cdo-review

---
name: cdo-review
description: Use when validating training-data consent before an AI project, choosing a data architecture, or valuing a data asset for productization or M&A.
category: C-Level Advisory
version: 1.0.0
tools: []
---

# CDO Review — Data Strategy Interrogation

Pressure-test any plan touching training data, data architecture, data productization, or data hiring with six forcing questions before commitment.

## When to run

Before approving an ML training run using customer data; before signing a multi-year data-infrastructure contract (warehouse, ELT, orchestration); before productizing customer data (benchmark report, embedding endpoint, license); before a major data hire (head of data, data PM, ML engineer); before M&A diligence, yours or theirs; or whenever "monetize" gets used near "data."

## The six questions

1. **What decision does this data drive?** "We might need it later" or "it feels like a moat" aren't decisions — name the specific business call this data unblocks, or question why it's being collected, trained on, or productized at all.
2. **What's the consent provenance for every source?** For each source, know its origin, consent flow, data class, and intended use. First-party-TOS-only consent is weaker than first-party explicit opt-in, and a bundled TOS doesn't cover a materially new purpose like training foundation models on PII — audit this whenever an AI use case is in scope.
3. **Who consumes this internally, and across how many functional domains?** This drives the architecture choice: fewer than 5 consumers argues for warehouse-only, 5-25 for a lakehouse, 25+ with a federated culture for a data mesh. A premature architecture choice is the top cause of data-team burnout.
4. **What's the M&A diligence impact?** If an acquirer asked about this data corpus tomorrow: is there a documented anonymization process, what percentage of customers have MSA carve-outs, and are training-data provenance logs current? Revisit quarterly.
5. **Can the model, decision, or report be re-run without this source?** If yes, blast radius is low and consent posture can change later; if no, the organization has structurally committed to that source and it needs harder vetting up front.
6. **What role unblocks this, and is it the right next hire?** Hiring a data scientist when an analytics engineer is the right answer costs roughly a year of productivity — map the decision to the specific role and confirm prerequisites are staffed (a data engineer before an ML engineer, an analyst before a data scientist).

## Output structure

State the decision in one sentence (training, architecture, asset, or hire), then report: for training, the count of NO-GO / MITIGATE / GO sources and the top remediation; for architecture, the recommended model (warehouse / lakehouse / mesh), a build-vs-buy summary, and the criteria that would trigger revisiting it; for asset valuation, a strategic-value score, a moat rating (strong/medium/weak), and an M&A multiplier range against ARR; for hiring, the next role, why it beats the alternative, and whether prerequisites are staffed. Close with a verdict — SHIP, SHARPEN, or BLOCK — and three concrete next steps.

## Related reviews

Route productization or licensing paths to the legal reviewer; route architecture changes touching customer data to the security reviewer; route build-vs-buy TCO and M&A valuation math to the finance reviewer; route data-team hiring to the people reviewer. Freeze multi-year infrastructure contracts for review before signing.