01Prompted and unprompted
A survey question tells the respondent what to think about. A conversation between strangers does not. That difference shapes everything that follows.
Unprompted talk is messier, but it shows what people care about enough to bring up on their own.
02What we lose
Observational data skews towards people who post. We correct for that, but we never claim it represents everyone. It represents the conversation.
“We report what people say to each other, not what they would say to us.”
03What we gain
Scale, timing and language. We see reactions within hours, in the words people actually use.
Key findings
2hmedian time from a product event to the first measurable shift.
300+distinct phrases we track as stand-ins for satisfaction.