Synthetic Research + ODI: 2 Weeks of User Research in 30 Minutes

A founder came to me with 15 feature ideas and zero clarity on which one to build. Normally that’s a 2-week research project. We figured it out in 30 minutes for $0.

Here’s what happened.

I was consulting for a product team. Tiny dev resources, lots of ideas, no data. The usual answer would be: recruit a panel, run surveys, wait for responses, clean data, analyze. Two weeks if you’re fast, a few thousand dollars for a decent sample.

I learned about synthetic research from Bayram Annakov, who’s been sharing great work on LLM-based consumer research. The idea: LLM-generated personas answer open-text questions about your product, then Semantic Similarity Rating maps their responses to a Likert scale. PyMC Labs validated this against 57 real surveys (9,300 actual humans) - 90% correlation.

But here’s what got me excited. I combined this with Tony Ulwick’s Opportunity Scoring (ODI). You measure two things for each job-to-be-done: how important is it, and how satisfied are users with current solutions. The gap = your opportunity.

30 minutes later I had a ranked feature list. Three features scored above 15 (strong opportunity - build now). Two scored below 5 (already well-served, skip). The founder’s personal favorite? Scored 7. Not terrible, not urgent.

We made the call with data instead of opinions. One coffee break instead of two weeks.

I still do real user interviews. But when you need directional confidence before committing eng resources, this is something else.

The tool is open-source. I forked Bayram’s original repo and added ODI scoring: github.com/tolmachevmaxim/synthetic-market-research

How do you decide what to build first - gut, stakeholder loudness, or something more structured?