How to choose a Microsoft Fabric partner.

Start with the only test that matters

Every partner you talk to will say the same things. Senior team. Proven methodology. Deep manufacturing experience. Microsoft Gold, or whatever the current badge is called. None of it separates them, because none of it is falsifiable in a sales meeting.

One thing is. Ask each of them to build something real on your data, on a fixed price, before you commit to the large engagement. What you learn in that exercise is worth more than every reference call combined:

• Whether the people who impressed you in the pitch are the people who actually show up
• Whether your data is in the shape everyone assumed it was in
• How they behave when something goes wrong, which it will
• Whether they can explain a technical problem to your CFO without condescension

Partners who work this way will say yes quickly. Partners whose model depends on long discovery phases will explain why it is not possible. That answer is the information you were looking for.

Seven questions to ask before you sign

1. Who actually does the building?

The single most common gap between the pitch and the delivery. Ask for the names and LinkedIn profiles of the people who will touch your project, not the org chart. Then ask what percentage of their time is committed to you. A good answer is specific and small. A bad answer describes a pool, a bench, or a centre of excellence.

2. Is the price fixed, and who carries the risk if it runs long?

Time and materials moves every estimation error onto you. Fixed-bid moves it onto the partner, which is why partners who quote fixed-bid tend to scope more carefully. If a firm will only work time and materials on a well-defined problem, they are telling you they do not trust their own estimate.

3. What happens if the result is wrong?

Most contracts define what gets delivered. Few define what happens when the delivered thing does not answer the question. Ask directly. The answer tells you whether they have ever stood behind an outcome as opposed to a set of hours.

4. How soon will I see something working on my own data?

Not a demo, not a sample dataset, not a slide. Your data. If the answer is measured in quarters, you are buying a program rather than a result, and programs are where mid-market budgets go to die quietly.

5. What do I keep if we part ways?

You should own the code, the data model, the documentation, and the environment. Ask whether anything they build sits inside proprietary tooling you cannot run without them. Some accelerators are genuinely useful. Some are lock-in with a friendly name.

6. Which Microsoft credentials do you hold, and can I verify them?

Every Microsoft credential is checkable in the Microsoft partner directory. Ask which ones they hold, then look them up while you are still on the call. The next section explains what each one is worth.

7. What will you tell me that I do not want to hear?

The most useful question on the list, and the one nobody prepares for. A partner who has done this work knows your data has problems they will have to raise. If they cannot name a single uncomfortable thing, either they have not looked, or they are managing you.

What the Microsoft credentials actually mean

Microsoft's partner tiers changed names several times, which is why the badges on partner websites are hard to compare. Here is what each level actually requires.

The practical move: ask for the Advanced Specialization, in the discipline closest to your problem, and verify it yourself in Microsoft's directory rather than trusting a logo on a slide.

Red flags

• A discovery phase priced like a project. Assessment work has its place, but a six-figure audit that produces a document is a way to bill while deciding whether they can help you.
• The demo runs on their data. Any platform looks excellent on clean sample data. Yours is not clean. That is the entire problem.
• Named seniors who vanish after signature. Ask for their time commitment in writing.
• Tool-first answers. If the recommendation arrives before the questions do, you are buying whatever they had in stock.
• No willingness to be wrong. See question seven.

When a Big 5 firm is the better choice

It would be convenient to argue that specialists always win. They do not, and pretending otherwise costs you credibility with your own board.

Choose a large global consultancy when the work spans many countries and legal entities, when you need hundreds of people mobilised at once, when the data project is a small piece of a business transformation with change management attached, or when your board simply requires a global brand on the contract. Those are real reasons and a specialist cannot match them.

Choose a specialist when the problem is a defined data problem, when you want senior people building rather than managing, and when you would rather have a working result in weeks than a governance structure in months. Most mid-market manufacturers, the ones running between fifty million and a billion in revenue with a complex ERP and a decade of acquisitions behind them, are firmly in the second case.

How this maps to what we do

Full disclosure, since you are reading this on a consultancy's website. Think AI is the specialist in that comparison, and the questions above are the ones we want to be asked.

We hold the Microsoft Advanced Specialization in Data Warehouse Migration to Microsoft Azure, our senior architects do the building, our engagements are fixed-bid, and our answer to question four is one week. If the dashboard does not answer the question we agreed on, on your real data, you pay nothing.

If that is useful, the Proof of Value is where it starts. If you take this list to four other partners instead, the article did its job anyway.