# Markaigen Benchmark Intelligence — Data Acquisition Manifest

**Evidence cut-off:** 8 October 2026  
**Release:** v0.5 Evidence + v0.6 Modelled

## Acquisition policy

1. Prefer publisher-owned first-party sources and official statistical offices.
2. Record the exact population, period, unit, sample (where applicable), and limitations before a value enters `evidence-master`.
3. Do not average incompatible metrics simply because they have similar labels.
4. Store official enterprise statistics separately from marketing-professional surveys.
5. Use public Gartner/DDMA summary statistics only; no proprietary underlying dataset is redistributed.
6. Treat forecasts/expectations as expectations, not observed outcomes.
7. Modelled values are generated only after the observed evidence layer is frozen for the release.

## Sources acquired and verified

| ID | Publisher | Acquisition mode | Evidence used | Model role |
|---|---|---|---|---|
| S01 | Eurostat | Publisher report + Eurostat datasets | Country/size/sector AI adoption; purposes | Primary calibration |
| S02 | CBS | Official statistical report / open-data ecosystem | NL microbusiness AI context | Context / validation |
| S03 | Statbel | Official statistical release / be.STAT ecosystem | Belgium AI/data/cloud context | Cross-check / context |
| S04 | DDMA | Published independent survey summary + disclosed methodology | NL marketing AI/governance | Marketing calibration |
| S05 | Gartner | Public CMO Spend Survey summary | AI budget/readiness | Soft enterprise constraint |
| S06 | Gartner | Public automation survey summary | Automation expectations | Scenario context |
| S07 | OECD | Intergovernmental comparative release | Cross-country firm adoption | Cross-check |

## Refresh design

Official API-capable sources (Eurostat, CBS StatLine, Statbel be.STAT) should be refreshed automatically in production. Independent research summaries are versioned manually after publisher verification. Any revised source value creates a new evidence release rather than silently rewriting a historical model.

## Audit fields

Every evidence record retains: `metric_id`, `metric_name`, `dimension`, `geography`, `population`, `segment`, `reference_period`, `value`, `unit`, confidence bounds if published, `source_id`, observation type, confidence tier, model-use flag and transformation notes.