Module 4’s quiet-week cut showed our new search facet was a launch firework. We kept the facet; we stopped putting it in the board pack.
96 written notes · mean 4.3
What people said after critique
These are edited for length, not for glow. At least one person below is still annoyed about SQL time. That is allowed.
★★★★★
Helen stopped me using “engagement” as a column header. I now write “accounts that completed two exports in 14 days”. Our data team sent flowers. I am not joking.
Short lab, heavy homework. The empty-state walkthrough in Activation Signal Lab is the only reason we found the hidden settings link.
I wanted more live SQL. The Cohort Retention Atlas still cleaned up our rolling-30 chart. I would not call the evenings relaxing.
Experiment Reading taught me to ask whether the tooltip was the treatment. Our “adoption lift” vanished when we labelled it correctly. Slightly bruising. Accurate.
Field study: a payments overlay in a GB neobank
A student in intake 11 brought a “smart split” overlay that looked beloved on week one. Eligible set was anyone with two accounts. After excluding staff and the first seven days, repeat use sat at 11% of eligible customers. The narrative pack recommended keeping the overlay for power users and stopping a planned homepage banner. Six months later they wrote to say the banner budget moved to statement redesign. We did not run their experiment; we taught them to read it.
Field study: warehouse picking on a tablet
A logistics analyst used Cohort Retention Atlas on a new “scan to confirm” step. First-week scans were high because trainers hovered. Cohort curves by shift lead showed night shift never formed the habit. The limitation they published: tablets shared logins, so person-level adoption was guessed, not known. That sentence went into their internal wiki. It is the kind of honesty the studio grades.