Qualitative research at scale, for the first time

iFood Pago is iFood's financial services arm. In the fiscal year ending March 2026 it accounted for roughly a quarter of the group's revenue, with R$ 52.6 billion in transaction volume. It is an operation that decides fast and pays dearly for mistakes — and a good share of those decisions runs through the design team.

The catch is that qualitative research doesn't scale at the pace of a company shipping financial products. A traditional round means recruiting, scheduling, running and transcribing six to eight conversations. Weeks of work for a sample you can count on one hand.

In March 2026 the team used Grunn to investigate adoption of a new consumer credit product. The first five interviews landed within the first hour. Within 24 hours there were 26. The round closed with 33 in-depth interviews across a 30-hour window between the first and last response.

The second study, in July 2026, looked at payment behavior at checkout. The script was five minutes long. The average conversation ran 8.8 minutes — people talked nearly twice as long as planned, with no one having to push. Together, the two studies produced more than five hours of recorded conversation.

Volume creates a new problem, though: someone has to read all of it. That's where automatic synthesis stops being a convenience and becomes a precondition for the method to work at all.

In practice this changed the order of the work: qualitative research stopped being the step that holds up the schedule and started running in parallel with the quantitative one, feeding the same deck with real quotes and recordings.

Scale and depth stopped being a trade-off. That is exactly what Grunn exists to solve.