Shortlist a data analytics consulting company on who will still own the pipelines in month nine, not on the capability deck. The decks all read alike because the work is the same everywhere: ingest, model, serve, explain. What differs is the business model behind the quote, and that shows up as a 4x spread in day rate and a wider spread in what you are left holding when the engagement ends.
Five models selling the same words
| Category | What you actually buy | Indicative EU day rate | Who owns the output |
|---|---|---|---|
| Big-4 style consulting practice | Method, governance, a partner’s signature | €1,400 to €2,600 | Joint, plus a handover pack |
| Global systems integrator | A scoped delivery programme | €900 to €1,600 | Them, until a transition milestone |
| Boutique analytics firm | Two or three specialists with domain knowledge | €800 to €1,400 | Usually you |
| Rented senior engineers | Named people inside your team, billed by the day | €450 to €750 | You, from week one |
| Talent marketplace | A profile and a contract you administer | €250 to €550 | You, including the vetting risk |
Rates move with country, seniority and notice period; our daily rates page carries current figures by role.
What the rate difference buys
Most of the spread is a staffing pyramid, not a quality gradient.
Large practices and integrators quote a blended rate across a team whose shape was decided before they met you: a partner, one or two managers, and a bench of analysts two to four years out of university. On a typical analytics programme, 60% to 75% of billed days go to that bench. You pay a senior rate for an average, and the senior part of the average spends much of its time reviewing the junior part.
Boutiques and rented engineers price per person, so the rate you see attaches to the person you met. Marketplaces do too, but the vetting moves to you: budget two to four weeks of your own engineers’ time per hire, and plan for a share of hires not lasting the first month.
The second driver is who estimates. Fixed-scope programmes carry 20% to 35% contingency because the vendor eats the overrun, a reasonable trade when the scope is genuinely knowable. Analytics scope rarely is, because nobody knows the state of the source data until someone profiles it. How each category of consulting firm prices covers the mechanics of that contingency.
How each one fails
- Consulting practices: the diagnosis is accurate and the build never starts. You get a target operating model and a long assessment deck, then a second procurement round to find someone who writes SQL.
- Systems integrators: scope locks against an assumption about your source systems that breaks in week three. Change requests then price at 1.3x to 1.6x the original rate, because you have no alternative supplier mid-programme.
- Boutiques: capacity. Two excellent people, both on another client’s incident during your month-end close.
- Rented engineers: they build what you specify. Specify the wrong metric and you get a well-engineered wrong metric, on time.
- Marketplaces: continuity. The contractor leaves, the semantic layer is undocumented, and the next person rebuilds it.
Four questions that sort a shortlist in one call
- Name the people who will do the work and their years of hands-on experience. A blended rate with no names means a pyramid.
- What is your ratio of billed days to people with more than five years’ experience?
- Walk me through a transformation repo you built that later got decommissioned, and why.
- In month four, when we change the grain of the core fact table, what happens commercially?
Question four separates the categories. Fixed-scope vendors raise a change request. Time-and-materials teams change the model on Tuesday.
Where renting engineers is the wrong choice
Our model fits when you know what you want measured and lack the hands to build it. It fits badly in three cases.
If you have no analytics leadership internally, rented engineers will produce exactly what an unclear brief asks for. Hire or borrow a head of data first. If a regulator or your board needs an independent opinion signed by a firm, buy the opinion from a firm; a data engineering team billed by the day cannot provide it. And for diagnostic work under 15 days, onboarding cost swamps the rate advantage, so a boutique that already knows your sector will come out cheaper.
Everything else, the multi-quarter platform rebuild included, tends to land cheaper and more maintainable under staff augmentation for data teams than under a scoped programme. You stop paying for the estimate and start paying for the work.