A cohort that finishes on time is not the same as a cohort that changed. When a workshop is tuned to move everyone across the finish line, the thing quietly discarded is the exact capacity we came for: the room to break, notice, and repair.
The default metric for adult learning programs is completion rate. It is legible, it is auditable, and it looks good on a slide. It is also a proxy that, over the last decade of workforce-learning research, has been repeatedly documented to correlate weakly with on-the-job behavior change (Class E: the Kirkpatrick-model critique tradition, from Alliger & Janak 1989 through Bates 2004, points at exactly this gap between Level 1 or 2 completion and Level 3 or 4 transfer).
Optimizing for completion introduces a specific pressure on facilitators: keep the pace, do not linger, get to the next module. That pressure is not neutral. It cuts against the thing the workshop is actually for.
What gets traded away
The capacity that makes a workshop stick is repair capacity. Repair capacity is the time a facilitator has to notice that a participant did not follow, to slow down, to double back, to let the group correct itself before moving on. It is unglamorous. It looks, on the outside, like a session running behind.
When completion becomes the target, three things systematically happen (Class C: this is what we observe in the cohort telemetry from our own workshop runs, cross-referenced against facilitator debriefs).
First, hard questions get deferred to office hours that most participants will not attend. Second, participants who fell behind in week two are quietly written off by week four, because catching them up would jeopardize the completion number for the ones still on pace. Third, facilitators learn to read the room for readiness to move on, rather than for the confusion that is the actual signal of learning happening.
The last one is the most expensive. Confusion, held in the room long enough to be worked through, is how a shared understanding gets built. Confusion, skipped past to protect the timeline, becomes a private failure the participant carries out of the room.
The completion number is a bad witness
A cohort with a 95 percent completion rate and a 15 percent behavior-change rate is a worse outcome than a cohort with a 70 percent completion rate and a 55 percent behavior-change rate. The first cohort feels successful to the org paying for it. The second cohort actually changes what happens on Monday morning.
We do not have clean industry-wide numbers on this comparison. That is a real limitation (Class U: the transfer-of-training literature has not converged on a stable ratio, and the studies that exist are heavily weighted toward compliance training, not toward the kind of capacity-building work a shared-build workshop attempts). What we have is a pattern, replicated across our own runs and consistent with what the transfer-of-training literature has been saying for thirty years: completion is a floor, not a ceiling, and treating it as a ceiling makes the ceiling lower.
What we do instead
In the workshop cluster this blog is part of, we treat the cohort as a shared build. That means we publish, up front, that some participants will finish the arc later than the calendar says, and that this is not a failure mode. It is the design.
Facilitator debrief time is protected. Participants who fall behind get a scheduled repair conversation, not a nudge email. The metric we watch, alongside completion, is what the participant can do in week eight that they could not do in week one, evaluated by someone other than the facilitator who taught them.
That does not scale as cleanly. It also does not quietly hollow out the thing we came to build.
None of this is a claim that completion is a bad metric. It is a claim that completion, on its own, is a bad witness to what a workshop is for. A single-number target that any competent operator can hit by lowering the ceiling is not a target worth optimizing.