Across the first four cohorts, attendance held near ninety percent for two weeks, then fell off a cliff. Not gradually — specifically, in week three, and specifically among the students who had been most engaged in weeks one and two.

What week three asks

Weeks one and two are additive. You list your options and you gather evidence, and both feel productive in the way that research always feels productive. Week three is the first week that takes something away: you write the criteria that will kill options you're still fond of.

That's not a motivation problem. It's a grief problem wearing a scheduling excuse. The messages we got were all logistics — a work trip, a sick child, a deadline — and the pattern underneath them was identical.

Three structural fixes

Name the dip in week one. We now tell every cohort, on day one, that week three will feel like the wheels coming off and that this is the design working. Predicted difficulty is survivable in a way that surprise difficulty isn't.

Move the ceremony earlier. The pod check-in used to sit at the end of week three. It now sits before the criteria session, so nobody writes their disqualifiers alone.

Shrink the artifact. The week-three deliverable went from a full written filter to five lines. The work is the same; the blank page is smaller.

The result

Week-three attendance across the last eight cohorts sits at eighty-eight percent, and completion overall moved from the sixties into the low nineties. The industry average for self-paced courses is around thirteen percent. Structure is doing that, not willpower.