TheySaid 4.0 - See how users will react, right where you build.
by•
Put what you’re building in front of real customers or synthetic users trained on them when you need an answer in minutes without leaving Claude, Cursor, or Jira. Nothing ships on a guess.
Replies
"Validation Debt" is a great name for it. As a solo maker who launched today, I know how easy that step is to skip. Question about the synthetic users: does a test tell you how confident it is, or flag when a result should be checked with real people? For a niche B2B audience I'd trust a synthetic "this flow is confusing" more than a synthetic "I'd pay for this".
@leadscout congrats on your launch today! and honestly your instinct is the right one, synthetic is strongest on behavior, "this flow is confusing", way more than on willingness to pay. on the confidence part, there isn't a single confidence score today, but every issue in the results has to point back to the actual sessions where a tester hit it, if it can't cite one it gets dropped, and issues are ranked by severity and then by how many testers ran into it. each one links straight to the recording so you can watch the moment yourself. for a niche b2b audience i'd use synthetic to clean up the flow fast, then put the pricing and "would you buy" questions in front of a handful of real people
@arnab_karmakar1 Thanks Arnab, that makes sense! I’ll test it out during the LeadScout onboarding, as that’s where I have the most questions. Is it possible to set up a persona like "SDR in a B2B company," or does it have to be configured using specific electrical/technical data?
@leadscout yep, plain language works, no technical data needed. when you set up the audience you just fill in things like job title (SDR), company size and industry, then a few lines on their motivation, how they see the world and what they already know, e.g. "books meetings all day, skeptical of new tools, lives in the CRM and linkedin". you can also nudge sliders like patience, tech trust and how direct they are, so you get a mix of testers instead of one. that's plenty for onboarding friction, and like Chris said, if you later want them to match your real buyers' preferences more closely, you can feed in more of your own customer data
@leadscout Congrats on your launch today! Skipping validation is the truest crime especially if it means you end up shipping something that is buggy or something that customers don't even want. Regarding your question, TheySaid flags responses from both human and synthetic testers and shows the impact of each finding. That way you can tell whether an issue came from one tester or many. Even a single response can surface something worth checking, and that’s your cue to validate it with another tester, or with a real person if it came from a synthetic one. Try it out and see how a synthetic tester answers both of your burning questions: “What’s working and what’s not?” and “Would you pay for this?”
@leadscout I'll add to what Arnab said. If you want preference matching between human and synthetic, you'll need to provide additional data for context and fine tuning. We can walk you through the data needed to approximate your actual customer behavior.