We killed two markets in five months. The lesson wasn't about demand.

We spent almost three months in India and two in the UAE before moving DataWyse

entirely to the US. Both were the right call and neither was obvious at the

time, so here's what actually happened.

It started with my co-founder's father, who runs a bakery chain in my hometown. He wanted to know which products actually made money, which

branches were worth keeping, how much to produce for festival season. No

analyst, nobody to hand it to. We built him something rough and it worked.

So we went looking for more of him. Closed another mid sized customer in India. Then had to

let them go.

Here's the part I didn't see coming. We validated demand correctly. Every

business we spoke to had the problem, wanted it solved, and would talk to us

about it. What we hadn't validated was whether the data our product needed

actually existed anywhere.

It mostly didn't. No systems of record. Data entered inconsistently or not at

all. We'd connect and find ourselves doing weeks of cleaning and normalising

before we could run a single analysis. We weren't selling financial analysis,

we were selling a data migration project we hadn't priced or scoped.

Then the second problem: the price these businesses could support put us

underwater on inference costs alone. Not thin margins. Negative ones.

We spent two months in the UAE hoping it was an India-specific thing. Same

pattern.

The US mid-market is the inverse. The data is already sitting in NetSuite,

QuickBooks, Salesforce, HubSpot, structured and current, and nobody can get

answers out of it. Same problem, opposite constraint, and one we can actually

solve.

If you're building anything agentic, the thing I'd check earlier than we did:

your product inherits your customer's data hygiene. Talking to customers

tells you whether they want it. It doesn't tell you whether their systems can

support it, and for AI products that second question can kill you after

you've already got the first answer right.

Curious whether anyone else has hit this. Did you validate the data before

the demand, or find out the hard way like we did?