The CDO's data checklist
You are accountable for data the business owns, produced by systems you do not run, consumed by people who did not ask your permission. This is the shortest path from that position to something defensible.
01 The numbers you own
Five you should be able to defend
Not the longest list — the shortest one that covers the job. Each links to its full definition, including where the number lives and how it gets read wrong.
Governed metric coverage
The practical measure of whether a single version of the truth exists at all.
metrics_with_a_governed_definition ÷ metrics_used_in_reportingDQData quality score
Whether the numbers underneath every other metric can be relied on.
weighted pass rate across completeness, validity, uniqueness, timeliness and consistency rulesLCLineage coverage
The answer to "where did this come from", which decides audits and AI sign-off.
critical_data_elements_with_end_to_end_lineage ÷ total_critical_data_elementsTTITime to insight
The only metric in your set the business actually feels.
elapsed time from a business question being asked to a trusted answerFAForecast accuracy
The business-facing proof that better data changed a decision.
1 − ( Σ|actual − forecast| ÷ Σactual )02 Questions to ask
Six questions for your next review
Ask them of your own team before anyone asks them of you. The uncomfortable answers are the useful ones.
- What share of the figures in the board pack have one written definition and a named owner?
- Can any critical figure be traced to its source system entry without a person in the loop?
- How long does a real business question take from asked to answered and accepted?
- Which datasets have a published SLA, and what happens when one is missed?
- If an AI assistant answered from our data tomorrow, which answers could we defend?
- What are we spending on reconciliation that would disappear if definitions were governed?
03 First 90 days
A plan that finishes something
Narrow and finished beats broad and half-built. Each stage produces evidence that funds the next one.
04 Where the data lives
The systems behind your numbers
| System | What it holds | Key records |
|---|---|---|
| Source systems (ERP, CRM, core platform) | Where the records are actually created, and the only honest starting point for lineage. | Transaction, Master record, Change log |
| The platform and semantic layer | Where definitions live, and where every consumer should be getting them from. | Metric definition, Model, Owner, Version |
| Catalogue, quality and lineage | What exists, whether it is right, and how it got here. | Dataset, Rule, Steward, Lineage node |
| Consumption (BI, reports, agents) | Where governance either holds or is quietly bypassed by an export to a spreadsheet. | Report, Dashboard, Query, Agent run |
05 Questions
Frequently asked
Where should a new CDO start?
With the numbers the business already argues about, not with the platform. Governing the ten figures in the board pack produces visible value in a quarter and earns the mandate for everything after it.
How do I show progress in a role with no direct revenue?
Time to insight and governed metric coverage, published monthly. The first is felt by every executive; the second predicts it. Both move for reasons you control.
What is the relationship between governance and AI?
Governance is what makes an AI answer signable. An agent inherits every unverified hop in the data beneath it, so lineage and quality coverage are the practical gate on deployment, not a policy preference.
Bring your hardest number
We will trace one of your own figures back to source, live, on your data.