In Journey Atlas we treat replays as a qualitative companion to event paths, not as entertainment. The rule is simple: you may not present a clip until you have written three observations that do not assign motive. “User paused 14 seconds on the PromptPay screen” is allowed. “User was confused” is not, until a later interview says so.
Rage taps are a good example. They often look like anger. They are also how some people explore a frozen webview, how others double-submit because the spinner is the same colour as the background, and how a child in a shared-device household plays with a parent’s grocery app. If your journey report promotes every rage tap into a UX emergency, you will ship a bigger spinner and miss the shared-device pattern.
A review ritual that keeps us honest
We watch with the sound off first, then with any on-screen copy. One person logs timestamps. Another is forbidden from speaking until the clip ends. Only then do we map the timestamps onto the event stream. If the events do not show the pause, the instrumentation is the finding — not the user’s character.
This is slower than a highlight reel in a growth review. It also stops the team from laughing a real checkout failure off the stage. Customer journey reporting for apps already fights a bias toward clean funnels. Replays should not add a second bias toward mockery.
What we still cannot see
Replays will not tell you why someone opened Line in the middle of payment. They will not tell you that the wifi dropped in a condominium lift. They will not tell you that the person is buying for a parent. Journey Atlas keeps interviews on the syllabus for that reason. Events and replays describe movement. Motives remain a research problem.
If you want the worksheet we use to annotate clips, it ships with the Reading Room and with Atlas Desk. It is a dull spreadsheet. That is a feature.