We schedule forty-minute conversations, not surveys with twenty tools listed. The first ten minutes reconstruct yesterday’s work in order: which IDE window stayed open, which CI failure they ignored, which internal CLI they copied from a teammate’s notes.
Then we show a sanitized usage extract for the same period—only if the participant agrees. The mismatch is usually mild and useful: people overestimate how often they use documentation portals and underestimate how often they bounce between two overlapping chatops bots.
For application analytics for developer tool usage inside engineering platforms, interviews are not soft decoration. They explain why an event spike coincides with a release freeze or why a “popular” plugin only fires from one team’s custom wrapper. Without that context, dashboards punish the wrong owner.