Conversion funnel: build it, instrument it, fix it

"Our site converts at 1.8%." That number says nothing actionable: where do the 98.2% get lost? The conversion funnel turns that black box into a step-by-step diagnosis. Well built, it points to the exact step to fix — and in what order to work.

Quick Answer: how do you build a conversion funnel?

Three decisions, then measurement:

  1. Pick the final goal — confirmed purchase, sent quote, activated sign-up. One conversion = one funnel.
  2. Split into 4 to 7 observable steps, each matching a real visitor action: product page view → add to basket → checkout start → details → payment → confirmation.
  3. Instrument each step in your analytics and read the pass-through rates between steps, segmented by device and source.

The reading: a step clearly below its benchmark (e.g. 25% basket → checkout when the norm is 40–60%) is the priority leak — CRO starts there, not on the page that "seems" improvable.

Conversion funnel in Mirage Analytics: pass-through rates and leaks per step

Building: the three rules of a funnel that teaches

One step = one observable action

"Interested in the product" is not a step; "viewed a product page" is. Every step must match an unambiguous measurable event — page reached, action performed. Psychological stages belong to the marketing journey; the operational funnel only measures facts.

The right grain: neither telescope nor microscope

Four to seven steps. Below, the leak stays vague ("somewhere between arrival and basket"); above, each step loses statistical significance and the analysis crumbles. Checkout often deserves its own sub-funnel (details → delivery → card → confirmation) when the global step leaks.

Funnels per goal, not one universal funnel

A site has several conversions: purchase, but also account creation, newsletter sign-up, contact request. Each has its funnel. Mixing them produces uninterpretable averages — the conversion rate guide covers micro and macro conversions.

Reading: pass-through rates and their benchmarks

The useful reading is never the global rate but comparing each step against its norm:

Step (e-commerce) Usual pass-through
Site arrival → product page view 40 – 60%
Product page → add to basket 8 – 12%
Basket → checkout start 40 – 60%
Checkout start → confirmation 50 – 70%

A step within the norm is not "done" — but a step far below is always the priority. Then segment: a decent global funnel can hide a mobile collapsing at the card step, offset by an excellent desktop. Device, source channel and new/returning are the three segmentation axes that pay.

Fixing: from leaking step to validated fix

The funnel locates; it doesn't explain. The full sequence on a deficient step:

  1. Qualify the leak: who abandons? If it's massively mobile, the work is ergonomic; if it's one specific channel, it's upstream (the ad's promise).
  2. Watch the friction: a heatmap of the page involved (dead clicks, scroll), then 5–10 replays of sessions abandoned at that step. That's where late-revealed fees, the fatal form field or the invisible button show up.
  3. Fix and test: the diagnosis-born hypothesis goes to an A/B test, judged on final conversion — not the step alone (inflating add-to-basket with a gimmick can crash payment).
  4. Re-measure the whole funnel: a repaired step moves the priority leak; the CRO cycle restarts.

The classic leaks, step by step

  • Arrival → product: confusing navigation, failing internal search, categories that don't speak the customer's language.
  • Product → basket: incomplete information (sizes, delivery, price), poor photos, no reassurance at the moment of doubt.
  • Basket → checkout: THE bad-surprise step — late shipping fees, forced account creation, promo codes sending people hunting elsewhere. Our cart abandonment guide is dedicated to it.
  • Checkout → confirmation: overlong form, missing payment methods, silent technical errors (watch through measurement, not complaints).

In practice with MirageFunnels are defined in a few clicks on your real goals, with pass-through per step, source/device/country segmentation, and direct access to each step's abandoned sessions in replay. Measured cookieless, hence on 100% of potential buyers — including banner refusers. Free 30-day trial.

FAQ

What is a conversion funnel?

The sequence of steps a visitor must pass to accomplish the site's goal: for example arrival → product page → basket → details → payment → confirmation. Instrumented in an analytics tool, it reveals each step's pass-through rate — and therefore the exact place where the site loses its buyers.

How many steps should a funnel have?

In measurement: between 4 and 7 — enough to locate the leak, not so many that analysis fragments. In the actual journey: as few as possible; every step removed from the real process (mandatory account, interstitial page) translates almost mechanically into conversion gains.

What pass-through rate is normal between steps?

In e-commerce, the usual orders of magnitude: 8–12% from product page to basket, 40–60% from basket to checkout start, 70–85% between checkout steps. A step far below its category is your priority — compare each step with its benchmark, not the whole funnel with one global number.

Are the marketing funnel and the conversion funnel the same thing?

The marketing funnel covers the whole cycle (awareness → consideration → decision → loyalty), often multi-channel. The conversion funnel is its website portion, measurable step by step in analytics. The former structures strategy; the latter gets fixed operationally — this article covers the latter.