By the end, we agree exactly where users drop off and pick 1–2 changes to test.
"When I open my ERP, I want to connect my bank once and trust it works, so I stop entering transactions by hand."
Don't show the data yet. Each person guesses. Tap a step every time someone calls it — let the room commit out loud.
Reveal one bar, then ask "what do you notice?" before you say a word. Click a number to type Dominio's real value.
◆ Dominio · past 30 days · first attempt per (user × bank) · n=6,167Tap each step to reveal who renders it. The drop sits on a screen we don't draw — that's the whole reframe.
The cliff is Confirm — the bank's screen, not ours to redesign. The one step we control right before it is Prepare. That is the lever.
"The biggest loss we can measure is on a screen we control. So what do we change first?"
Two independent signals, same step
Almost every failure before the bank login is the user not completing input — auth-pending and timeout dominate — and it concentrates in the biggest banks (BB, Itaú, Caixa).
The same step surfaces in what firms told us — clarity on what to share, and getting their data in. Sentiment and behavior agree.
"How might we get the user ready and confident before the bank screen, so more push through?"
Run the brainstorming on the shared Excalidraw board so everyone sketches at once. Come back here to prioritize and commit.
| Experiment | Owner | Metric | By when |
|---|
Restate the shared win — moving that ~50% up — and read back the experiments and their owners. Then capture the follow-up date and export.