Funnel leakage
The one funnel step where far more people disappear than expected, pointing to a likely bottleneck, bug, or tracking gap.
See it
What it is
Every funnel narrows, but leakage is the one step that drops disproportionately hard. If the surrounding steps keep 80% of users and payment keeps 35%, payment is the leak. The comparison can be against neighboring steps, an older period, another device, or a known baseline. The point is the cliff, not the ordinary slope. Some of the slope is just the category: the Baymard Institute's running average of documented studies puts ecommerce cart abandonment near 70 percent.
Reach for it when overall conversion moved and you need one place to investigate. Read both the percentage and the number of people lost, then break the step down by browser, device, country, plan, and experiment variant. The strongest leak usually gives design, reliability, and support teams a much smaller search area.
Gotcha: analytics can manufacture a leak. A missing success event, an optional step forced into strict order, duplicate identities, or a window shorter than the real buying cycle all record valid users as drop-offs. Verify the instrumentation and watch a few session replays before redesigning the screen.
Ask AI for it
Create a PostHog Funnel insight for [ordered events] using unique users, a 7-day window, and the same eligibility filters at every step. Show step-to-step conversion, cumulative conversion, and absolute users lost. Flag the step whose conversion fell furthest versus the previous 28-day period, then break that step down by browser, device type, country, plan, and experiment variant. Link five matching PostHog session replays and verify that both the entering and completing events fire before calling the step a product leak.