A small business running Power BI properly usually needs four tables of data, two questions answered reliably, and Power BI Pro at £10.80 per user per month. That is genuinely it for the first year or more. No warehouse, no capacity, no platform, no twelve-month programme.
I say that as someone who sells Power BI work, so treat it with appropriate suspicion. But the most common thing I see in small businesses is not underinvestment, it is an expensive system built for a company three times the size, which nobody opens, bought because the alternative sounded unambitious.
What does a small business actually need?
| What you need | What often gets sold instead | Why the difference matters |
|---|---|---|
| Two questions answered reliably | A 40-page "single pane of glass" | Nobody reads 40 pages, so nobody trusts any of it |
| Four to six tables, modelled properly | Every table in the ERP, imported wholesale | Wide imports are slow, fragile and confuse everyone |
| One agreed definition per key term | Each report defining terms its own way | Two reports disagreeing kills trust in both |
| Pro licences, about £11 per person | Fabric capacity "for the future" | Capacity costs multiples of per-user at your size |
| A scheduled refresh someone watches | A refresh nobody owns | Silent failure means stale numbers in real decisions |
| Someone who can change it next quarter | A finished system with no owner | Reporting needs are never actually finished |
Licensing prices are Microsoft's published UK figures as of September 2026, excluding VAT. For ten people that is £108 a month, which for most small businesses is not the number worth agonising over. The full licensing breakdown covers when that changes.
Which two questions?
Every business I have worked with has two questions that genuinely drive decisions, and a long tail of things people say they would like to see.
For a wholesaler it might be which products are actually making money after discounts, and which customers are quietly buying less than last year. For a services business, which jobs ran over and where the capacity goes next month. For a charity, where the money came from and what it did.
The tail is real, but it is not where the value is, and building for it first is what produces the 40-page report nobody opens. Answer the two properly, on numbers people trust, and the tail either turns out to matter or quietly goes away. Both outcomes save you money.
The test for whether a question belongs in the two: if the answer changed, would somebody do something differently this month? If not, it is interesting rather than useful, and interesting can wait.
Do you need a data warehouse?
Almost certainly not yet.
A warehouse earns its keep when several systems have to agree with each other, or when volumes get large enough that querying the source directly becomes painful for the source. Most small businesses have neither problem. They have one accounting system, maybe a CRM, and a couple of spreadsheets that matter more than anyone admits.
Power BI can model that directly, and will do so happily for years. If someone proposes a warehouse before looking at your data, ask which specific problem it solves that a well-built model does not. There are good answers to that question. There is also a version of the conversation where the answer is that warehouses are what the supplier builds.
Do you need Microsoft Fabric?
Not at small-business scale, in almost every case I see.
Fabric is genuinely good, and it is where Microsoft is heading. It is also priced and shaped for organisations with data engineering workloads, hundreds of report readers, or both. Buying a capacity at ten users means paying for headroom you cannot use, while your Pro licences would have cost a fraction of it.
The exception is real but specific: if you have hundreds of people who need to read reports, the capacity threshold at F64 changes the licensing arithmetic entirely, because free licences can then view content. That is a genuine tipping point rather than a fashion, and when to move to Fabric covers where it sits.
What does it cost a small business to get started?
The honest answer is that you can start for almost nothing and find out whether this is worth more.
Power BI Desktop is free. If someone in your business is comfortable with spreadsheets and has a few hours, they can connect it to your accounting export and produce something useful this week. That is not a consultant's answer, but it is the true one, and plenty of good small-business reporting started exactly there.
Where it goes wrong is the step after: the thing works, more people want it, and now it needs to refresh on a schedule, be shared properly, and produce numbers that match the accounts. That is the point where a few hours of help is worth more than a few months of trial and error.
For that step my entry point is an hour on your own data at £195, which gets a working dashboard running on your figures and your questions answered. A full health check is £950 if you already have reports and want to know what state they are in. Neither is a commitment to anything larger.
What breaks first in small-business Power BI?
Three things, reliably.
The refresh, silently. A password changes, a file moves, and the report keeps showing last Tuesday's numbers without any warning that it is doing so. This is the most dangerous failure in small-business reporting because it does not look like a failure. Somebody presents month-old figures in a meeting and nobody notices. Whatever else you skip, make sure one named person gets told when a refresh fails.
The one person who built it leaves. In a small business the reporting usually lives in one person's head and one person's laptop. When they go, nobody can change anything, and often nobody can even find the file. Write down where things live and who owns them, even informally. Who owns your Power BI estate covers the version of this that involves consultants.
Definitions drift. Sales starts counting orders at quote stage, finance counts them at invoice, and two reports that were both right last year now disagree by 15%. Agree the definitions once, write them down, and put them in the model rather than in each report.
When should a small business spend real money on this?
When the reporting is costing you more than it would to fix. That sounds circular, so here are the observable versions.
Somebody spends two days a month assembling a report by hand. That is roughly 24 days a year, and it will not get shorter.
Decisions get delayed because nobody trusts the number, or two versions of it are circulating and the meeting is spent reconciling rather than deciding.
You have started making commitments (to a bank, a board, an acquirer) that depend on figures you cannot produce reliably.
Any one of those justifies spending properly. None of them are fixed by buying a bigger platform, which is worth saying because the platform is what gets pitched. The signs you have outgrown your reporting go through the thresholds in more detail.
Common questions
Is Power BI suitable for a small business?
Yes, and it is arguably better suited to small businesses than to large ones, because a small business can hold its whole data model in one person's head and get to a trustworthy answer quickly. The risk is not that Power BI is too big for you, it is that people build for a company they hope to become rather than the one they are.
How much does Power BI cost for a 10-person company?
Ten Power BI Pro licences at £10.80 per user per month is £108 a month excluding VAT, on Microsoft's published UK prices as of September 2026. If only three people need to build and share reports, and the rest can use content another way, it can be less. The software is rarely the meaningful cost at this size, the time is.
Can I use Power BI for free?
Power BI Desktop is free to download and use, and you can build whatever you like in it. What you cannot do on a free licence is share it with colleagues through the Power BI service. For a business of one, or for evaluating whether this is worth anything, free is genuinely enough to start.
Do I need a consultant to set up Power BI?
Not to start. To make it reliable, usually yes, and the value is less in the building than in avoiding the model shape that makes everything slow and fragile a year later. If you can get something useful working yourself, do that first. Bring someone in when it needs to be trustworthy rather than just interesting.
What is the smallest useful Power BI project?
One report, answering one question, on data that refreshes automatically, that a named person looks at weekly. That is a complete and genuinely valuable project, and it beats a large system that nobody opens by every measure that matters.
Should I wait until my data is clean?
No, because the data will not clean itself and reporting is usually what exposes the problems. Start with the cleanest source you have, accept that the first version will surface disagreements, and treat those as findings rather than failures. Waiting for clean data is how businesses spend three years not reporting on anything.