Most funnel analytics ask you to define the funnel. You list the steps you believe visitors take, and the tool faithfully reports drop-off along the path you invented.
Which is fine, until the path is wrong. And it frequently is — because you know your site, and knowing your site is the thing that prevents you from seeing it.
The problem with hand-built funnels
They encode your assumptions. You built landing → pricing → signup. Visitors go landing → docs → pricing → back to landing → signup, and the third of them who took the detour appear in your data as a leak that isn't there.
They hide leaks outside the path. Anything happening off your defined steps is invisible by construction. The step where people actually leave might not be in the funnel at all.
They rot. You ship a new page, the funnel keeps measuring the old one, and nobody notices for a quarter.
What discovery does instead
Funnel discovery starts from behaviour. It watches the sequences visitors actually walk, finds the common paths, and reports where each one loses people — including the paths you'd never have thought to define.
Two things fall out of that immediately.
The real flow. Often close to what you assumed, occasionally very different, and the difference is usually the most interesting fact you'll learn that quarter.
Leaks in places you weren't looking. A step that isn't in anyone's mental model — an interstitial, a redirect, a country selector — quietly costing you visitors.
Structure first, behaviour after
There's an honest sequencing here that matters, because it's where most tools overpromise.
Before any traffic, you can suggest a sensible starter flow from a site's structure alone: an ecommerce site has a cart and a checkout; a SaaS site has a signup and an activation. That's a reasonable hypothesis, available on day one, and it needs no snippet.
After enough traffic, and only then, you can measure the real flow and compare it against the hypothesis. If they differ, that difference is the finding.
This is how Defrixa works. A structural flow is suggested immediately from your site. Once the tracking snippet has seen enough visitors, the real flow is measured, and where the two disagree, you're shown the flow that was found rather than the one that was assumed.
What discovery still can't do
It finds the step. It doesn't, by itself, find the field inside a form, or explain why the step fails. Discovery narrows the search space to a place; the cause still needs a human looking at what that step asks of a visitor.
And it needs traffic. Below a certain volume, "the common paths" are a handful of sessions, and reading them as a pattern is superstition.
Common questions
You may want to, for the flows you care about commercially. Discovery is what tells you whether your definition matches reality.
Enough that a path is a pattern rather than a coincidence. Small sites should treat early discovered flows as suggestions.
No. Recordings show you individual journeys, one at a time, and you generalize by watching. Discovery aggregates the journeys and reports the shape.