Most tools score a page. The trouble is that nobody buys from a page — they move through a sequence, and they give up somewhere in the middle of it. Flow analysis measures the sequence.
One page tells you almost nothing
A landing page with a 47% bounce rate is not a diagnosis. Some of those people were never going to buy, and some of the ones who stayed dropped out two steps later, where it actually cost you something.
A conversion flow makes the loss legible. If 1,299 visits open registration, 691 reach the details step and 497 finish, you are not looking at a bounce rate any more — you are looking at 608 visits stopped at one joint, and one place to go and fix.
Rank by visits, not percentages
A 60% drop at the first step is often mostly people who arrived by accident. A 20% drop at payment is made entirely of people who intended to buy. Later leaks are worth more because more of the loss is recoverable.
What conversion flow tracking measures
Once a small script is on the site, each step of a defined flow is counted from your own traffic — not modelled, not sampled, not estimated. The unit is a visit: the tracker is cookieless and stores nothing persistent on a device, so visits are deliberately not deduplicated into unique people.
- Reach and drop per step — how many visits arrived, how many continued, and how many stopped between the two. A visit that does not continue has not necessarily left the site.
- Where they came from — the page they were on immediately before, and the off-site source that sent them.
- Where they went instead — the page people opened when they left the flow, which usually names the reason.
- Who finished by any route — everyone who reached the last step however they got there, so a mistyped first step does not hide your real finishers.
Three views over the same data
The same counts read three ways: the ordered walk through the flow, a ranked list of pages by how many visits stop on them, and the outcomes view — everyone who completed, out of every session on the site.
Monitoring: being told, instead of checking
A flow that worked in March can quietly break in June — a redesign moves a button, a form gains a required field, a script fails on mobile. Nobody notices until the month is over.
- Structural checks need no traffic. If a protected page changes and its score falls materially, that is caught on the next pass, the same day.
- Measured checks need evidence. A drop in a step's rate has to clear a sample floor, be large enough to sit outside normal variation, and persist for about two weeks before it is called a regression.
- Seasonality is accounted for. A quiet week, or a change in where traffic came from, is not reported as a broken flow.
The point of the thresholds is to make an alert mean something. A tool that cries wolf at every daily wobble gets muted within a fortnight.
Then measure whether the fix worked
Fixing the worst step is only half of it. Defrixa compares the same step before and after the change, in your own traffic, and reports the verdict plainly — including when the answer is that nothing moved, or that it got worse.
Changing several things at once makes this impossible, which is why the ranked list is there to read but you ship the top item first. That is the only way the before-and-after means anything.
Common questions
What is flow analysis?
Flow analysis measures a whole journey rather than a single page. Instead of scoring your homepage in isolation, it counts how many visits reach each step of a real path — signup, registration, checkout, booking — and how many fall out between one step and the next.
How is flow tracking different from page analytics?
Page analytics tells you a page had 1,299 visitors and a 47% bounce rate. Flow tracking tells you 1,299 started, 691 reached the next step, and 608 left at one specific joint — so you know which step to fix, not just which page was busy.
How much traffic do I need before a conversion flow can be monitored?
A structural change is caught immediately with no traffic at all, because it compares the page itself against a protected baseline. Judging a change in a measured step rate needs roughly 200 sessions through that step and about two weeks, so a normal quiet week is not mistaken for a real drop.
See where your flow is losing people
Start with a free Friction Score on any page. No signup, no card.