Signup funnel: the steps to track

Signup page viewed. The funnel's entry. A pageview of /signup or whatever your equivalent is, which any analytics tool records already. The only decision is whether a signup modal that opens over another page counts, and it should, which means firing an event when the modal opens rather than relying on URL changes.

Form started. The first interaction with a field. This step exists to split "nobody wants to" from "people try and give up": a big gap between page views and form starts points at the pitch, the price, or the page itself, while a big gap between starts and submits points at the form.

Form submitted. The visitor pressed the button and the request went out. Fire it once per genuine attempt, not per click.

Email verified. Only if you require verification. It is the step most likely to be missing from the funnel entirely, because it happens in the inbox where your site cannot see it. The event has to come from your backend when the confirmation link is hit.

Account active. The person made it inside: first login, or the first screen of the product actually rendered. This is the step that catches the quiet failure where signup technically succeeded and the handoff into the product did not.

The stretch after this point, from first login to first value, is a different funnel with different problems, the activation territory covered in the activation benchmark.

Form analytics is one of the few corners of this subject with disclosed data. Zuko's benchmark, built from 93 million form sessions, puts overall view-to-completion at 45 percent and completion among people who start the form at 66 percent. Those cover forms of every kind, not signup forms alone, but the shape transfers: the biggest loss is usually before the first keystroke, and of those who start, roughly a third still leave. The same dataset's field-level finding is worth knowing for signup specifically, because the password field is its single worst performer, abandoned by about one in ten of the people who reach it.

Treat these as orientation, not targets. The honest benchmark question, what share of visitors should end up as signups, has its own post with figures split by signup motion in visitor-to-signup rate.

Two tools watching the same form will report different start counts, because there is no standard for what starting means. Some fire on the first click or tap into a field, some on the first keystroke, some on any interaction including tabbing through. The difference is not cosmetic: focus-based counting includes everyone who clicked into a field and typed nothing, keystroke-based counting excludes them, and the gap between those two groups is precisely the hesitant audience you are trying to measure.

Pick one definition, write it down, and keep it. First field focus is the more useful choice for signup forms, because the person who clicked into the email field and left without typing is real drop-off you want to see. Whatever you choose, the number is only comparable to itself over time, a point that generalises to most of why analytics tools disagree.

Four failure modes account for most broken signup funnels, and all four produce plausible-looking numbers rather than obvious errors.

Counting the CTA click as a form start. The "Sign up" button on the homepage is an ad for the form, not the form. If clicking it fires your form-start event, the start count inflates and the start-to-submit rate craters, and the funnel now says the form is broken when the form is fine.

Double-firing on validation. Forms that validate before submitting can emit a submit-shaped event on the validation pass and again on the real submission. Every rejected password attempt becomes a phantom submit, and the funnel reports more attempts than people.

Single-page-app amnesia. When the signup flow is client-side routed, pageview-based steps stop firing on navigation, and a multi-screen signup can register as one long stay on step one. Funnel steps in an SPA need explicit events on each screen change, the same issue that breaks pageview tracking generally.

Trusting the frontend alone for the final step. Confirmation emails bounce, verification links expire, and account-creation calls fail after the browser already showed success. The account-active step should come from the backend, where an account either exists or does not. When the frontend count and the backend count disagree, the backend is right, and the size of the disagreement is itself a useful number.

Signup with a Google or Apple button is a different path through the same funnel, and it should be tagged as one. Record the method on every signup event, email against OAuth, and read the two paths separately: they fail differently, since there is no password field to abandon but there is a consent screen and a redirect that can eat people. One documented technical caveat: Safari's tracking prevention caps first-party cookies set by script at seven days, and OAuth flows embedded in iframes can break silently under it, so a signup form that works in Chrome deserves a test in Safari before you trust the split.

The reading rules are ordinary funnel analysis: find the biggest step-to-step loss, segment by device and traffic source before concluding anything, and compare against your own last month rather than a stranger's average. Signup funnels add one wrinkle, small numbers. A site doing thirty signups a week cannot read a five-step funnel daily, because every step is a handful of people and the percentages jump around. Weekly or monthly windows keep you from chasing noise.

The step the numbers cannot explain is where recordings earn their place. In Bigdelta, any tracked event can be a funnel step, the drop-off between every pair of steps is computed for you, and a click on the leaking step opens the session recordings behind it, so you can watch the password attempts fail instead of inferring them. Signups also close a loop specific to this funnel: once someone signs up, their anonymous pre-signup history merges into their profile, so you can trace a paying customer back to the funnel they originally stumbled through.

Five events: page viewed, form started, form submitted, email verified, account active. Fire the start on first field focus, the submit once per real attempt, and the last two from the backend. Write down your definitions, tag the signup method, and give the numbers a week before reading them. An afternoon of instrumentation buys you the difference between "signups are down" and "mobile Safari starts collapsed on Tuesday", and only one of those is fixable by Friday.