What an ISG quadrant requires (and what it teaches about choosing a Data & AI partner)

If you’re going to hire a Data & AI partner in the coming months, a scene awaits you: four or five proposals on the table, all promising speed, results and experience. All with polished case studies. All calling themselves a benchmark.

And then comes the practical problem: how do you separate those who deliver from those who just present well?

After years on both sides of that table, I’ve arrived at a simple yardstick: distrust what vendors say about themselves. Trust what third parties have verified and what clients declare in numbers. 

Promises are free. Audits aren’t.

What an ISG quadrant really requires

I use ISG as an example because it’s a process I know from the inside, and because almost nobody explains what it involves.

Getting into an ISG quadrant isn’t about filling out a form or paying a fee. It means going through a battery of independent assessments: project volume backed by evidence, verified team qualifications, financial health open to analysis, client references checked.

The analyst has no interest in praising you; their job is to compare you with the biggest players in the market, coldly.

Dataside has been a quadrant leader for two consecutive years, competing for position with KPMG, Deloitte, BRQ and CI&T. I mention this less for the trophy and more for what it means by the yardstick of this article: this kind of result can’t be manufactured in a single sales cycle.

It takes an operation that works, at scale, for years, in front of those who verify it.

How to read a credential (any credential)

The tech market is full of badges, and not every badge is worth the pixel it takes up. The right way to evaluate one is to ask: what had to be proven, and to whom? Here are some examples of what each type of credential reveals.

Vendor advanced specialization. Microsoft’s AI Advanced Specialization, which we were the first Brazilian company to earn, requires a practice audit, evidence of delivered projects and team certifications. It proves technical capability verified by the people who built the platform.

Restricted partnership selection. Being one of two Databricks Partner Select companies in Brazil proves something different: out of everyone, the vendor chose who it trusts to represent its technology. Scarcity is a signal.

Management certifications. ISO 42001 (AI management), 27001 and 9001 prove process: that there is a documented method, audited annually, for governance, security and quality. For anyone about to put sensitive data in a third party’s hands, this is the most underrated credential on the list.

Individual recognition for pioneering work. Brazil’s first Azure OpenAI MVP is on our team. It proves depth of talent, the one asset no corporate slide can fake.

Notice the pattern: each of these required proof in front of an external verifier. That’s what separates a credential from a logo.

The script: five questions for your next vendor selection

Save these five and ask all of them, to every candidate. The answers will sort the table in no time.

1. “Who audited you, and what was verified?” A good answer names the assessor, the scope and how often it recurs. A bad answer names a paid media award.

2. “Which clients declare results, in numbers, on their own KPIs?” A case study written by the vendor is marketing. A number declared by the client is proof. Ask to talk to one of them.

3. “How do you measure ROI after go-live, and for how long?” If the answer ends at delivery, the return will end there too.

4. “What do you turn down?” A vendor that takes on everything has no criteria, and criteria are what protect your budget. Ask for real examples of projects they turned down.

5. “What deadline commitment do you sign?” Not the estimate: the commitment, with written consequences for whoever fails to meet it. The difference between the two is the difference between marketing and engineering.

We answer all five in writing, and by design: an assessment before the proposal, numbers declared by the client, Measure after go-live, a public list of what we turn down, and a working

MVP in up to 4 weeks, with a guarantee. But the script holds even if Dataside isn’t on your table at all. It exists to protect your decision, not to win us the deal.

The final yardstick

In the end, every selection process comes down to a trade: you’re going to entrust part of your company’s results to a third party. The least that third party owes you is verifiable proof that it has done this before, at scale, under audit, with clients willing to sign off on the numbers.

Everything else is just presentation.

Anyone can make a promise. Not everyone can pass an audit.