How to use Funnel Analysis
Funnel analysis breaks your store's purchase path into stages and measures the drop-off at each one. This guide shows how to define stages, read the numbers, and turn diagnosis into a test backlog.
Funnel Analysis
Funnel analysis is the practice of splitting a conversion path into stages and measuring drop-off at each transition.
Funnel analysis is the diagnostic method that sits underneath every conversion-rate decision. You take a goal — usually a purchase — and decompose the path to it into discrete stages: landing, product view, add-to-cart, checkout start, checkout complete. For each transition between stages you calculate a stage conversion rate, and the gaps between those rates tell you where customers are leaving.
It's the step that precedes optimization. You can't prioritise a redesign, a new payment method, or a copy test without first knowing which stage is actually leaking. A good funnel analysis turns a vague "checkout is broken" hunch into a number: 62% of carts are abandoned at the shipping step on mobile.
Most stores already have the raw data — GA4, Shopify analytics, a heatmap tool — but the data sits in different places and the stages don't line up across tools. The work is less about collecting events and more about defining a clean stage model and reading the deltas honestly.
Funnel analysis is the diagnostic half of funnel optimization. Optimization is what you do with the findings — running tests, rewriting copy, fixing forms. Without the analysis, optimization becomes opinion.
Define the stages before you measure anything
A funnel is only as useful as its stage definitions. For a Shopify apparel store, a workable model is five steps: session start, product detail view, add to cart, checkout initiated, order completed. Each stage must be a clean, unambiguous event you can pull from your analytics consistently.
Resist the urge to add more stages than you can act on. A 12-stage funnel looks thorough but produces drop-off numbers so small that statistical noise dominates the signal. Five to seven stages is the band where the math stays robust and each stage maps to a decision.
Split the funnel by the dimensions that actually change behaviour: device (mobile vs desktop), traffic source (paid social vs organic vs email), and new vs returning visitor. A single aggregate funnel hides the fact that mobile paid-social traffic converts at a third of the rate of desktop email — and those are different problems with different fixes.
Watch your event hygiene
If your add-to-cart event fires on quick-add buttons but not on PDP add-to-cart, your funnel will lie to you. Spend half a day in tag manager validating that each stage event fires exactly once per intended action, on every template. Bad event hygiene is the single most common reason funnel numbers don't match Shopify's order count.
Read the transitions, not the absolute rates
The number that matters at each step is the stage conversion rate — the percentage of users who move from stage N to stage N+1. Compare that to industry norms and to your own historical baseline. A 45% checkout-to-purchase rate sounds bad in isolation; if last quarter it was 38%, you're improving.
Look for the steepest cliff. If product-view to add-to-cart drops 8% but checkout-initiated to order-completed drops 55%, the checkout is where the money is. Fixing a 55% leak by even five points typically returns more revenue than a 20% lift on a higher-up stage, because you're recovering high-intent users.
Typical stage drop-off on a Shopify apparel funnel
In this pattern the PDP-to-cart step looks like the obvious leak — only 11% continue. But that's normal for apparel browsing. The real anomaly is often inside checkout: a shipping step that drops from 72% to a lower number signals a cost-shock problem, not a UI one.
Diagnose the cause of each leak
Knowing where the leak is doesn't tell you why. Once you've isolated the worst stage, layer on qualitative signals: session recordings of users who dropped at that step, on-exit surveys, and form-field analytics. The quantitative funnel narrows the search; the qualitative tools explain it.
Benchmark your stage rates against typical ranges before declaring a stage broken. The table below shows the bands you should expect on a Shopify store doing €1M–€15M in revenue. If you're already inside the top band on a given stage, the next test won't move it much — invest the effort where you're below median.
Typical stage conversion rates on Shopify (apparel & beauty, €1M–€15M revenue)
| Stage transition | Bottom quartile | Median | Top quartile |
|---|---|---|---|
| Session → Product view | 28% | 42% | 58% |
| Product view → Add to cart | 6% | 11% | 18% |
| Add to cart → Checkout start | 32% | 48% | 63% |
| Checkout start → Order complete | 38% | 55% | 72% |
| Overall session → Purchase | 0.8% | 1.6% | 3.1% |
Pair the benchmarks with segment splits. A mobile checkout completion of 42% against a 55% median is a 13-point gap — almost always a payment-method or address-form problem on small screens. Desktop sitting at 58% against the same median is a different conversation entirely.
Turn findings into a prioritised test backlog
Funnel analysis ends where funnel optimization begins: a ranked list of hypotheses, each tied to a specific stage and a specific expected lift. For every leak, write one sentence — "If we add Klarna at the payment step, checkout completion on mobile will rise from 42% to 50%" — and rank by traffic × current rate × expected lift.
Re-run the funnel analysis after every shipped test, on the same stage definitions. Drift in stage events is the silent killer of long-running CRO programs; six months in, half of teams discover their "add to cart" event has been double-firing since a theme update. Treat the funnel itself as a maintained asset.
Use historical data on day one
You don't need to wait 30 days to build a funnel — if you've had GA4 running, your stage events already exist in history. Import a year of data and you can rank leaks, segment by device and source, and ship the first test in the first week instead of the second month.
Frequently asked questions
Funnel analysis is the diagnostic step — you measure where users drop off. Funnel optimization is what you do with the findings — running A/B tests, rewriting copy, fixing forms. Analysis tells you where; optimization changes the number.
Five to seven is the sweet spot for most online stores. Fewer than five hides where leaks happen; more than seven produces stage rates so small that noise overwhelms signal. Start with session → PDP → cart → checkout → order and split checkout further only if it's your worst-performing area.
Yes — always. Mobile and desktop funnels typically differ by 2-3x at the checkout stages, and aggregating them hides device-specific problems like small-screen form friction or missing mobile wallets. Segment by device first, then by traffic source, then by new vs returning.
GA4's Funnel Exploration report is workable for top-level stage tracking, but it struggles with checkout-step granularity on Shopify and doesn't easily combine with session recordings. Most teams use GA4 for the quantitative funnel and pair it with a session-replay or analytics tool that can join the two views.
Across apparel and beauty stores, 65-75% of carts are abandoned before checkout completion. Below 60% is excellent; above 80% suggests something specific is broken — usually a shipping cost shock, a forced account creation, or a missing payment method.
Re-check the funnel weekly for active tests and at least monthly for baseline drift. Theme updates, app installs, and tag changes silently break stage events; a monthly QA on event firing catches most issues before they corrupt a quarter of decisions.
On apparel and beauty stores, yes — mobile checkout completion is typically 10-15 points below desktop. The gap closes significantly when Apple Pay, Google Pay, and Shop Pay are enabled, because they remove the address and card-entry friction that hurts mobile most.
On mid-size Shopify stores, the largest recoverable leak is almost always the shipping step inside checkout — where users see total cost including shipping for the first time. Reducing shipping-cost surprise (free-shipping thresholds shown in cart, clearer earlier signalling) consistently produces the biggest stage-rate gains.
You need a separate view, not a separate funnel. Use the same stage definitions but filter by traffic source. Paid social converts differently from organic search at the PDP-to-cart step in particular, and aggregating them gives you a misleading average that matches neither audience.
If a stage has more than 1,000 weekly transitions and sits in the bottom quartile of benchmarks, act now — the signal is reliable. Below 1,000 weekly events, give it 2-4 weeks of data before declaring a problem, because day-of-week and campaign noise can swing small samples by 20% or more.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.