Trials are converting worse and the cause is upstream

The drop is not in the trial experience. It is in who is starting trials at all.

Segmenting starts by source shows the decline is entirely within self-serve traffic from a single referral partner, whose audience does not match the product. Excluding that source, conversion is flat at 14.8%.

The recommendation is to stop treating this as a conversion problem. The trial funnel is doing what it did last quarter; the input changed.

What we would need to be wrong

If the excluded source were converting at its historical rate elsewhere, the argument fails. It is not — that cohort converts at 3.1% on every surface we can measure.