Almost no consultancy publishes prices, which leaves buyers unable to tell whether a quote is reasonable or whether they can afford to ask at all. A great many qualified enquiries never happen because somebody could not estimate the order of magnitude.
So here are real bands, what pushes a project up or down within them, and the costs that arrive after the invoice.
Why Nobody Publishes This
Two honest reasons and one dishonest one. Honest: the range genuinely is wide, because a dashboard on one clean source and a dashboard spanning six systems with contested definitions are different projects wearing the same name. And scope is discovered rather than specified — you find out what the data contains during the build.
The dishonest reason is that opacity lets a firm price to what a client looks like they can pay. That is real and it is why you should ask for a fixed price against a written scope rather than a day rate against an open one.
The Bands
These are for a competent practitioner, delivered remotely, working against reasonably accessible sources. Indian and Eastern European rates sit toward the lower end of each band; US and Western European onshore rates sit two to four times higher for the same scope.
| Project | Typical duration | What it includes |
|---|---|---|
| Performance audit | 2 weeks, fixed price | Diagnosis, prioritised findings, top measures rewritten, documentation |
| Single dashboard, one source | 3-4 weeks | Model, dashboard, refresh, handover |
| Finance or ops dashboard, 3-4 sources | 6-8 weeks | Integration, model, dashboard, validation, training |
| Semantic model rebuild | 4-6 weeks | Inventory, definition workshops, star schema, migration |
| Reporting automation | 3-5 weeks | Extraction, reconciliation rules, validation, distribution |
| ERP integration + reporting layer | 6-10 weeks | Extraction, warehouse, transformation, model, runbook |
| Full BI ecosystem | 3-6 months | Multiple subject areas, phased |
If a quote sits far below the relevant band, ask what is excluded — usually the data layer, which is where most of the work actually is. If it sits far above, ask who is doing the work and how many people are on the invoice.
What Moves the Number, in Order
Number of source systems
The single biggest driver, and it is not linear. Two sources is not twice one; it introduces reconciliation, which is where the time goes.
A project spanning four systems where nobody agrees which is authoritative can spend more time on that question than on everything technical combined.
How accessible the data is
A documented API or a database connection is cheap. A vendor portal somebody logs into and downloads a CSV from is expensive, because automating it means browser automation, which is fragile and needs maintenance.
Ask this before scoping. It moves estimates more than people expect.
Whether definitions are agreed
If finance and operations disagree about revenue, that is not a technical cost but it is a schedule cost, and it usually lands in the middle of the build.
Organisations with a documented metric catalogue get projects delivered noticeably faster and cheaper.
The state of what exists now
Greenfield is often cheaper than remediation. An existing model built on assumptions that no longer hold can cost more to untangle than to rebuild, and deciding which is the first week's work.
Licensing Is Separate, and People Forget It
Consulting cost is not total cost. Power BI Pro is a per-user monthly licence for anyone viewing content in a shared workspace. Beyond a certain user count, capacity-based licensing becomes cheaper than per-user, and the crossover point is worth calculating before you commit either way.
If you want Copilot, that requires paid Fabric capacity — it now runs from the lower capacity tiers rather than requiring the largest, but it is still a line item that did not exist under the older Q&A feature.
Add the source side too: a warehouse, a gateway machine, and whatever your ERP vendor's position is on indirect access. That last one has surprised more than one project.
The Costs After Go-Live
The part almost no proposal covers, and the reason some projects feel expensive in year two.
- Change requests. A used dashboard generates requests. Budget a few days a quarter or expect it to stagnate.
- Source changes. Someone renames a column and a refresh breaks. Unavoidable; the cost is whether it is caught by monitoring or by a user.
- Data growth. A model sized for today performs worse in two years. Incremental refresh defers this; it does not remove it.
- Knowledge decay. The single largest hidden cost. If one person understands the model and they leave, you pay for rediscovery. Documentation at handover is insurance against exactly this.
Where It Legitimately Gets Cheaper
Scope one report rather than a programme. The first build is how you discover what your data actually contains, and discovering that on one report is much cheaper than on twelve. The second project is always faster because the plumbing exists.
Do the definition work internally before the engagement starts. Two weeks of your own time agreeing what metrics mean can remove a week of consulting time and improve the result.
And be honest about freshness. Specifying real-time when daily would do can double an architecture's cost for a benefit nobody uses.
Red Flags in a Quote
A fixed price quoted without anyone looking at your data. That number contains a large contingency and you are paying for the vendor's uncertainty rather than for work. A short paid assessment first is cheaper for you even though it looks like an extra line item.
No data layer in the scope. If the proposal describes dashboards and says nothing about where data lands, how often, and what happens when a refresh fails, it is priced for the visible half of the work. The invisible half will arrive as a change request.
Day rates with no cap and no defined outcome. Discovery has no natural end. Cap it explicitly, or you are funding an open-ended investigation.
Handover defined as access to a workspace. Without documentation, transformation logic and training for a named person on your side, you have bought a dependency. That shows up as cost in year two, and it is the most common way a cheap project becomes expensive.
Versus Hiring Someone Internally
A fair comparison, because it is the real alternative and consultancies rarely make it honestly.
A competent BI developer is a permanent salary plus overhead, and they take time to recruit and longer to become productive in your environment. Against that, a project-based engagement has no ongoing cost and delivers faster because the practitioner has done the same thing before.
The honest answer depends on volume. If you will have continuous BI work for years, hire — and use a consultant for the initial build so your hire inherits something well structured rather than starting from scratch. If your need is lumpy, three months of intensive work then occasional changes, a permanent hire is overstaffed for the plateau.
The worst outcome is hiring one person, having them build everything alone with no review, and then losing them. That is the knowledge-decay cost above, concentrated into a single point of failure.
What We Would Do
We quote fixed price against written scope wherever the brief supports it, and where it genuinely needs discovery we cap that discovery — a short paid assessment producing a scope and a fixed price for the build. A vendor who will not work that way is telling you something about their estimating.
We also say when a project is not worth doing, which is the other half of pricing honestly.
Tell us the number of sources, how accessible they are, and whether the definitions are agreed. That is enough for a realistic band before anyone writes a proposal. Describe the project.