An effort model for producing a first data quality baseline for each critical data asset (CDA) in the baseline reporting return. Built from the GOV.UK policy and guidance, not from the earlier AI Mode answers. Every default is an assumption you can change.
Person-days for one CDA at the chosen depth and number of critical data items.
| Activity | DQAP step | Main role | A | B | C |
|---|
Rules exist for the critical data items and checks run on a schedule. A DQAP or equivalent exists. The work is mapping existing results into the return.
Some rules or ad hoc checks. Known issues are logged somewhere. No agreed targets and no complete baseline across dimensions.
No rules, no profiling, and often unresolved data access, ownership or sharing-agreement questions. Data access is the biggest source of delay.
The baseline template itself is not public, so this is the minimum set that the DQAP guide and issues framework support. Compare it with the real template to check the effort model fits.
| Field | Why it is needed | Source in GOV.UK guidance | Proportionate? |
|---|---|---|---|
| DQAP status (none, in progress, in place, equivalent) | Policy requires a DQAP or equivalent for every CDA | DAM policy; DQAP guide | Yes |
| Named data owner and steward | Accountable for meaning, content and quality | Identify and record CDAs; data ownership model | Yes |
| Number of critical data items measured | Shows baseline coverage | DQAP step 1 (Data Criticality Matrix) | Yes |
| Dimensions measured, with rule count | Six DAMA UK dimensions; not all apply to every asset | DQAP step 2 | 3 core |
| Baseline score per dimension (%) and date measured | The benchmark for every later report | DQAP step 3 | Yes |
| Target per rule, from importance × impact | Targets come from the combined score, not a flat percentage | Data quality issues framework | Later |
| Open issues by priority (critical, high, medium, low) | Drives funding and remediation order | Issues framework, priority score | Yes |
| Measurement frequency and method (automated or manual) | Shows whether the baseline can be repeated | DQAP step 7 | Yes |
| Coverage and equality check for personal data | Completeness can look high while groups are missing or misrecorded | Not in the guidance (recommended addition) | Add |
| Confidence in the baseline (high, medium, low) | Separates measured results from estimates | Not in the guidance (recommended addition) | Yes |
| Access route (open, restricted, on request, closed) | Q-FAIR Accessible: users need to know how to get the data | Not in the guidance (Q-FAIR addition) | Yes |
| Standards it conforms to, and shared identifiers used | Q-FAIR Interoperable: joins up only if formats and identifiers are shared | CDA metadata "conformsTo"; DAM interoperability requirement | Yes |
| Licence | Q-FAIR Reusable: no licence means no lawful reuse | Not in the guidance (Q-FAIR addition) | Yes |
| Provenance (where it comes from, how it is made) | Q-FAIR Reusable: users judge fitness from origin and processing | Not in the guidance (Q-FAIR addition) | Later |
| Q-FAIR rating per area: Good, Improving or Attention required | One comparable picture across organisations, using the Geospatial Commission's scale | Q-FAIR benchmark (Geospatial Commission, 2022) | Yes |
Q-FAIR adds Quality to the FAIR principles (Findable, Accessible, Interoperable, Reusable). The Geospatial Commission uses it to benchmark national geospatial data. Use the same three levels for any critical data asset. The toolkit fills these columns in the CDA return automatically from measured results and the asset record.
The DAM policy does not mention FAIR or Q-FAIR. Only the Geospatial Commission's six partner bodies are publicly known to have used Q-FAIR (2021 to 2022 benchmark). Ratings are an indicator, not an absolute measure.
Each CDA carries a small allowance for an equality and coverage check. Standard completeness and validity scores can hide exclusion: a record can be complete while sex, ethnicity, disability or language fields are defaulted, or while people with no fixed address or without a digital footprint never appear at all. For CDAs holding personal data, the baseline should report:
| Earlier claim | What the published guidance says |
|---|---|
| Flat targets such as 99.5% for critical fields, or sector bands such as 99.99% | Targets follow importance × impact (score 1–16), giving ranges from 75% to 100%. |
| Data owner must be Senior Civil Service | The owner is "usually a senior individual". No grade is mandated. |
| Mandatory metadata: UUID, steward, update frequency, AI readiness tags | Named fields include creator, created date, title, description, expires, supersededBy, temporal coverage, conformsTo and security classification. Update frequency, AI tags and UUIDs are not required. |
| Metadata "reported to GDS annually" | Returns must be maintained and reviewed every 12 months. No fixed date is published. |
| Sector calibration matrix and a generated rule-engine file | Not part of the policy. Useful as an idea, but it should not be presented as GOV.UK guidance. |
The departmental return workspace and its Dataverse schema can take the quality fields above as new columns, and the assurance dashboard can show baseline coverage by organisation.
Your Implementation Guide Playbook v1.0 already has a RACI and an implementation checklist that match the activity rows in this model.
Baseline quality scores make CDAs safer to reuse, and are the evidence the NDL needs to decide which assets to prioritise.