What a data strategy contains
A data strategy ties data work to business outcomes: which decisions and products the data serves, how it is managed, who owns it, and how value is measured. A tool list with no outcome attached is a technology plan.
| Component | Answers | Evidence of a gap |
|---|---|---|
| Business alignment | Which goals and decisions the data serves | No use case traced to a strategic objective |
| Use cases and value | Where data creates revenue, cost or risk outcomes | Value claimed, never measured |
| Architecture and platform | Where data lands, is modeled and served | Copies of the same entity in many systems |
| Governance | Decision rights, policies, standards, stewardship | No named data owner for core entities |
| Data quality | Fitness for use, measured | No quality metrics or thresholds |
| Security and privacy | Classification, access, retention, lawful use | Sensitive data unclassified or over-shared |
| People and literacy | Roles, skills, data culture | Analysts rebuild extracts by hand |
| Roadmap and metrics | Sequencing, funding, KPIs | No dates, owners or baselines |
DalleMule and Davenport (HBR, 2017) split data strategy into defense and offense. Each organization sets its own balance; regulated industries lean defensive, competitive retail and media lean offensive.
| Defense | Offense | |
|---|---|---|
| Goal | Control, compliance, risk reduction | Revenue, customer insight, new products |
| Core idea | Single source of truth (SSOT) | Multiple versions of the truth (MVOTs) for different uses, derived from the SSOT |
| Typical work | Master data, lineage, access control, regulatory reporting | Analytics, ML, personalization, data products |
| Owner | CDO, CISO, compliance | CDO, business units, product |
An improvement recommendation names the component, the current state with evidence, the target, the owner and the measure. "Improve data quality" alone is not a recommendation.
Word bank
All definitions →- Data strategy
- A plan linking data management and use to business objectives, with owners, sequencing and measures.
- MVOTs
- Multiple versions of the truth: purpose-built views derived from the SSOT for specific uses.
- SSOT
- Single source of truth: one authoritative, governed version of a data entity.
Knowledge check
A bank's regulator flags inconsistent customer records across systems. Which side of the data strategy is weakest?
Which is a complete improvement recommendation?