Most data governance budgets are spent on the wrong fifth of the problem. Policy design, ownership models and classification schemes account for roughly 15% of the effort needed to make governance real; the remaining 85% is engineering inside the catalogue, the pipelines and the access layer. If the proposal in front of you reverses that ratio, you are buying a document and an operating model you will then have to implement yourself.
Three work types hiding inside one phrase
Data governance consulting services bundle three things that have different durations, different skill profiles and wildly different prices.
- Policy and operating model. Classification scheme, ownership and stewardship roles, approval paths, regulatory mapping. Six to ten weeks of workshops for a mid-sized organisation, and you only do it once.
- Platform implementation. Catalogue population, lineage capture, access policies, data contracts, quality tests running in the pipeline rather than in a spreadsheet. Three to nine months of engineering, and the part that nobody volunteers to fund.
- Run. Access reviews, exception handling, onboarding new sources. Permanent, and it belongs inside your organisation.
Buyers who split the budget roughly 15 / 70 / internal get working governance. Buyers who spend 60% on the first line get a framework that nobody can enforce, because enforcement lives in code.
How each supplier category prices it
| Supplier category | What they sell | Indicative day rate | Where it breaks |
|---|---|---|---|
| Audit-linked advisory practices | Target operating model, regulatory mapping, board-level sign-off | €1,400–2,500 | Implementation passes to a junior pyramid or returns to you |
| Global systems integrators | Multi-year programme, tool rollout across domains | €900–1,600 | Per-system onboarding cost repeats and the programme outruns the business case |
| Platform vendor professional services | Their own catalogue, configured properly | €1,000–1,800 | Governance stops at that platform’s boundary; your mainframe extract is still ungoverned |
| Boutique governance consultancies | One named specialist plus a method | €800–1,400 | Two people deep, so one resignation stalls the programme |
| Rented senior data engineers | Implementation capacity against your design | €450–750 | You need external authority to settle an ownership dispute |
Bands are directional and move with seniority and market. Current figures for data and AI roles sit on our daily rates page.
What actually moves the estimate
Four variables explain most of the variance between a €200k programme and a €900k one.
- Source system count. Onboarding one system into a catalogue with lineage and quality rules costs three to ten engineer-days. Forty systems is a different project from eight, and the multiplier is close to linear.
- Whether access control is centralised. If every system carries its own permission model, policy enforcement has to be written five times. A single policy engine over a lakehouse cuts that line by 50% to 70%.
- The regulatory driver. GDPR Article 30 records, DORA register obligations and EU AI Act documentation each fix the scope for you. That is helpful: it stops scope drift, but it also removes your ability to descope.
- Data quality rule count. Counting the rules you intend to enforce is the estimate. Teams routinely plan 40 and discover they need 400.
Before you sign
- Ask what percentage of the quoted days is implementation, and what the implementers’ CVs look like. Not the partner’s.
- Ask for the per-source-system onboarding cost, in days. A supplier who cannot answer has not sized the work.
- Check whether the deliverable runs. A lineage diagram in a slide is not lineage.
- Name the internal owner who will run it afterwards. If that person does not exist, delay the programme.
When renting engineers is the wrong answer
Three situations, honestly. If your real problem is that two business units disagree about who owns customer data, you need an external party with the standing to force a decision, and our engineers do not supply that. If you need signed regulatory attestation, you need a firm that carries that liability. And if nobody internally owns governance after we leave, the catalogue will be 40% stale within two quarters, whoever built it.
Where our model fits: you have the policy, you have an owner, and you need senior people to build the catalogue, the lineage and the enforcement. That is ordinary data engineering work with a compliance deadline attached. The same logic that decides how data analytics consultancies price an engagement applies here, and you can staff it by the month with senior data engineers from Eastern Europe.