file reviews on the highest
LeadingShare of priority suppliers reporting primary data
LaggingEmissions intensity and assured disclosures
How it looks by industry
Industry
Business objective
Data play
Measure
Manufacturing
Cut energy and scrap per unit produced
Join meter, machine and scrap data by line and shift
Energy used per good unit produced
Banking
Measure and reduce financed emissions
Estimate emissions across the loan book from property and business sector data
Lending covered by an emissions estimate
Retail
Cut waste from unsold and returned stock
Buy and allocate closer to demand using forecasts and returns data
Stock written off as a share of sales
Healthcare
Reduce energy use and clinical waste across sites
Track energy, waste and theatre consumables by site and procedure
Energy use per occupied bed day
Telecom
Cut network energy use
Power down spare capacity in quiet hours using cell traffic forecasts
Network energy per terabyte carried
Insurance
Repair rather than replace after claims
Steer claims to lower-waste repair options using claims and supplier data
Claims settled by repair rather than replacement
People and productivity
Offence
Business objectives
Put insight in every role
Free time for higher-value work
Grow scarce skills
Data plays
Self-service analytics
AI assistants on governed data
Workforce analytics
Data products
Semantic layer and metricsKnowledge baseWorkforce
LeadingMonthly active users of governed data
LaggingHours released and output per employee
How it looks by industry
Industry
Business objective
Data play
Measure
Manufacturing
Put analytics skills in plant teams
Certify process engineers in self-service analytics on governed plant data
Process engineers certified in self-service analytics
Banking
Give colleagues trusted self-service data
Publish certified reports and dashboards from one governed source
Monthly active users of certified reports
Retail
Put analytics next to trading decisions
Embed analytics translators in every trading and marketing team
Teams with an embedded analytics translator
Healthcare
Match nurse staffing to patient demand
Forecast admissions and ward acuity to set rosters a week ahead
Shifts filled by temporary staff
Telecom
Equip business sellers with account insight
Give sellers one account view of services, spend and network experience
Sellers using account insight every week
Insurance
Free experienced handlers for complex claims
Route simple claims automatically so experienced handlers focus on complex cases
Complex claims handled per experienced handler
Illustrative. Source: SCIKIQ Data Strategy Framework.
2.2Balanced scorecard
Put the strategy on one map, linked by cause and effect
Four perspectives: learning and growth enables internal processes, processes deliver the customer value proposition, customers drive the financials. Information capital sits in learning and growth; that is where the data strategy plugs in.
01
Financial
To succeed financially, how should we appear to our owners? Data shows up here as value: revenue enabled, cost removed, risk avoided, capital released. Never as a data objective.
02
Customer
To achieve our vision, how should we appear to our customers? The customer value proposition (price, quality, availability, service, relationship) sets which customer data matters.
03
Internal process
To satisfy customers, which processes must we excel at? Operations, customer management, innovation, and regulatory and social processes: most data use cases live here.
04
Learning and growth
To excel at those processes, what people, information and culture do we need? Information capital sits here with human and organisation capital. This is where the data strategy plugs into the business strategy.
Exhibit 3Strategy map: four perspectives linked by cause and effect
Worked example
Hover or tap an objective to trace its causes and effects, and the data objectives that serve it. Illustrative.
Six rules for a strategy map that works
Write objectives as verb and noun"Reduce unplanned downtime", not "Downtime". An objective is something you do.
Keep it to one pageFifteen to twenty-five objectives. More than that and nobody can hold the strategy in their head.
Link by cause and effectEach link is a hypothesis: if we improve this, that will move. Reviews test the hypotheses.
Pair leading and lagging measuresLagging measures prove results; leading measures tell you early whether you will get them.
Fund initiatives by objectiveInitiatives are funded because they move an objective, not because a department asked.
Name an owner for every objectiveA person, not a committee, accountable for the measure and the initiatives behind it.
2.3Offence and defence
Choose your balance of control and growth deliberately
Defence protects the numbers the organisation is judged on; offence uses data to grow. Every strategy needs both, in a proportion set by regulation, competition and how trusted the data is today.
Exhibit 4Posture calculator: where should your data strategy sit?
How heavily regulated is your industry?
How much does your advantage depend on analytics, personalisation or AI?
How often do leaders argue about whose number is right?
How mature are data ownership and quality today?
How ambitious are your growth and AI plans for the next two years?
How complex is your estate: entities, systems, acquisitions?
DefenceOffence
50% offence
Answer the six questions
Your recommended balance between control and growth appears here as you answer.
Defence
Control, compliance, security and quality: one source of truth for the numbers the organisation is judged on.
Certified definitions for reported metrics
Ownership and quality rules on critical data
Lineage from report to source
Access control and privacy by design
Offence
Growth, customer value and AI: flexible, fast access to data, with many fit-for-purpose views built from the trusted core.
The same discipline, applied to data: value, stakeholder, data operations and capability. One or two measures per objective, leading and lagging, with measured baselines, targets, initiatives and owners.
Value
How does data move the business results on the strategy map?
Typical objectives
Enable growth in priority segments
Remove cost from core processes
Reduce regulatory and operational risk
LeadingUse cases moved from pilot to production; Decisions running on governed data
LaggingBenefit realised against baseline; Revenue, cost or risk outcome of each use case
Stakeholder
How do the business users and customers of our data see us?
Typical objectives
Make trusted data easy to find and use
Answer business questions fast
Earn customer trust in how we use their data
LeadingTime from question to trusted answer; Self-service share of questions
LaggingBusiness partner satisfaction; Adoption of data products by target users
Data operations
Which data processes must we excel at?
Typical objectives
Govern critical data with owners and rules
Deliver data products to service levels
Protect sensitive data by design
LeadingCritical data elements with owner and rules; Time to onboard a new source
LaggingQuality score of critical data; Data incidents and audit findings
Capability
What people, platform and culture do we need?
Typical objectives
Build data literacy in every role
Run a reusable, governed platform
Grow data product ownership in the business
LeadingLiteracy pulse score; Share of new work built on existing data products
LaggingRoles filled with the target skills; Run cost per data product
Exhibit 5A worked data strategy scorecard
Worked example
Illustrative. Source: SCIKIQ Data Strategy Framework.
Measures library
Value measures
Benefit realisedLag Measured benefit of data initiatives against baseline, confirmed with finance
Revenue enabledLag Revenue from use cases attributed with an agreed method
Cost removedLag Run-rate cost removed by data-enabled process change
Use cases in productionLead Use cases past the Scale gate and in use
Decisions on governed dataLead Priority decisions that now run on certified data
Value pipelineLead Estimated value of use cases at Discover and Prove
Stakeholder measures
Partner satisfactionLag Business partner rating of data and analytics services
Data product adoptionLag Target users actively using each data product
Time to trusted answerLead Elapsed time from business question to certified answer
Self-service shareLead Questions answered without the data team
Customer data trustLag Complaints and consent withdrawals related to data use
Request backlog ageLead Median age of open data requests
Data operations measures
Glossary coverageLead Share of critical data elements with an approved business definition and owner
Time to find dataLead Median time for an analyst to find and validate the data behind a new question
Critical data qualityLag Share of critical data elements passing their rules
Owned critical dataLead Critical data elements with owner, definition and rules
Certified metricsLead Headline metrics with one certified definition
Lineage coverageLead Critical reports traceable to source
Data incidentsLag Quality or availability incidents affecting the business
Time to onboard a sourceLead Days from request to governed, usable source
Capability measures
Literacy pulseLead Average score on the data literacy pulse check
Skills coverageLag Roles filled with the target data skills
Reuse rateLead New work built on existing data products
Domain ownershipLead Domains with a business data product owner
Run cost per data productLag Total run cost divided by active data products
Platform availabilityLag Availability of the data platform against its service level
Define every measure the same way
MeasureShort name everyone uses
Objective servedThe scorecard objective it measures
Why it mattersThe decision it informs
Definition and formulaNumerator, denominator, inclusions and exclusions
Lead or lagLeading (predicts) or lagging (proves)
SourceSystem of record and data product
OwnerAccountable person, and who reports it
FrequencyHow often it is refreshed and reviewed
BaselineMeasured value, and when
TargetValue and date
ThresholdsGreen, amber, red
CaveatsKnown data limitations
3.2Value at stake
Rank the portfolio on value and feasibility, then release money through gates
Move the weights to match your themes; the ranking updates. Money follows evidence: each gate releases the next tranche.
Exhibit 6Value-at-stake ranking: move the weights, watch the portfolio re-rank
Weight the value levers to your strategy
0 = irrelevant, 3 = critical. Feasibility counts for a third of the score.
Illustrative use cases with indicative scores. Score your own in the Strategy Cascade and Scorecard Workbook.
Fund through stage gates
1DiscoverIs the problem worth solving?
Named business owner
Decision to improve is explicit
Value hypothesis and baseline
Data availability checked
2ProveDoes it work on real data?
Pilot on production data
Measured effect against baseline
Users have changed a decision
Run cost estimated
3ScaleCan it run at scale?
Built on governed data products
Adoption target met
Controls and monitoring in place
Benefit in the finance forecast
4RunIs it still earning its keep?
Service levels met
Benefit tracked every quarter
Cost per use within plan
Retire if value fades
Exhibit 7Build once, use many: data products by use case
Data products that serve several use cases are funded as shared infrastructure, not rebuilt by each project.
Worked example
Illustrative. Source: SCIKIQ Data Strategy Framework.
Chair: Domain owner (business) Members: Stewards, data product owners, analysts
Decides: Definitions, quality rules, data product backlog, access requests
Fortnightly
Design authority
Chair: Chief architect Members: Platform, security, data engineering leads
Decides: Architecture standards, build or reuse, technology choices
Exhibit 8Decision rights across the governance forums
Decision
Exec council
CDO
Domain owner
Steward
Architecture
Risk and privacy
Set data strategy and priorities
A
R
C
I
C
C
Approve the funding envelope
A
R
C
I
I
Certify a business metric
I
C
A
R
C
Own a data product and its service level
C
A
R
C
I
Set architecture standards
I
C
C
A
C
Grant access to sensitive data
I
A
R
C
C
Retire reports and data sets
A
R
R
C
I
A accountable, R responsible, C consulted, I informed. One A per decision.
4.2Rhythm and the first 100 days
A strategy is a management system, not a document
Monthly scorecard reviews, quarterly strategy reviews that test the cause-and-effect assumptions and move money, an annual refresh of the cascade, and a hundred-day plan to start.
Weekly
Delivery
Inputs: Sprint progress, blockers, data incidents
Decisions: Unblock, re-sequence within the quarter
Monthly
Scorecard review
Inputs: Data scorecard, leading measures, initiative status
Decisions: Corrective actions on red measures
Quarterly
Strategy review
Inputs: Scorecard trends, tested assumptions, value realised, new opportunities
Decisions: Stop, start, re-fund; re-balance offence and defence
Annually
Strategy refresh
Inputs: Business strategy changes, maturity re-baseline, market shifts
Decisions: Refresh the cascade, objectives, targets and funding
The quarterly strategy review
Scorecard: which measures moved, which did not, and why
Hypotheses: which cause-and-effect links held, which need rethinking
Value: benefits realised against the plan, confirmed with finance
Portfolio: stage-gate decisions; stop, start or re-fund
Risks and triggers: anything that forces an early refresh
Asks: decisions and support needed from the executive team
Exhibit 9The first 100 days: the generic plan, and the worked example
Weeks 1-2
Mobilise
Sponsor and core team named
Interview schedule set with executives
Existing strategy documents and scorecards gathered
Maturity baseline launched
Weeks 3-6
Diagnose and anchor
Executive and business interviews complete
Draft cascade: vision to themes
Decision and capability heat map
Baseline of value, spend and maturity
Weeks 7-10
Choose and design
Workshop 1: strategy map and data objectives
Workshop 2: posture and prioritised portfolio
Draft data scorecard with measures and owners
Operating model and forums agreed
Weeks 11-14
Commit and launch
Roadmap and funding approved by the council
Strategy on a page published
First quick wins in delivery
Monthly and quarterly rhythm started
Compare with the worked example
Triggers for an early refresh
Business strategy changeNew goals, markets or business model: revisit anchor and prioritise.
Merger, acquisition or divestmentNew data estates and owners: revisit design and sequence.
Regulatory changeNew obligations: revisit governance, risk and priorities.
Missed measuresA KPI off track for two quarters: root-cause and re-plan.
Technology shiftA step change in platforms or AI: revisit architecture and value.
4.3How to build it
Six stages that produce the cascade
The method: baseline, anchor in the business, prioritise, design, sequence and mobilise, run as a cycle. Each stage names the template to use.
Stage 1 of 6
Baseline: Where are we now?
Take an honest view of the starting point: what the last plan delivered, which assumptions held, how mature data management is today and what the business thinks of the data function.
Key questions
What did the previous plan promise, and what did it deliver?
Which assumptions and risks played out, and which did not?
How mature are governance, quality, integration and analytics today?
How do business partners rate the data and analytics they receive?
Activities
Review previous initiatives and root-cause the results
Test last cycle's assumptions and risks
Run a data maturity assessment
Map where data and analytics money is spent, inside and outside the data team
Interview business partners on satisfaction and pain points
Outputs
Lessons-learned logMaturity baseline by dimensionSpend mapBusiness partner perception summary
Who is involved
Chief data officer or head of data
Data and analytics leads
Finance partner
Business partners
Use from the toolkit
Stage 2 of 6
Anchor: What does the business need?
Tie the strategy to the business strategy: the goals that matter, the capabilities that deliver them and the decisions inside those capabilities that data can improve.
Key questions
Which business goals will the next two to three years be judged on?
Which business capabilities deliver those goals?
Which decisions inside those capabilities are slow, inconsistent or made on instinct?
Which external trends change what data the business needs?
Activities
Interview executives and business partners
Map goals to business capabilities
Rate the decision health of each priority capability
Scan for market, regulatory and technology shifts
Outputs
Goal-to-capability mapDecision health heatmapTrend and implication listInterview synthesis
Who is involved
Executive sponsors
Business unit leaders
Strategy or transformation office
Data strategist
Use from the toolkit
Stage 3 of 6
Prioritise: Where will data create the most value?
Turn needs into a ranked portfolio of opportunities, scored for business value and feasibility, and decide what not to do.
Key questions
Which opportunities enhance today's business, and which could transform it?
What is each opportunity worth, and how sure are we?
Is the data ready, and do we have the skills to deliver?
What are we explicitly choosing not to do this cycle?
Set the data objectives and design the enablers that deliver them: data domains and products, governance, architecture, people and literacy, operating model and funding.
Key questions
Which data objectives follow from the chosen opportunities?
Which data domains and data products must exist, and who owns them?
What governance, platform and skills do they depend on?
How will the data function be organised and funded?
Activities
Write data objectives linked to business goals
Define priority domains and data products
Decide the eight strategy pillars
Choose the operating model
Size the investment
Outputs
Objectives linked to goalsData product and domain mapPillar decisionsOperating modelInvestment case
Who is involved
Head of data
Domain and data product owners
Enterprise architect
HR and learning partner
Use from the toolkit
Stage 5 of 6
Sequence: In what order, and how will we know?
Turn the design into a roadmap of initiatives across three horizons, each with an owner, a measure of success and the triggers that would force a rethink.
Key questions
What must come first because everything else depends on it?
What can show value within six months?
How will each initiative be measured, and against what baseline?
What events would make us revisit the strategy early?
Activities
Write an initiative-on-a-page for each initiative
Build the roadmap across three horizons
Set success measures and baselines
Agree review cadence and triggers
Outputs
Initiative-on-a-page setThree-horizon roadmapKPI scorecard with baselinesReview triggers
Who is involved
Programme lead
Initiative owners
Finance partner
PMO
Use from the toolkit
Stage 6 of 6
Mobilise: How do we bring people with us?
Put the strategy on one page, tell it as a story for each audience, and start the communication and review rhythm that keeps it alive.
Key questions
What is the one-sentence strategy, and the story behind it?
What does each audience need to hear, and what do we need from them?
Which channels and moments will carry the message?
How will we report progress and ask for help?
Activities
Write the strategy on a page
Build the narrative and the audience message map
Present to the board and executive team
Launch the communication and review cadence
Outputs
Strategy on a pageExecutive presentationMessage map by audienceCommunication calendar
Who is involved
Executive sponsor
Head of data
Internal communications
Business champions
Use from the toolkit
Six reasons data strategies stall
01
Technology first
The plan opens with a platform choice and works backwards to a reason.
Hover for the counter-move
Counter-move
Start from business goals and the decisions behind them; choose technology last.
02
Boil the ocean
Every domain, every system and every quality issue is in scope on day one.
Hover for the counter-move
Counter-move
Pick the few data domains that carry the most value and finish them.
03
No business owner
Data teams write the strategy, then try to sell it.
Hover for the counter-move
Counter-move
Co-author with business leaders; give every objective a business owner.
04
No line of sight
Initiatives cannot be traced to a business goal or a measure.
Hover for the counter-move
Counter-move
Link every initiative to a goal, a decision and a KPI before funding it.
05
Shelfware
A long document that nobody outside the data team has read.
Hover for the counter-move
Counter-move
Fit the strategy on one page and tell it as a story for each audience.
06
One and done
The plan is never revisited while the business moves on.
Hover for the counter-move
Counter-move
Review quarterly, refresh annually, and agree triggers that force an early rethink.
Metadata, value proposition, capability gaps, architecture, governance platform, DataOps, cloud and decision automation: the decisions every data strategy has to make, one track each.
Exhibit 758 tracks, plugged into the six-stage method
1BaselineWhere are we now?
2AnchorWhat does the business need?
3PrioritiseWhere will data create the most value?
4DesignWhat must be true to deliver it?
5SequenceIn what order, and how will we know?
6MobiliseHow do we bring people with us?
Pick a track to open it below. Each has a method, an interactive tool, six industry examples, measures and a deck.
Track 1 of 8 · Stage 4: Design
Metadata strategy
Metadata is the data about your data: what it means, where it came from, who owns it, how good it is and who uses it. A data strategy without a metadata strategy cannot prove that any objective was met.
Every level of the cascade below the strategy map depends on being able to find, trust and trace data. Data objectives become measurable only when the measures behind them have certified definitions and visible lineage back to source. Metadata is also the context that AI and agents need to act safely, so it now decides how far automation can go. Treat it as a strategic asset with an owner, a business case and a scorecard, not as a tool the data team buys.
So whatFund metadata as the traceability layer of the data strategy: start from the few critical measures on your scorecard, make their meaning and lineage visible end to end, and grow coverage only where a business sponsor can show the value.
Questions this track answers
Which measures on our scorecard must have a certified definition and lineage to source, and by when?
Who owns business meaning, and who owns the technical description of each critical data set?
Which business outcome will the metadata programme be accountable for, and what is its baseline?
How much of our metadata is captured passively, and where should it start to trigger action?
What context will AI assistants and agents need before we let them answer questions or act on data?
How will we prove the value to sponsors each quarter, in their language rather than ours?
Exhibit 10Four kinds of metadata, one picture
Four kinds of metadata, one picture
Each kind answers a different question. Value comes from joining them, so a business term leads to the tables, the pipelines, the quality score and the people who rely on it.
01
Business metadata
What does it mean? Terms, definitions, business rules, owners, sensitivity and the decisions a data set supports.
02
Technical metadata
What is it physically? Systems, schemas, tables, columns, types, keys and the code that transforms them.
03
Operational metadata
What happened to it? Load times, freshness, volumes, failures, quality results and lineage from source to report.
04
Usage metadata
Who relies on it? Queries, reports, downstream consumers, ratings, comments and how often each asset is used.
So whatStart with the business kind for your critical measures; technical and operational metadata can largely be harvested automatically, meaning cannot.
Exhibit 11From passive record to active metadata
From passive record to active metadata
Most organisations document metadata after the fact. The strategic shift is to let metadata drive what happens next.
Passive metadata
Captured in a catalogue and read by people when they remember to look
Refreshed by periodic scans or manual documentation drives
Answers where data is and what it means
Value depends on adoption, which often fades after launch
vs
Active metadata
Continuously harvested and analysed as systems and usage change
Triggers action: alerts owners on drift, blocks a release that breaks a certified measure, tunes pipelines
Feeds policies at the point of access, such as masking sensitive columns
Supplies grounded context to AI assistants and agents at run time
So whatPlan the move in steps: catalogue first, then alerts on critical lineage, then policy enforcement, then context for agents.
Exhibit 12Six capabilities the strategy should name
Six capabilities the strategy should name
Decide which of these you need for your themes, and in what order. Few organisations need all six on day one.
01
Data catalogue
A searchable inventory of data assets with owners, descriptions, ratings and access paths.
02
Business glossary
Agreed terms and definitions, linked to the physical data that implements them.
03
Lineage
The path from source system through every transformation to the report, model or agent that uses it.
04
Semantic consistency
One meaning for a concept such as customer or margin, reconciled across systems and reports.
05
Rules management
Quality, privacy and retention rules recorded once, attached to data, and measured.
06
Active metadata
Signals from usage, quality and change that trigger alerts, policies and automation.
Exhibit 13Tell a story a sponsor will fund
Tell a story a sponsor will fund
A metadata programme gets funded when it is framed as a business outcome with a named sponsor, not as a catalogue.
Business story
Metadata it relies on
Natural sponsors
Outcome to measure
Personalise the customer experience
Automated discovery of customer data, end-to-end lineage, modelled relationships between customer entities
Customer experience lead, CDO, CFO
Time to launch a personalised journey
Grow sales with predictive analytics
A common semantic view across sources, reuse of master data definitions
Sales leadership, CDO, CIO
Forecast accuracy and campaign conversion
Reduce regulatory and conduct risk
A catalogue of reported data, lineage readable by business users, workflow evidence
CEO, CDO, data protection officer
Effort and elapsed time to answer a supervisory request
Make analysts more productive
A catalogue, lineage and managed rules for trusted data sets
Analytics centre of excellence, CDO, CFO
Time spent finding and validating data per project
So whatPick one or two stories that match your strategic themes and let them set the scope of the first release.
Exhibit 14Build the business case in seven moves
Build the business case in seven moves
A business case for metadata is a chain from a sponsor's goal to a number finance will sign. Skip a link and the case collapses at review.
1
Anchor in a sponsor's vision
Find the executive whose goal depends on trusted, traceable data, and agree whether the effort is top-down, bottom-up or both.
2
Choose business measures
Select the measures that sponsor already reports, not metadata activity counts.
3
Measure the baseline
Record today's value of each measure, how it was measured and by whom. Without it, no improvement can be claimed.
4
Agree the target
Negotiate a realistic improvement with the sponsor and the people who run the process.
5
Translate into money
Convert the improvement into revenue gained, cost avoided, risk reduced or capital released, using finance's assumptions.
6
Cost the whole thing
Model total cost of ownership: software, people, stewardship time, integration, training and run costs over several years.
7
Show the return and payback
Compare value with cost, show when it pays back, and state the assumptions you will test each quarter.
So whatBuild the case with the sponsor and finance in the room; a case written by the data team alone is rarely believed.
Exhibit 15Leading and lagging measures for metadata
Leading and lagging measures for metadata
Lagging business measures prove the value; leading measures tell you early whether you will get there.
Horizon
Measure
What it tells you
Lagging
Revenue or cost impact attributed to faster, trusted insight
Whether the programme paid for itself
Lagging
Effort to answer audit and supervisory requests
Whether lineage and definitions reduced risk work
Leading
Time to complete the data selection phase of an analytics project
Whether people find and trust data faster
Leading
Reuse of certified data sets and measures
Whether teams stop rebuilding the same data
Leading
Critical measures with a certified definition and lineage
Whether the scorecard is becoming traceable
Leading
Active users and searches in the catalogue and glossary
Whether adoption is real or fading
Exhibit 16Map stakeholders by influence and support
Map stakeholders by influence and support
Metadata touches every function, so expect both champions and blockers. Place each stakeholder, then plan a move for each box.
Low influenceHigh influence →
Influential opponents
Meet early, understand the objection, and show how the programme reduces a risk or cost they own.
Influential supporters
Make them sponsors: give them the story, the scorecard and a visible role in quarterly reviews.
Weak opponents
Keep informed and address concerns through their managers; do not let them set the agenda.
Weak supporters
Turn them into stewards and champions, and give them the recognition and time to contribute.
OpposesSupports →
So whatSpend most effort moving influential opponents toward neutral; a single senior blocker can stall the whole programme.
Exhibit 17Metadata business case calculator
Metadata business case calculator
Estimate the value of time recovered by analysts and in audit work, against the cost of the programme. Replace every default with your own baseline.
Exhibit 18Metadata strategy in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Plants describe the same part, defect and downtime reason differently, so quality and delivery measures cannot be compared across the five sites.
A shared glossary for part, defect and downtime codes, linked to each plant's systems, with lineage from shop floor to the delivery scorecard.
Plant measures on the scorecard with one certified definition
Retail bank
The supervisory review found risk reports could not be traced to source or reconciled quickly.
Business-readable lineage and certified definitions for every regulatory risk measure, with owners and quality rules attached.
Elapsed time to answer a supervisory data request
Retailer
Stores and online plan stock separately and each team defines sell-through and margin its own way.
One semantic definition of product, stock and margin across channels, published in the catalogue and reused by pricing and range teams.
Reuse of certified product and stock data sets
Hospital group
A privacy incident and a coding audit exposed unclear ownership and sensitivity of patient data.
Sensitivity classification and ownership recorded for patient data, with lineage behind coded activity and access policies driven by metadata.
Patient data sets with owner, classification and access policy
Telecom operator
Mobile, broadband and business run on separate billing and CRM stacks, and network inventory drifts from the field.
A catalogue across the three stacks with a common customer and service definition, and change alerts when inventory records drift.
Time for an analyst to assemble a single customer view
Insurer
Regulatory and actuarial reporting relies on manual reconciliations between policy, claims and finance systems.
Lineage from policy and claims systems to reserving and regulatory measures, with reconciliation rules recorded once and monitored.
Manual reconciliation effort per reporting cycle
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Certified critical measuresLead Share of scorecard measures with an approved definition, owner and lineage to source
Lineage coverageLead Critical reports and models traceable end to end, source to output
Time to find trusted dataLead Median time for an analyst to locate and validate the data for a new question
Catalogue adoptionLead Active monthly users and searches as a share of the target audience
Certified asset reuseLead Number of downstream uses per certified data set or measure
Breaking changes caughtLead Schema or logic changes flagged before they reached a certified report
Audit request effortLag Hours and elapsed days to answer an audit or supervisory data request
Realised valueLag Benefit confirmed with finance against the business case baseline
Traps to avoid
01
Catalogue as the goal
A tool is bought and populated, adoption peaks at launch and then fades.
Hover for the counter-move
Counter-move
Tie every release to a sponsor's business story and a measure they already report.
02
Boiling the ocean
The team tries to document every table before anyone gets value.
Hover for the counter-move
Counter-move
Start with the critical measures on the scorecard and the data behind them; grow by demand.
03
Meaning left to IT
Technical metadata is rich but business definitions are missing or disputed.
Hover for the counter-move
Counter-move
Give business owners accountability for definitions, with stewards and a fast approval path.
04
No baseline
The programme claims benefits nobody can verify, and funding is cut at the next review.
Hover for the counter-move
Counter-move
Measure the baseline before go-live and report against it every quarter with finance.
05
Stale metadata
Documentation drifts from reality and users stop trusting the catalogue.
Hover for the counter-move
Counter-move
Harvest technical and operational metadata automatically and alert owners on drift.
06
AI without context
Assistants answer from raw tables and give confident, wrong numbers.
Hover for the counter-move
Counter-move
Ground AI in certified definitions, lineage and policies, and log the context used for each answer.
The first 90 days
Days 1-30
Anchor and baseline
Agree the sponsor and the one or two business stories that set the scope
List the critical measures on the data scorecard and their current owners
Measure the baseline for the chosen business measures and for time to find data
Map stakeholders by influence and support
Days 31-60
Prove it on the critical path
Harvest technical and operational metadata from the systems behind the critical measures
Certify definitions for those measures with business owners
Publish end-to-end lineage for the first critical reports
Complete the business case with finance, including total cost of ownership
Days 61-90
Make it active and report
Switch on alerts for breaking changes to certified measures
Open the catalogue and glossary to the first analyst community
Report leading measures and early wins to the sponsor
Agree the next domains to cover, ranked by business value
Where SCIKIQ fits
Enterprise Discovery Engine
Harvests technical and business metadata, classifies data automatically, traces lineage including SQL lineage from views and procedures, and records usage, owner and criticality.
Semantic and Ontology Construction
Holds the taxonomy, glossary terms and business definitions, concepts and rules in a governed editor with maker-checker approval and provenance on every object.
Metrics and Knowledge Fabric
Parses BI reports into a governed metrics layer with formulas, definitions, lineage from BI to metric and report usage, reconciled across systems.
Context Engine and Governance Gate
Assemble context packs of data, rules, metrics and policies for AI and agents, and enforce classification, access, audit and lineage before anything acts.
Track 2 of 8 · Stage 2: Anchor
Value proposition
Data can play three different roles for an organisation: a dependable foundation, an enabler of better processes, or an engine for new growth. Deciding the mix is the first strategic choice, because it changes what you build, how you govern it and how you prove it worked.
Most data strategies fail quietly because nobody decided what data is for. The platform team optimises for uptime, the business wants faster answers and the board expects new revenue, and each judges the strategy by a different yardstick. Naming the value proposition, and its mix, gives architecture, funding, governance and measures a single reference point. It also sets expectations: a foundation strategy is judged on reliability, a growth strategy on experiments that turn into products.
So whatWrite down the mix of foundation, enabler and growth engine explicitly, then make every architecture, funding and measurement choice consistent with it.
Questions this track answers
Which role does data play for us today, and which role does the business strategy need it to play?
What mix of foundation, enabler and growth engine do we commit to for the next three years?
How does that mix change our architecture, governance, organisation and funding choices?
Which measures prove value for each role, so we stop judging a foundation on revenue or a growth bet on uptime?
Which strategic goals does each data initiative serve, and through which role?
Exhibit 19Three roles data can play
Three roles data can play
Every organisation uses all three to some degree. The strategy decides which one leads.
01
Foundation
Data as a dependable shared service the business runs on: available, fast to reach, quick to extend with a new source or interface. Success is invisible: nobody waits and nothing breaks.
02
Enabler
Data that makes existing processes work better: sharper targeting and conversion, lower operating cost, earlier fraud detection, maintenance before failure. Success shows in the process measures the business already owns.
03
Growth engine
Data that creates something new: products, services, pricing models or revenue streams that did not exist before. Success is a pipeline of experiments, a fair share of which become repeatable offers.
So whatName the leading role first; the other two become supporting roles with smaller budgets and clear limits.
Exhibit 20The mix changes every other choice
The mix changes every other choice
The same capability is built differently depending on which role leads.
Dimension
Foundation
Enabler
Growth engine
Architecture
Standardised, resilient, reusable platform; strong integration and service levels
Data products shaped around specific processes; embedded in operational systems
Flexible sandboxes, fast access to new and external data, easy promotion to production
Governance
Central standards, certified definitions, strict change control
Fit-for-purpose quality set by each process owner, shared core definitions
Light guardrails for exploration, tighter gates when an idea goes live
Organisation
Strong central platform team, service management discipline
Data people embedded in business teams, hub-and-spoke support
Cross-functional product teams with a product manager and commercial owner
Funding
Run budget, priced like a utility service
Business co-funding tied to process outcomes
Venture-style tranches released at stage gates
Skills
Engineering, operations, reliability
Domain analysis, process redesign, change management
Data science, experimentation, product and commercial skills
Measures
Availability, time to access, time to onboard a source
Conversion, cost per transaction, loss avoided, process cycle time
Experiments run, share that repeat, revenue from new data-led offers
So whatIf your architecture, funding and measures point at different roles, the strategy will pull itself apart.
Exhibit 21Where your current initiatives sit
Where your current initiatives sit
Plot today's portfolio by how established the process is and how new the value is.
Shared serviceSpecific business outcome →
Enabler
Process-specific improvements: pricing, fraud, maintenance, targeting. Owned by the process owner.
Growth engine
New offers built on data, owned by a commercial lead and funded in tranches.
Foundation
Shared platforms, integration, quality and access that every other initiative relies on.
Platform for growth
New shared capabilities, such as external data or partner interfaces, that open future offers. Fund deliberately and sparingly.
Improves what existsCreates something new →
So whatA portfolio crowded in one quadrant tells you the mix you actually run, whatever the strategy says.
Exhibit 22Link goals to initiatives to measures
Link goals to initiatives to measures
Every initiative should trace to a strategic goal through one named role.
1
Start from the goal
Take each strategic goal from the business strategy, in the business's own words.
2
Name the role
Decide whether data serves that goal as foundation, enabler or growth engine.
3
Attach initiatives
List the data initiatives that serve the goal; reject any that cannot name a goal.
4
Choose the measures
Pick measures that fit the role: service measures for foundation, process measures for enabler, pipeline and revenue measures for growth.
5
Assign an owner
One business owner per goal and role combination, accountable for the measures.
So whatAn initiative with no goal or no role is a cost, not a strategy; stop or reframe it.
Exhibit 23Example measures by role
Example measures by role
Role
Leading measures
Lagging measures
Foundation
Time to onboard a new source; time from request to access; share of requests self-served
Platform availability; business hours lost to data outages; cost to serve per data product
Enabler
Adoption of the data product in the process; decisions made with it; data freshness
Conversion gained; operating cost removed; losses avoided; cycle time cut
Growth engine
Experiments started; time from idea to first test; share of experiments repeated
Revenue from new data-led offers; customers using them; margin on them
So whatJudge each initiative by the measures of its role, never by another role's yardstick.
Exhibit 24Signs the mix is wrong
Signs the mix is wrong
Too much foundation
Excellent platform, few business decisions changed
Business units build their own shadow analytics
The data team is seen as a cost centre to be squeezed
vs
Too much growth engine
Many pilots, few reach production
Every experiment rebuilds its own data pipeline
Trust in reported numbers falls while innovation spend rises
So whatRebalance at the annual refresh: growth bets need a trusted foundation, and a foundation needs bets to justify it.
Exhibit 25Which role should lead your data strategy?
Which role should lead your data strategy?
Answer six questions about your business strategy. The result is a starting point for the leadership debate, not a verdict.
Exhibit 26Value proposition in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Margins squeezed by input costs and price-downs, with a large installed base of parts in the field.
Lead as enabler on pricing and plant performance, with a growth bet on service revenue from the installed base.
Margin recovered on priced quotes; service contracts sold from installed-base data.
Retail bank
Supervisory findings on risk data aggregation and reporting; growth must come from existing customers.
Lead with foundation: certified risk data and lineage first, then enabler use cases on primary relationships.
Regulatory reports produced from certified data; products held per primary customer.
High churn, evenly spread network capital, separate billing and CRM per line of business.
Lead as enabler on churn and network investment, with a growth engine for business customers.
Churn among targeted segments; return on network capital; business revenue from new offers.
Insurer
Combined ratio above target, claims leakage and fraud, brokers frustrated by slow quotes.
Lead as enabler on pricing, claims and fraud, on a foundation of trusted regulatory data.
Loss ratio by segment; leakage and fraud avoided; broker quote turnaround.
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Initiatives traced to a goalLead Share of funded data initiatives that name a strategic goal and a role
Time to onboard a sourceLead Working days from request to a new source being available and governed
Time to accessLead Working days from a user's request to usable access to existing data
Data product adoptionLead Share of target process decisions made using the intended data product
Experiment repeat rateLead Share of growth experiments that become repeatable offers or processes
Process value realisedLag Benefit confirmed by finance from enabler initiatives against baseline
Revenue from data-led offersLag Revenue from products and services that depend on data to exist
Platform availabilityLag Share of business hours the shared data services meet their service level
Traps to avoid
01
One yardstick for everything
Foundation work is judged on revenue and growth bets on uptime, so both look like failures.
Hover for the counter-move
Counter-move
Assign each initiative a role and measure it with that role's measures.
02
The unspoken mix
Different leaders assume different roles for data and judge the strategy accordingly.
Hover for the counter-move
Counter-move
Agree the mix in a leadership workshop and write it on the strategy on a page.
03
Growth without a foundation
Pilots multiply, each with its own pipeline, and trust in reported numbers falls.
Hover for the counter-move
Counter-move
Tie every growth bet to shared, governed data products and fund the foundation it needs.
04
Foundation as an end in itself
The platform is excellent but few decisions change, and funding is cut.
Hover for the counter-move
Counter-move
Pair every foundation investment with an enabler use case that shows value within a quarter.
05
Copying another organisation's mix
The strategy borrows a growth narrative the business strategy does not support.
Hover for the counter-move
Counter-move
Derive the mix from your own business goals, regulation and trust in today's data.
The first 90 days
Days 1-30
Diagnose the mix you run today
Map current initiatives to foundation, enabler and growth engine
Interview theme owners on what they expect data to deliver
Note where measures and roles do not match
Days 31-60
Agree the target mix
Run a leadership workshop with the picker results as the starting point
Agree the leading role and the limits on the others
Link each strategic goal to initiatives through a named role
Days 61-90
Make it operational
Set role-appropriate measures and baselines for each initiative
Align funding models to roles: run, co-funded, staged
Publish the mix on the strategy on a page and schedule the annual review
Where SCIKIQ fits
Logical Enterprise Model
Maps value chains and processes to data domains and decisions, so each initiative can be traced to the goal and process it serves.
Metrics and Knowledge Fabric
Holds certified measures with definitions, so foundation, enabler and growth measures are defined once and reported consistently.
Virtual Data Fabric
Connects and federates sources in place, shortening time to access and time to onboard a source for foundation work.
Outcome and Observability
Tracks outcomes of decisions and actions, giving enabler and growth initiatives evidence of realised value.
Track 3 of 8 · Stage 1: Baseline
Capabilities and gaps
Assess each data capability twice: where it is today, and where the chosen value proposition actually needs it to be. The gaps that matter become the roadmap; the rest are deliberately left alone.
Maturity assessments usually produce a long list of weaknesses and an implied instruction to fix them all. That spreads money thin and ignores the strategy. Comparing current capability with the level the strategy requires turns a maturity score into a set of choices: invest where a gap blocks a strategic goal, hold where the capability is good enough, and stop over-investing where it already exceeds what is needed.
So whatFix only the gaps the chosen value proposition depends on, and write down the gaps you are deliberately leaving open.
Questions this track answers
Which capabilities does our chosen value proposition depend on most?
Where is each capability today, on evidence rather than opinion?
What level does the strategy actually need for each, not the highest possible level?
Which gaps block a strategic goal now, and which can wait?
Where are we over-invested relative to what the strategy needs?
Exhibit 27From assessment to roadmap
From assessment to roadmap
Five steps turn a maturity score into strategic choices.
1
Choose the capabilities
Use a short, stable list that covers strategy, governance, meaning, quality, integration, analytics, people and operating model.
2
Score today on evidence
Rate each capability on a five-level scale, citing artefacts and examples, not impressions.
3
Set the needed level
For each capability, decide the level the value proposition requires within the strategy horizon.
4
Classify the gap
Mark each capability as ahead of need, on target, or short of need.
5
Prioritise and sequence
Turn only the gaps that block strategic goals into roadmap items with owners and measures.
So whatThe needed level is a strategic choice; debate it with theme owners, not just the data team.
Exhibit 28Three states, three responses
Three states, three responses
01
Ahead of need
Capability exceeds what the strategy requires. Hold investment steady or redeploy effort to a gap.
02
On target
Capability matches the need. Sustain it, monitor it, and avoid gold-plating.
03
Short of need
Capability falls below what the strategy requires. Fix it if a strategic goal depends on it; otherwise record it as an accepted gap.
So whatReporting all three states keeps the conversation balanced and makes over-investment visible.
Exhibit 29The needed level depends on the value proposition
The needed level depends on the value proposition
Illustrative emphasis: the same capability can need a different level under each role.
Capability
Foundation leads
Enabler leads
Growth engine leads
Governance and trust
High: certified and controlled
Medium: core shared, rest fit for purpose
Medium: guardrails, tight at go-live
Metadata and meaning
High: full lineage and glossary
Medium: around priority processes
Medium: discoverability first
Data quality
High across reported data
High in the processes served
Variable: good enough to test
Integration and architecture
High: resilient and standard
Medium: process-shaped products
High: fast access to new data
Analytics and AI
Low to medium
Medium to high
High
People and literacy
Engineering and operations
Domain and process skills
Data science and product skills
So whatDo not set every capability to the top level; the strategy needs a profile, not a perfect score.
Exhibit 30Prioritise the gaps
Prioritise the gaps
Plot each gap by how much the strategy depends on it and how large it is.
Strategy depends on it a littleStrategy depends on it a lot →
Quick fixes
Close these early; they unblock strategic goals at modest cost.
Strategic programmes
Fund as roadmap programmes with owners, stage gates and measures.
Monitor
Accept for now; review at the next refresh.
Accepted gaps
Record the decision not to fix and the risk it carries; revisit if the strategy changes.
Small gapLarge gap →
So whatWriting down the accepted gaps is as important as funding the strategic programmes.
Exhibit 31Scoring on evidence
Scoring on evidence
Evidence that counts
Artefacts in use: policies, definitions, lineage, service levels
Decisions changed by the capability, with examples
Independent views from business users
vs
Evidence that does not
Tools purchased but not adopted
Plans and roadmaps not yet delivered
The data team's own optimism
One strong team generalised to the whole organisation
So whatScore the weakest common practice, not the best example, or the gap will be hidden.
Exhibit 32Capability gap map
Capability gap map
Set where each capability is today and the level your strategy needs. Gaps that matter rise to the top. Defaults are placeholders: replace them with your own evidence.
Exhibit 33Capabilities and gaps in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Strong plant systems but weak shared product and customer definitions across five plants.
Close the gap in metadata and quality on product and price data; hold plant integration at its current level.
Quotes priced from certified product and cost data.
Retail bank
Supervisory findings expose weak lineage and quality on risk data.
Treat governance, metadata and quality as strategic programmes; accept a gap in advanced analytics for a year.
Risk reports with full lineage and passing quality rules.
Retailer
Advanced analytics talent, but stock data split between stores and online.
Fix integration of one stock pool first; analytics capability is already ahead of need.
Share of stock visible and sellable across channels.
Hospital group
Privacy incident and coding audit reveal governance and quality gaps.
Raise governance, quality and literacy for clinical coders and managers; defer predictive work.
Coding accuracy and privacy incidents per quarter.
Telecom operator
Good network analytics, separate billing and CRM per line of business.
Close the gap in a shared customer view and metadata; hold network analytics steady.
Customers with a single view across mobile, broadband and business.
Insurer
Strong pricing models, weak claims data quality and slow broker data flows.
Prioritise quality and integration in claims and broker data; models are already ahead of the data feeding them.
Claims records passing quality rules; broker quote turnaround.
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Capabilities assessed on evidenceLead Share of capabilities scored with cited artefacts and measured results
Strategic gaps with ownersLead Share of strategy-blocking gaps with a named owner and roadmap item
Gap closure progressLead Average movement toward the needed level on prioritised capabilities, per quarter
Accepted gaps recordedLead Share of unaddressed gaps with a written decision and risk note
Over-investment flaggedLead Number of capabilities ahead of need with spend reviewed
Goals unblockedLag Strategic goals whose blocking capability gaps have closed
Capability profile metLag Share of capabilities at or above the needed level at the annual refresh
Traps to avoid
01
Fix everything
Every weakness becomes a project and money is spread too thin to move anything.
Hover for the counter-move
Counter-move
Fund only the gaps that block strategic goals and record the rest as accepted.
02
Top level everywhere
Every needed level is set to the maximum, so the gaps look enormous and arbitrary.
Hover for the counter-move
Counter-move
Set the needed level from the value proposition, capability by capability.
03
Scoring by opinion
Scores reflect who filled in the survey, not what is in use.
Hover for the counter-move
Counter-move
Require evidence for each score and calibrate across business units.
04
Ignoring over-investment
Capabilities ahead of need keep absorbing budget.
Hover for the counter-move
Counter-move
Report the ahead-of-need state and redeploy effort to the gaps that matter.
05
A one-off exercise
The assessment is filed and never repeated, so progress is unknown.
Hover for the counter-move
Counter-move
Repeat it at each annual refresh and show movement on the scorecard.
The first 90 days
Days 1-30
Assess on evidence
Agree the capability list and the five-level scale
Collect evidence and score current levels with business input
Calibrate scores across units in one session
Days 31-60
Set the needed profile
Set needed levels from the chosen value proposition
Classify each capability as ahead, on target or short
Plot gaps by strategic dependence and size
Days 61-90
Build the gap roadmap
Turn strategy-blocking gaps into roadmap items with owners
Record accepted gaps with their risks
Add gap closure measures to the data scorecard
Where SCIKIQ fits
Enterprise Discovery Engine
Profiles and maps the existing estate, giving an evidence base for current capability scores rather than survey opinion.
Semantic and Ontology Construction
Builds the shared glossary, concepts and business rules that close gaps in metadata and meaning.
Governance Gate
Applies policies and approvals consistently, raising governance and trust capability without a manual process for each team.
Outcome and Observability
Measures quality, usage and outcomes over time, so gap closure is tracked on evidence.
Track 4 of 8 · Stage 4: Design
Architecture and principles
Govern every data initiative with a small set of stable principles, then shape the architecture around how well the data and the questions are already understood.
Roadmaps change every quarter; principles should not. A short, agreed set of principles lets many teams build in different places and still end up with one coherent estate. Architecture that is chosen use case by use case, without them, ends in duplicated pipelines, competing copies of the truth and a platform bill nobody can explain. The principles and the target state belong in the strategy because they are how the strategy is enforced between reviews.
So whatAgree eight to ten principles and a one-page target state before choosing tools, and test every new initiative against both at the stage gate.
Questions this track answers
Which principles will every data initiative be held to, whoever runs it?
How well understood are the data and the questions behind each priority use case?
Which integration, storage and analytics styles fit each kind of work?
What must the target-state architecture contain, and what is deliberately left out?
How does our value proposition change where the architecture puts its effort?
Exhibit 34Nine principles that govern the data estate
Nine principles that govern the data estate
Our starting set. Keep the ones that fit, reverse the ones that do not, and write down why.
01
Decisions first
Every dataset, pipeline and model exists to improve a named decision; if none can be named, it is not built.
02
Enterprise asset, domain owned
Data belongs to the organisation, and a named business domain is accountable for each data product.
03
Open by default, protected by design
Data is discoverable and shareable unless a stated reason says otherwise; classification and access rules travel with it.
04
One meaning per term
Shared business terms, entities and metrics are defined once and reused everywhere they appear.
05
Fit for its purpose
Quality is set by the decision the data serves, published as a score, and checked where the data is produced.
06
Reuse before rebuild
New use cases start from certified data products; a new copy needs a reason the architecture board accepts.
07
Connect before you move
Query data where it lives when that is good enough; copy or materialise it only for performance, history or isolation.
08
Traceable end to end
Every published number can be traced to its sources, its transformations and its owner.
09
Responsible use
Lawful is the floor, not the bar: sensitive and automated uses pass an ethics and privacy check before release.
So whatPrinciples only govern if the stage gate checks them; add a principles review to every funding decision.
Exhibit 35Four zones: how well do we know the data and the question?
Four zones: how well do we know the data and the question?
Place each use case by asking two things: is the data understood and curated, and is the question already defined?
Data new or uncuratedData understood and curated →
Explain
Trusted data, new questions: analysts investigate why. Governed preparation, a searchable catalog, virtual views over the warehouse and visual exploration.
Run
Trusted data, known questions: the reporting and operational core. Batch and replicated integration, the warehouse, real-time feeds where latency matters, standard and embedded reporting.
Discover
New data, open questions: experiments and data science. Raw landing in a lake, self-service preparation, virtualisation for quick prototypes, notebooks and augmented analytics.
Expand
New data, known questions: adding sources to answer what is already asked. Streaming and new connectors, virtualisation to test value before building, then promotion into the core.
Question still openQuestion defined →
So whatDo not hold discovery work to run-zone controls, or run-zone work to discovery freedoms; design both, and a path between them.
Exhibit 36Which styles fit each zone
Which styles fit each zone
Zone
Integration
Storage and management
Access and analytics
Run
Scheduled batch, replication, change capture
Warehouse, real-time store, master data
Standard reports, dashboards, embedded analytics
Explain
Virtual views over curated data, governed preparation
Warehouse plus semantic layer
Visual exploration, natural-language query
Expand
New connectors, streaming, virtualisation to test value
Staging area, then promotion to the core
Prototype reports, then the standard set
Discover
Self-service preparation, raw ingestion
Lake or lakehouse sandbox
Notebooks, augmented analytics, model training
So whatMost estates need all four zones; the strategy decides how much of each, and who may move work between them.
Exhibit 37The path from discovery to the core
The path from discovery to the core
Work moves zone to zone as it proves its worth; each move has an entry test.
1
Prove it in Discover
A hypothesis, a sandbox, a time box and a business sponsor.
2
Promote to Expand
The value is shown and the sources are named, with owners and a first quality check.
3
Certify into Run
Definitions in the glossary, lineage captured, quality rules live and a service level agreed.
4
Retire what is not used
Reports, pipelines and copies with no recent use are switched off on a schedule.
Exhibit 38Target-state architecture checklist
Target-state architecture checklist
A one-page target state must answer each of these.
Area
What to capture
Test of done
Sources
Internal systems by domain, and external data with its licence terms
Every critical source has an owner and a criticality rating
Integration styles
Batch, replication, streaming, virtualisation and APIs, and when each is used
One written rule per style, not one per project
Management styles
Warehouse, lake or lakehouse, semantic layer and data products
Each data product has a home and a service level
Access styles
Portal, analyst workbench, data science lab, embedded analytics, APIs to other systems
Each persona has a named way in
Master and reference data
Golden records for core entities and shared code lists
One agreed customer, product and supplier record
Quality and observability
Rules at source, published scores, pipeline monitoring
Scores visible to the people who use the data
Security and privacy
Classification, access control, masking, consent and retention
Policies enforced by the platform, not by habit
Metadata
Catalog, glossary, lineage and usage, kept current automatically
Any published number traceable to source
Exhibit 39Your value proposition changes the emphasis
Your value proposition changes the emphasis
Value proposition
Architecture puts its effort into
Principle that dominates
Foundation: data the business runs on
Availability, cost, speed of onboarding new sources, service levels
Reuse before rebuild
Enabler: data that improves existing processes
Trusted definitions, quality at source, embedded analytics in workflows
Fit for its purpose
Growth engine: data that creates new offers
Sandboxes, fast experimentation, external data, APIs and data sharing
Connect before you move
So whatIf the architecture effort does not match the value proposition the executives chose, one of the two is wrong.
Exhibit 40Zone finder: where does this use case belong?
Zone finder: where does this use case belong?
Answer five questions about one use case; the finder suggests its zone and the patterns to start from.
Exhibit 41Architecture and principles in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Plant historians and quality data sit in each plant; field data on the installed base is unseen.
Run zone for delivery and quality reporting across plants; a Discover sandbox for installed-base and sensor analytics.
Share of plants reporting from one certified quality data product
Retail bank
Risk reports are rebuilt by hand from several ledgers, and lineage cannot be shown to the supervisor.
Principles of traceability and one meaning per term enforced first; regulatory metrics moved into the certified Run zone.
Share of regulatory metrics with end-to-end lineage
Retailer
Stores and online plan stock separately; pricing experiments run on ad hoc extracts.
One product and stock data product in the Run zone; pricing and range tests run in Discover with a promotion path.
Time from proven pricing test to production use
Hospital group
Patient-flow data is spread across hospital systems, and a privacy incident has raised the bar.
Protected-by-design principle applied first; bed and discharge data virtualised into an Expand view before building the core.
Share of sensitive data assets classified with access policies enforced
Telecom operator
Mobile, broadband and business run on separate billing and CRM stacks; inventory drifts from the field.
Connect before you move: a virtual customer view across the three stacks, then a mastered customer record in the Run zone.
Share of customers with one resolved identity across products
Insurer
Actuarial and regulatory reporting rely on manual reconciliations between policy and claims systems.
Certified policy and claims data products in the Run zone; fraud and leakage models built in Discover and promoted.
Manual reconciliation steps removed from the reporting cycle
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Principles compliance at gateLead Share of funded initiatives that passed the principles review
Source onboarding timeLead Median days from request to a governed, usable new source
Data product reuseLead Average number of use cases served by each certified data product
Duplicate pipelines retiredLead Redundant pipelines and copies switched off in the period
Promotion cycle timeLead Median days from proven experiment to certified production use
Lineage coverageLag Share of critical reports traceable to source
Platform cost per active userLag Run cost of the data platform divided by monthly active users
Traps to avoid
01
Principles on a poster
Agreed in a workshop, never checked again.
Hover for the counter-move
Counter-move
Make the principles a stage-gate criterion and log every exception with an owner and expiry.
02
Tool-led target state
The target architecture is a list of products.
Hover for the counter-move
Counter-move
Describe styles and capabilities first; products are chosen to fit them, not the reverse.
03
One zone for everything
Experiments are forced through production controls, or production runs on sandbox code.
Hover for the counter-move
Counter-move
Design the four zones explicitly, with an entry test for each move between them.
04
Copy by default
Every project builds its own extract, and nobody knows which copy is right.
Hover for the counter-move
Counter-move
Connect in place first and require a reason for each new copy.
05
No retirement plan
The estate only grows; cost and confusion grow with it.
Hover for the counter-move
Counter-move
Schedule retirement reviews and switch off unused reports and pipelines.
The first 90 days
Days 1-30
Draft and agree the principles
Inventory current architecture decisions and exceptions
Draft eight to ten principles with the architecture board
Test them against three live initiatives
Agree them at the executive council
Days 31-60
Map the zones and the target state
Place the priority use cases in the four zones
Draft the one-page target state against the checklist
Name owners for critical sources and data products
Days 61-90
Make them enforceable
Add the principles review to the stage gates
Publish the promotion path from Discover to Run
Start the first retirement review
Report principles compliance on the data scorecard
Where SCIKIQ fits
Virtual Data Fabric
Connects, federates and queries data where it lives, with caching and on-demand materialisation, so connect-before-you-move is the default.
Semantic Data Layer
Business objects, semantic models and natural language to SQL give the Explain zone one meaning over curated data.
Enterprise Discovery Engine
Discovers technical and business metadata, classification and lineage across sources, keeping the target state honest.
Governance Gate
Classification, access control, masking and policy checks enforce the protected-by-design principle in the platform.
Track 5 of 8 · Stage 4: Design
Governance platform
Governance capabilities that once lived in separate tools are converging into platforms. Decide which policies matter most, what each persona needs, and which tools to consolidate.
Governance fails when it lives in documents, spreadsheets and half a dozen disconnected tools: policies are written but not enforced, and nobody can see whether they work. A data strategy that commits to trusted data must say how governance will be operated day to day, and on what. Choosing the platform from the policy priorities and the people who use it, rather than from feature lists, keeps governance tied to the strategy.
So whatPrioritise two or three policy types, write the requirements by persona, and consolidate governance tooling around them rather than buying another point tool.
Questions this track answers
Which policy types matter most to our strategy: quality, security, privacy, ethics, retention or access?
What does each persona need from governance to do their job?
Which governance capabilities do we have, and in how many separate tools?
Where are policies written but not enforced or monitored?
What would we consolidate first, and what would it free up?
Exhibit 42From scattered tools to one governance platform
From scattered tools to one governance platform
Scattered today
Policies in documents nobody can find
A catalog here, a quality tool there, access rules in each system
Stewardship by email and spreadsheet
No single view of whether policies are working
Each tool with its own glossary
vs
Converged target
Policies authored once, linked to the data they govern
Discovery, classification, lineage and quality on shared metadata
Stewardship workflow with tasks, approvals and escalations
Monitoring and scores reported to the governance forums
One glossary used by every capability
So whatConsolidation is a strategy decision, not a procurement one: it changes who does governance work and how it is seen.
Exhibit 43Six policy types to prioritise
Six policy types to prioritise
Few organisations can lead on all six at once. Choose two or three from the strategy and posture.
Policy type
What it governs
Lead when
Quality
Rules, thresholds and remediation for critical data
Reported numbers are disputed or decisions are delayed by checking
Security
Classification, protection and handling of sensitive data
Sensitive data is widely copied or breach exposure is high
Privacy
Consent, purpose, minimisation and data subject rights
Customer or patient data drives growth plans
Ethics and responsible AI
Fairness, transparency and human oversight of automated decisions
Models or agents are starting to act on decisions
Retention
How long data is kept, archived and destroyed
Storage cost, legal holds or discovery requests are rising
Access
Who may see and use what, for which purpose
Self-service is growing faster than access reviews
So whatThe posture calculator points the way: defence-led strategies lead on quality, security and privacy; offence-led ones cannot skip access and ethics.
Exhibit 44Requirements by persona
Requirements by persona
Write the checklist from the people who will use the platform, then score options against it.
Persona
Needs from governance
Sign it is working
Executive
A view of policy health and risk on the data that matters
Governance is reported on the data scorecard
Data owner
Clear accountability, approval requests and quality scores for their domain
Owners approve access and changes in the platform
Data steward
Task queues, issue workflow, rules authoring and glossary editing
Issues are closed within agreed times
Data engineer
Policies as code, lineage captured automatically, quality checks in pipelines
Pipelines fail fast on policy breaches
Analyst
Find trusted data, see its meaning, quality and permitted use
Analysts start from the catalog, not from a colleague
Risk and privacy officer
Classification, consent, retention evidence and audit trails
Audit evidence is produced from the platform, not assembled by hand
Exhibit 45The capabilities a governance platform brings together
The capabilities a governance platform brings together
Capability
What it does
Linked to
Policy authoring
Write policies once, in business language, with scope and owner
Every other capability
Workflow and stewardship
Tasks, approvals, issue management and escalation
Owners, stewards, forums
Discovery and classification
Find data across systems and tag sensitivity automatically
Security, privacy, access
Lineage
Trace data from source to report, model and action
Quality, impact analysis, audit
Monitoring and scoring
Measure policy compliance and quality over time
Scorecard, forums
Reporting
Policy health, open issues and trends by domain
Executive council, domain councils
Automation
Enforce policies in pipelines, queries and agent actions
Engineering, AI governance
Exhibit 46How to consolidate
How to consolidate
1
Inventory
List every governance tool, spreadsheet and manual process, and what each covers.
2
Prioritise policies
Pick the two or three policy types the strategy depends on.
3
Write persona requirements
One checklist per persona, weighted by the priority policies.
4
Score options
Score the current estate and candidate platforms against the checklist.
5
Migrate by domain
Move one domain at a time, starting where trust problems cost most.
6
Switch off
Retire the old tools and spreadsheets as each domain moves.
So whatRetiring tools is part of the plan, not an afterthought; otherwise consolidation adds one more tool.
Exhibit 47Governance capability gap
Governance capability gap
Set where each capability is today and where the strategy needs it. Red bars are the deficits to put on the roadmap.
Exhibit 48Governance platform in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Quality and supplier data rules differ by plant and live in local spreadsheets.
Lead on quality policies; one stewardship workflow for product and supplier data across the five plants.
Share of critical product data elements with published quality scores
Retail bank
Supervisory findings on risk data aggregation; evidence is assembled by hand each quarter.
Lead on quality, lineage and access; consolidate governance tooling so audit evidence comes from the platform.
Days to produce supervisory evidence for a risk metric
Retailer
Loyalty data drives personalisation, but consent is recorded differently online and in store.
Lead on privacy and access: one consent policy enforced wherever customer data is used.
Share of personalised campaigns checked against recorded consent
Hospital group
A privacy incident and a coding audit put data governance on the board agenda.
Lead on privacy, security and quality; classify patient data first and report policy health to the board.
Share of patient data assets classified with enforced access policies
Telecom operator
Consent records are incomplete across three billing and CRM stacks.
Lead on privacy and retention; resolve consent per customer and enforce it in every campaign and model.
Share of active customers with a complete, resolved consent record
Insurer
The conduct supervisor expects evidence of fair pricing and claims handling.
Lead on ethics and quality: oversight of pricing and fraud models, with lineage and decision audit trails.
Share of pricing and claims models with documented oversight and lineage
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Policies under managementLead Share of priority policies authored, owned and linked to data in the platform
Classification coverageLead Share of sensitive data assets classified automatically
Issue closure timeLead Median days to close a data issue in the stewardship workflow
Tools consolidatedLead Governance tools and spreadsheets retired against the plan
Policy compliance scoreLag Share of monitored checks passing, by domain and policy type
Audit evidence effortLag Person-days to produce evidence for a regulatory or internal audit
Trust in dataLag Share of surveyed users who trust the data they use for decisions
Traps to avoid
01
Feature-list buying
The platform is chosen on the longest feature list.
Hover for the counter-move
Counter-move
Score options against persona requirements weighted by the priority policies.
02
All policies at once
Six policy programmes start together and none finishes.
Hover for the counter-move
Counter-move
Lead with two or three policy types tied to the strategy; sequence the rest.
03
One more tool
A new platform is added but the old tools stay.
Hover for the counter-move
Counter-move
Plan retirement by domain and track tools consolidated on the scorecard.
04
Governance by documents
Policies exist but are never enforced or measured.
Hover for the counter-move
Counter-move
Link each policy to data, monitor compliance and report it to the forums.
05
Stewards without time
Stewardship is added to day jobs with no capacity.
Hover for the counter-move
Counter-move
Size stewardship effort per domain and agree it with the domain owner.
The first 90 days
Days 1-30
Inventory and prioritise
List governance tools, spreadsheets and manual processes
Choose the priority policy types with the executive council
Name owners and stewards for the first domain
Days 31-60
Requirements and scoring
Write persona requirement checklists
Score the current estate and candidate options
Agree the consolidation plan and the first domain
Days 61-90
First domain live
Move priority policies for one domain into the platform
Start the stewardship workflow and quality scores
Retire the first redundant tool
Report policy health at the monthly review
Where SCIKIQ fits
Governance Gate
Identity and access, classification, privacy, policy and compliance checks, approvals and audit before anything is executed.
Enterprise Discovery Engine
Automatic classification, lineage including SQL lineage, ownership and criticality across sources.
Semantic and Ontology Construction
A governed glossary and ontology editor with maker-checker approval and provenance on every object.
Outcome and Observability
Measures outcomes against baselines, so policy health and its effect can be reported to the forums.
Track 6 of 8 · Stage 6: Mobilise
DataOps
DataOps is a working practice that brings communication, integration, automation, observability and operational discipline to the flow of data between the people who produce it and the people who use it.
A strategy that is right on paper still fails if every new dataset takes a quarter to arrive and nobody trusts it when it does. DataOps is how the data team turns the roadmap into a steady flow of usable, decision-quality data. It belongs in the strategy because it sets the delivery speed and the level of trust that every other objective depends on.
So whatMeasure how long it takes to deliver a trusted dataset and how often it breaks, then fix the practices that move those two numbers before buying anything new.
Questions this track answers
How long does it take today to turn a request for data into a trusted, usable dataset?
How often do pipelines break or deliver wrong data, and who notices first?
Which software delivery habits can the data team adopt this year?
Who owns each data product end to end, from source to consumer?
Where should DataOps start so that it shows value quickly?
Exhibit 49The friction DataOps removes
The friction DataOps removes
Traditional delivery fails on both sides of the pipeline.
Producing trusted data
Hand-built pipelines that only one engineer understands
Changes deployed by hand, tested in production
Breaks discovered by the business, not by the team
Every request treated as a new project
vs
Consuming decision-quality data
Long queues before a new source or field arrives
No clear definition or owner for the data received
Reports that disagree, so meetings debate the numbers
Analysts rebuilding the same extracts in spreadsheets
So whatIf both columns sound familiar, the bottleneck is how data is delivered, not which platform delivers it.
Exhibit 50The outcomes leaders want but rarely measure
The outcomes leaders want but rarely measure
Outcome
What to measure
Why it matters
Speed
Time from request to a trusted dataset in use
Sets how fast the strategy can move
Throughput
Usable datasets and data products delivered per quarter
Shows capacity, not just effort
Quality
Data incidents reaching consumers
Trust is lost one wrong number at a time
Rework
Share of effort spent fixing or rebuilding
Rework is capacity the strategy never gets
Reuse
Consumers per data product
Reuse is where delivery pays back
So whatPick two of these, measure a baseline this month, and put them on the data scorecard.
Exhibit 51Lessons borrowed from software delivery
Lessons borrowed from software delivery
Software teams solved the same friction years ago. Most of the habits transfer directly.
01
Version everything
Pipeline code, transformations, schemas and tests live in version control with review.
02
Test automatically
Every change runs tests on structure, volume, freshness and business rules before it ships.
03
Deploy continuously
Small changes move through development, test and production by pipeline, not by hand.
04
Observe in production
Monitor freshness, volume, schema drift and quality, and alert the team before consumers notice.
05
Work in small batches
Ship a usable slice in weeks, then extend, rather than a large release after months.
06
Learn from incidents
Blameless reviews turn each break into a test or a control so it does not recur.
Exhibit 52Four foundations for introducing DataOps
Four foundations for introducing DataOps
1
Align to the value proposition
Run the trusted core with strict controls; let exploration move faster with lighter ones.
2
Align the roles
Agree who engineers, who stewards, who tests and who consumes, in each flow.
3
Build composable data products
Reusable products with contracts and service levels, assembled into many uses.
4
Organise in product teams
Cross-functional teams with a product owner who answers for value, not tickets.
So whatThe four foundations change how work is organised; tools only make them cheaper to run.
Exhibit 53An illustrative data product team
An illustrative data product team
Role
Accountable for
Sits in
Product owner
Value, priorities and consumer satisfaction
The business domain
Data engineers
Pipelines, tests and deployment
The product team
Analytics engineer
Models, metrics and semantic definitions
The product team
Data steward
Definitions, quality rules and access
The business domain
Platform engineer
Shared tooling, environments and observability
The central data team
Consumers
Feedback and acceptance of each slice
The business domain
So whatPlacing the team close to the business that uses its products shortens every feedback loop.
Exhibit 54Start small, with sponsorship
Start small, with sponsorship
01
One domain, one product
Choose a data product a sponsor cares about and that breaks often today.
02
Measure before and after
Baseline delivery time and incidents, so the improvement is visible.
03
Ride on other changes
Pair DataOps with literacy and governance work already in flight.
04
Scale the pattern
Turn what worked into templates, then spread team by team.
Exhibit 55DataOps practice gap
DataOps practice gap
Rate where each practice is today and where the strategy needs it. Deficits are listed in order, with a first fix.
Exhibit 56DataOps in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Plant quality and delivery data arrive late and differ by plant.
A product team for the delivery-performance data product, with automated tests on every plant feed.
Days from plant month-end to trusted delivery figures
Retail bank
Supervisory findings on risk data aggregation and manual report fixes.
Version-controlled, tested pipelines for regulatory metrics, with an incident log reviewed monthly.
Data incidents reaching regulatory reports
Retailer
Stores and online plan stock from separate, conflicting extracts.
One stock-position data product with a contract and freshness alerts, owned by merchandising.
Freshness of the stock position at trading start
Hospital group
Bed and discharge data are compiled by hand each morning.
An automated patient-flow pipeline with tests on admissions and discharge feeds.
Time from shift change to a trusted bed position
Telecom operator
Separate billing and CRM stacks break churn models whenever a field changes.
Data contracts on billing and CRM feeds, with schema-drift alerts to the owning teams.
Churn model runs lost to upstream changes
Insurer
Actuarial and regulatory reporting relies on manual reconciliations.
Automated reconciliation tests in the claims and policy pipelines, deployed by pipeline.
Analyst hours spent on manual reconciliation
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Lead time to trusted dataLead Days from an accepted request to a dataset in use with definitions and tests
Deployment frequencyLead Pipeline changes released to production per month
Automated test coverageLead Share of priority pipelines with automated quality tests
Contracted feedsLead Share of critical feeds covered by a data contract
Incidents reaching consumersLag Data breaks found by consumers rather than by monitoring
Time to restoreLag Hours from detection of a data break to a corrected dataset
Rework shareLag Share of team effort spent fixing or rebuilding existing pipelines
Data product reuseLag Average number of consuming use cases per data product
Traps to avoid
01
Tool first
A new orchestration tool is bought and the same manual habits move onto it.
Hover for the counter-move
Counter-move
Change the practices on one product first; choose tooling once the way of working is proven.
02
Big bang
DataOps is launched across every team at once and stalls in training.
Hover for the counter-move
Counter-move
Start with one domain and one sponsor, measure the gain and spread the template.
03
No consumer in the loop
Pipelines are faster but still deliver data nobody asked for.
Hover for the counter-move
Counter-move
Put consumers and a business product owner in the team and accept work slice by slice.
04
Speed without trust
Releases accelerate and incidents rise with them.
Hover for the counter-move
Counter-move
Gate releases on automated tests and track incidents alongside lead time.
05
Same rules everywhere
Exploration is slowed by controls designed for regulatory reporting.
Hover for the counter-move
Counter-move
Set control levels by value proposition: strict for the trusted core, lighter for experiments.
The first 90 days
Days 1-30
Baseline and choose
Measure lead time, incidents and rework on current delivery
Choose one priority data product and its sponsor
Name the product owner and form the team
Days 31-60
Put the practices in place
Move the product's pipelines into version control with review
Add automated tests and freshness alerts
Write the first data contracts with source owners
Start the incident log and review
Days 61-90
Prove and template
Automate deployment end to end for the product
Report before-and-after measures to the sponsor
Turn the setup into a template for the next two teams
Where SCIKIQ fits
Virtual Data Fabric
Connects and federates sources in place, so new data products start without building another copy pipeline.
Semantic Data Layer
Business objects and semantic models give every data product one shared definition for consumers.
Outcome and Observability
Measures outcomes and feeds learning back, so the team sees what changed and what it delivered.
Governance Gate
Checks that access and actions are compliant before they run, so faster delivery does not bypass control.
Track 7 of 8 · Stage 5: Sequence
Cloud data strategy
Moving data and analytics to the cloud changes cost, speed and control at once. Plan it as a sequence of deliberate choices about what moves, in which order, on which ecosystem, and how spend is governed.
Cloud decisions fix the platform, the cost model and much of the governance for years. Made as an infrastructure project, they move workloads without moving value. Made inside the data strategy, they follow the priorities on the roadmap, respect data gravity and come with financial governance from the first month.
So whatSequence the move by data gravity and value, choose the ecosystem style deliberately, and run financial governance from day one with price-performance as the measure.
Questions this track answers
Which workloads should move first, and which belong together?
Which cloud ecosystem style fits our commitments, skills and data spread?
How will we keep spend predictable when pricing is consumption-based?
Which on-premises habits must change, and who will change them?
What do regulation, sovereignty and latency rule in or out?
Exhibit 57Cloud is different
Cloud is different
Habits from the data centre carry over badly. Three differences matter most.
01
The implementation is hidden
You choose services, not hardware. Tuning means understanding how each service consumes resources.
02
Pricing is synthetic
Units, credits and tiers stand between your workload and its cost. The same query can cost very different amounts.
03
Old best practices must change
Sizing for the peak, keeping everything running and copying data freely all cost money every hour.
Exhibit 58Benefits and concerns, weighed together
Benefits and concerns, weighed together
What draws organisations in
Rapid provisioning and fast access to new capability
Elastic scale and paying for what is used
No capital outlay and fewer infrastructure specialists
Faster, more agile development
vs
What holds them back
Regulation, data sovereignty and security
Governance across many services and accounts
Latency and compatibility with systems that stay on site
Loss of control and fear of lock-in
So whatNew financial models and the cultural shift appear on both sides: they are the benefit if managed and the risk if not.
Exhibit 59Four hybrid and multicloud patterns
Four hybrid and multicloud patterns
01
Lifecycle split
Development, test or recovery run in the cloud while production stays where it is, or the reverse.
02
Use-case specific
Particular workloads, such as data science or a new data product, run in the cloud.
03
Spanning architecture
One architecture spans on-site and cloud, with data and processing placed where each fits best.
04
Intercloud
Workloads and data are spread across more than one cloud provider by design.
Exhibit 60Let data gravity set the migration order
Let data gravity set the migration order
Systems that exchange data constantly form clusters. Move one and the rest of its cluster follows.
1
Map the flows
Chart which applications and data stores exchange data, and how often.
2
Find the clusters
Group tightly entangled systems; these move together or not at all.
3
Score each cluster
Weigh value to the roadmap against effort, risk and the cost of splitting it.
4
Sequence by wave
Move low-entanglement, high-value clusters first; plan bridges for the rest.
So whatMigrating one system out of a tight cluster creates latency and transfer cost that erode the case for moving at all.
Exhibit 61Cloud data ecosystem styles
Cloud data ecosystem styles
Style
Strengths
Watch-outs
Provider-native
Deepest out-of-the-box integration; fastest potential time to value
Lock-in to one provider; harder to reach data in other clouds
Mixed native and independent
Weaker native services can be swapped for better ones; a common path to multicloud
More integration effort up front
Self-contained independent software
Easiest lift and shift; suits an existing software commitment; some portability
More work to use surrounding provider services; lock-in to the software supplier
Intercloud
Uses each provider's unique strengths; fits data already spread across clouds
Hardest to integrate and manage; complex cost governance; transfer fees and latency
So whatEvery style still rests on the provider's identity, billing and monitoring services, so integration with them is never optional.
Exhibit 62Pricing models by workload
Pricing models by workload
Match the pricing model to how predictable each workload is. Good fit, consider, or poor fit.
Workload
Resource-based
Consumption-based
Bounded or burst
Blended
Operational and transactional
Good
Poor
Consider
Good
Dashboards
Good
Consider
Good
Good
Standard reports
Good
Consider
Good
Good
Development and test
Consider
Good
Good
Consider
Exploratory
Poor
Good
Consider
Consider
Ad hoc
Poor
Good
Good
Consider
So whatSteady workloads favour reserved resources; unpredictable ones favour consumption with spending limits.
Exhibit 63Operate for scarcity in a world of abundance
Operate for scarcity in a world of abundance
01
Price-performance is the measure
Judge services by cost per unit of useful work, not by speed or list price alone.
02
Prove it before you commit
Run proofs of concept on real workloads to learn how each service behaves and bills.
03
Govern spend from day one
Budgets, tags, alerts and owners before the first workload, not after the first overrun.
04
Expect early overruns
Plan for learning costs in the first quarters and review spend weekly until it settles.
Exhibit 64Which cloud ecosystem style fits you?
Which cloud ecosystem style fits you?
Answer six questions. The style with the strongest fit is recommended, with what it implies for the strategy.
Exhibit 65Cloud data strategy in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Plant systems stay on site; analytics on field and quality data needs scale.
A spanning architecture: plant data stays local, quality and installed-base analytics move to the cloud.
Cost per quality-analytics run against the on-site baseline
Retail bank
Regulators expect control over where risk and customer data sit.
Use-case specific moves for analytics, with risk data aggregation kept under one governed design.
Regulatory data held in approved locations, evidenced
Retailer
Seasonal peaks swamp fixed capacity for pricing and stock analytics.
Consumption pricing with spending limits for peak analytics; reserved capacity for daily dashboards.
Analytics spend per trading week against budget
Hospital group
A privacy incident makes patient data placement a board concern.
Move non-identifiable planning and patient-flow analytics first; keep clinical records on site until controls are proven.
Workloads moved with privacy review passed
Telecom operator
Separate billing and CRM stacks are tightly entangled with network data.
Map data gravity first and move billing, CRM and churn analytics as one wave.
Cross-environment transfer cost per month
Insurer
Pricing models need burst compute; reporting runs on steady schedules.
Burst pricing for pricing-model runs, reserved capacity for actuarial and regulatory reporting.
Price-performance of pricing model runs
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Spend against budgetLag Monthly cloud data and analytics spend compared with plan, by owner
Price-performanceLag Cost per unit of useful work for each major workload, against baseline
Tagged spendLead Share of cloud spend tagged to an owner and a data product
Workloads moved by planLead Migration waves completed on the sequenced roadmap
Cross-environment transfer costLag Monthly cost of moving data between clouds and on site
Time to provisionLead Days from approved request to a working environment
Idle resource shareLead Share of provisioned capacity left unused
Traps to avoid
01
Lift and shift everything
Workloads move as they are and cost more than they did on site.
Hover for the counter-move
Counter-move
Re-plan each wave for cloud pricing and switch off what runs idle.
02
Breaking a cluster
One system moves and its entangled neighbours start paying latency and transfer fees.
Hover for the counter-move
Counter-move
Map data gravity and move tightly coupled systems in the same wave.
03
Governance after the bill
Spend control starts only after the first budget breach.
Hover for the counter-move
Counter-move
Set budgets, tags, alerts and owners before the first workload goes live.
04
Multicloud by accident
Teams pick providers independently and the estate fragments.
Hover for the counter-move
Counter-move
Choose the ecosystem style centrally and approve exceptions explicitly.
05
Speed as the only test
Services are chosen on benchmark speed and cost overruns follow.
Hover for the counter-move
Counter-move
Choose on price-performance from proofs of concept on real workloads.
The first 90 days
Days 1-30
Plan
Map data flows and entangled clusters
Agree regulatory, sovereignty and latency constraints
Set up financial governance: budgets, tags, alerts and owners
Days 31-60
Choose and prove
Choose the ecosystem style using the roadmap and constraints
Run proofs of concept on two real workloads
Pick pricing models per workload from the fit table
Days 61-90
Move the first wave
Migrate one low-entanglement, high-value cluster
Review spend weekly and tune
Report price-performance against the baseline to the sponsor
Where SCIKIQ fits
Virtual Data Fabric
Connects and federates data wherever it lives, on site or across clouds, so not everything has to move.
Enterprise Sources
Connects in place to enterprise applications, databases and cloud stores, keeping data where it already sits.
Enterprise Discovery Engine
Discovers metadata bottom-up, showing which systems and data are connected before migration waves are set.
Governance Gate
Applies the same access and compliance checks to every request, whichever environment holds the data.
Track 8 of 8 · Stage 3: Prioritise
From insight to decision automation
Most data strategies still stop at the dashboard: a person reads a number and decides what to do. This track decides which decisions should stay with people, which should be recommended and which should be taken by systems that act within guardrails, with people supervising.
Value is created when a decision changes, not when a report is published. As analytics, AI and agents mature, the bottleneck moves from producing insight to turning it into action, reliably and accountably. A data strategy that does not choose its decision modes deliberately will either automate the wrong things or leave its best models unused on a shelf.
So whatPick the decisions first, set the mode for each one on frequency and the cost of being wrong, and fund the guardrails and outcome measures together with the models.
Questions this track answers
Which of our recurring decisions create or destroy the most value, and who makes them today?
For each, what is the right mode: describe, diagnose, recommend, act with approval, or act within guardrails?
What must be true of the data, the rules and the explanations before a system may act?
Who is accountable when an automated decision goes wrong, and how is it reversed?
How will we measure the outcome of a decision rather than the activity around it?
Which partners can help us redesign decisions in our domain, rather than install more tools?
Exhibit 66Open loop versus closed loop
Open loop versus closed loop
The same insight, used in two very different ways.
Open loop: decision support
A dashboard or report shows what happened
A person interprets it, often late and inconsistently
Action happens in another system, if at all
Nobody records which insight drove which decision
Success is measured in reports built and users logged in
vs
Closed loop: decision automation
The decision itself is modelled: inputs, rules, options, owner
A system recommends or acts, within limits a person has set
Action lands in the operational system directly
Every decision is recorded with its evidence and outcome
Success is measured in outcomes, and the loop learns from them
So whatClosing the loop is not about removing people; it is about making every decision traceable from evidence to outcome.
Exhibit 67The decision ladder: five modes
The decision ladder: five modes
Each rung hands a little more of the decision to the system, and asks a little more of the data and the guardrails.
1
Describe
Show what happened, consistently, from certified data. The person does all the thinking.
2
Diagnose
Explain why it happened: drivers, segments, exceptions. The person still decides what to do.
3
Recommend
Propose the best option with its reasons and expected effect. The person chooses and acts.
4
Act with approval
Prepare the action in the target system; a named person approves before it runs.
5
Act within guardrails
Act autonomously inside set limits, escalate anything outside them, and record everything for review.
So whatMove a decision up the ladder one rung at a time, and only when the evidence from the rung below supports it.
Exhibit 68Three phases of the shift
Three phases of the shift
How technology, skills and the operating model change as organisations move from analysis to action.
Phase
What the technology does
Skills that matter
Operating model
Augmented analysis
Automates data preparation, finds patterns and explains drivers in plain language
Analysts who frame good questions and judge evidence
A central team serving the business with faster answers
Augmented decision-makers
Puts recommendations inside the tools where decisions are taken, for everyone, not just analysts
Business owners who trust, challenge and override recommendations
Domain teams own decisions; the data team owns the platform and standards
Generative and agentic
Agents converse, plan and take governed actions across systems, with people supervising
Decision designers, policy owners and supervisors of automated work
Decisions are products: designed, owned, monitored and retired
So whatMost organisations run all three phases at once in different domains; plan skills and governance for the most advanced one you intend to reach.
Exhibit 69Two ideas that are converging
Two ideas that are converging
Explained plainly, because both are often sold as products rather than understood as practices.
01
Decision intelligence
Treat decisions as things you design, measure and improve: model the inputs, the options, the rules and the expected outcomes, then learn from what actually happened.
02
Data fabric
Connect data where it lives through shared metadata and meaning, so any decision can draw on the right, governed data without a new pipeline each time.
03
Why they meet
Decision intelligence says what a decision needs; the fabric delivers it, in context, at the moment of decision. One without the other stalls.
04
Composable architecture
Build decisions from reusable parts: data products, rules, models and actions that can be recombined, rather than one monolith per use case.
So whatFund the shared meaning and the reusable parts once; every new automated decision then costs less than the last.
Exhibit 70Which decisions to automate
Which decisions to automate
Place each candidate decision by how often it is made and how costly a wrong call would be.
Low cost of a wrong callHigh cost of a wrong call →
Support and deliberate
Rare, high-stakes calls such as a major investment or a market exit. Describe and diagnose; keep people firmly in charge.
Automate with approval
Frequent and costly, such as credit limits or large claims. Recommend and prepare the action; a person approves, with tight guardrails.
Leave it simple
Rare and low-stakes. A good report is enough; automation would not repay the effort.
Automate within guardrails
Frequent and low-stakes, such as replenishment or routing. Let the system act, monitor outcomes and escalate exceptions.
Made rarelyMade very often →
So whatStart where decisions are frequent and the cost of error is low: that is where automation pays back fastest and safely builds trust.
Exhibit 71Guardrails and accountability
Guardrails and accountability
What must exist before any system is allowed to act.
01
A named decision owner
A business person is accountable for each automated decision, its limits and its results.
02
Written limits
Thresholds, amounts and conditions beyond which the system must stop and escalate.
03
Explanations
Every recommendation or action carries the evidence and the rule that produced it, in business language.
04
Reversal and override
A known, fast way to undo an action and to switch the automation off.
05
A complete record
Inputs, decision, approver, action and outcome are logged for audit and for learning.
06
Outcome monitoring
Drift, bias and unexpected effects are watched continuously, not discovered in the next audit.
So whatIf you cannot name the owner, write the limit and reverse the action, the decision is not ready to automate.
Exhibit 72Choosing delivery partners
Choosing delivery partners
Judge partners on the decisions and outcomes they can change in your domain, not on the number of tools they install.
Look for
Instead of
Ask them
Outcome-based measures in their proposals
Effort, licences and milestones only
Which business measure will move, and how will we both know?
Depth in your industry and its decisions
Generic platform skills
Which decisions in our industry have you redesigned before?
A view beyond the current project
Delivery of the brief as written
What will we need in two years that this design prepares for?
Willingness to adapt their team to yours
A fixed delivery model
How will you staff alongside our domain owners?
Focus on a few strengths
Claims to do everything
Where would you recommend another partner instead of yourselves?
So whatNo partner is strong in every part of data and analytics; choose for the decisions you are changing, and say no to the rest.
Exhibit 73Decision mode picker
Decision mode picker
Pick one recurring decision and answer six questions. The picker suggests the rung of the ladder it is ready for today.
Exhibit 74From insight to decision automation in six industries
Worked example
Industry
Situation
The move
Measure
Manufacturer
Input costs swing, price-downs bite and service decisions on the installed base are made case by case from spreadsheets.
Recommend price and surcharge moves to account managers, and automate spare-part replenishment within stock and value limits.
Share of pricing decisions taken on the recommendation; parts availability at agreed stock cost
Retail bank
Supervisors question risk data, and relationship offers are decided by campaign calendars rather than customer need.
Keep credit decisions at act with approval with full explanations, and let next-best-action recommendations reach advisers in their tools.
Approval decisions fully evidenced on first review; offers accepted per relationship
Retailer
Markdowns rise and stores and online plan stock separately, so price and allocation calls are late and inconsistent.
Automate store replenishment within guardrails, and recommend markdown timing and depth to merchants who approve by category.
Full-price sell-through; share of replenishment orders placed without manual change
Hospital group
Delayed discharges block beds and every automated step touching patients raises privacy and clinical questions.
Diagnose and recommend discharge readiness to ward teams; never automate clinical decisions, only scheduling within agreed rules.
Discharges completed before midday; recommendations reviewed by the ward team each day
Telecom operator
Churn is high and retention offers are generic; network spend is spread evenly rather than where it earns most.
Automate low-value retention offers within margin limits, and recommend network investment cases to a monthly capital forum.
Saves per retention contact at target margin; capital approved against modelled return
Insurer
Claims inflation, leakage and fraud slip through a high-volume operation, and brokers wait on slow referrals.
Fast-track simple claims within guardrails, route suspected fraud to investigators with explanations, and recommend referral terms to underwriters.
Simple claims settled without handling; leakage found per investigated claim; broker quote turnaround
Illustrative organisations, the same six as the worked examples.
Measures for the data scorecard
Decisions inventoried and ownedLead Share of high-value recurring decisions with a named owner and chosen mode
Recommendation acceptanceLead Share of recommendations accepted, with override reasons captured
Decision latencyLead Time from the triggering event to the decision being acted on
Guardrail coverageLead Share of automated decisions with written limits, explanations and a reversal route
Hand-back rateLead Share of automated decisions handed back to a person, tracked against the agreed band
Decision outcomeLag The business measure each decision exists to move, against its baseline
Reversed decisionsLag Automated decisions reversed or overturned on review, by cause
Value realised from automationLag Benefit confirmed with finance for decisions moved up the ladder
Traps to avoid
01
Automating the dashboard
Teams add alerts to existing reports and call it automation, but nothing acts.
Hover for the counter-move
Counter-move
Model the decision first: its inputs, options, owner and the system where the action lands.
02
Jumping to autonomy
A model goes straight to acting on its own, the first bad call makes headlines and the programme is frozen.
Hover for the counter-move
Counter-move
Climb the ladder one rung at a time, with evidence from the rung below before moving up.
03
No owner, no off switch
When an automated decision goes wrong, nobody can say who is accountable or how to stop it.
Hover for the counter-move
Counter-move
Name the decision owner, write the limits and test the reversal route before go-live.
04
Measuring activity
Success is reported as models deployed and recommendations served.
Hover for the counter-move
Counter-move
Report the outcome each decision exists to move, and the acceptance and override rates behind it.
05
Buying tools instead of redesigning decisions
Partners and platforms multiply but decisions stay the same.
Hover for the counter-move
Counter-move
Choose partners on the decisions they can change in your domain, and measure them on outcomes.
The first 90 days
Days 1-30
Inventory the decisions
List the recurring decisions behind each strategic theme
Name an owner and record the current mode for each
Place them on the frequency and cost-of-error grid
Pick two frequent, low-risk decisions as pilots
Days 31-60
Design two decisions
Model inputs, options, rules and the target system for each pilot
Write limits, explanations and the reversal route
Set the outcome measure and its baseline
Agree the approval path with risk and the decision owner
Days 61-90
Run and prove
Run the pilots at recommend or act with approval
Record every decision, override and outcome
Review results with the owner and decide whether to move up a rung
Add the decision portfolio to the quarterly strategy review
Where SCIKIQ fits
Decision and Reasoning Engine
Uses business context, rules and models to determine what should happen, so each decision is explicit and explainable.
Governance Gate
Checks policy, limits and approvals before anything runs, which is what makes act with approval and act within guardrails safe.
Action Fabric
Executes approved actions in the enterprise systems where the decision lands, closing the loop instead of stopping at a report.
Outcome and Observability
Measures the outcome of every decision and feeds it back, so modes can be moved up or down the ladder on evidence.
Twenty-two decks: the framework, the templates, six worked examples and eight tracks
View any deck here, or download it to adapt.
Framework and templates
Framework
The SCIKIQ Data Strategy Framework
The full framework: the strategy cascade, theme library, balanced scorecard for data, offence and defence, value at stake, stage gates, governance, rhythm and the first 100 days.
Templates from vision and choices to the strategy map, data objectives, posture, scorecard, use case and data product canvases, RACI and the 100-day plan.
Metadata is the data about your data: what it means, where it came from, who owns it, how good it is and who uses it. A data strategy without a metadata strategy cannot prove that any objective was met.
Data can play three different roles for an organisation: a dependable foundation, an enabler of better processes, or an engine for new growth. Deciding the mix is the first strategic choice, because it changes what you build, how you govern it and how you prove it worked.
Assess each data capability twice: where it is today, and where the chosen value proposition actually needs it to be. The gaps that matter become the roadmap; the rest are deliberately left alone.
Govern every data initiative with a small set of stable principles, then shape the architecture around how well the data and the questions are already understood.
Governance capabilities that once lived in separate tools are converging into platforms. Decide which policies matter most, what each persona needs, and which tools to consolidate.
DataOps is a working practice that brings communication, integration, automation, observability and operational discipline to the flow of data between the people who produce it and the people who use it.
Moving data and analytics to the cloud changes cost, speed and control at once. Plan it as a sequence of deliberate choices about what moves, in which order, on which ecosystem, and how spend is governed.
Most data strategies still stop at the dashboard: a person reads a number and decides what to do. This track decides which decisions should stay with people, which should be recommended and which should be taken by systems that act within guardrails, with people supervising.
387 slides in 22 decks. Editable PowerPoint and PDF, free to use and adapt. Template fields are in [square brackets]; worked examples are illustrative.
Twenty-one statements across seven areas. You get a score by area, the weakest links, and the deck to fix each one. About five minutes.
Strategic alignmentScorecard and measuresPortfolio and valueData foundationsMetadata and meaningGovernance and operating modelRhythm and adoption
Self-assessment for orientation; answers stay in your browser.
Frequently asked questions
What is a data strategy?
The set of choices that decides how data creates value for the business: which outcomes it serves, which decisions it improves, which data matters most, and the people, platform, governance and funding needed to deliver it, in order.
How is a data strategy different from a data maturity assessment?
A maturity assessment tells you where you stand today. A data strategy decides where you need to be for the business strategy, and how to get there. The assessment is an input to the Baseline stage of the strategy.
How long does it take to build a data strategy?
A focused first version usually takes weeks, not months: baseline and interviews, one leadership workshop to make the choices, then design, roadmap and the strategy on a page. It is then reviewed quarterly and refreshed annually.
Can we use the SCIKIQ templates?
Yes. The framework deck, workbook, interview guide, workshop kit, presentation template, measurement pack and communication guide are free to view and download in PowerPoint and PDF.