How to use Drop-Off Analysis
Drop-off analysis pinpoints the funnel stage bleeding the most users so you fix the real leak, not the one intuition suggests. Here's how to run it, what numbers to expect, and how to turn findings into tests.
Drop-Off Analysis
A funnel diagnostic that identifies which step loses the highest percentage of users, so optimisation effort targets the real leak.
Drop-off analysis is the process of measuring the user loss between consecutive steps in a conversion funnel — product view to add-to-cart, cart to checkout, checkout to purchase — and ranking those gaps by severity. The stage with the steepest drop is where revenue is leaking fastest, and where a single fix typically returns the most.
It is the first diagnostic in any CRO programme because it answers the only question that matters before you start testing: where is the problem actually located? Without it, teams test homepage hero variants while the real loss is happening on a broken mobile shipping form.
Most CRO programmes fail not because the tests are bad, but because they're aimed at the wrong stage. You can't out-test a checkout that loses 70% of shoppers by polishing the product page above it.
Drop-off analysis sits inside the broader practice of funnel analytics, but it's the specific lens that turns a funnel chart into a prioritised list of problems. The output isn't a number — it's a target.
How to run a drop-off analysis
Start by defining the funnel as the user actually experiences it, not as your analytics platform defaults to. For a Shopify apparel store that usually means: landing → product view → add-to-cart → checkout start → shipping → payment → purchase. Seven steps, not three.
Then calculate two numbers for each transition: the step conversion rate (users who continued ÷ users who reached this step) and the absolute drop-off (raw users lost). You need both — a 40% drop on a stage 100 people reach is smaller than a 20% drop on a stage 10,000 people reach.
Segment everything by device, traffic source, and new vs returning. A 65% overall checkout conversion can hide a 35% mobile-paid-social conversion that's dragging the whole funnel down. Aggregates lie; segments tell you where to look.
Don't trust your highest-traffic step
Teams instinctively optimise the page with the most pageviews. But drop-off analysis often shows the worst leak is two steps deeper — on a page only 8% of visitors ever see. That's where the test-impact-per-effort is highest.
What drop-off patterns typically look like
Across e-commerce funnels, the drop-off curve is rarely linear. You usually see a soft loss at product browsing, a sharp cliff at add-to-cart, a second cliff at checkout-start (where guest vs account friction hits), and a slower bleed through the payment stages.
The two cliffs are where intent meets friction. A visitor who adds to cart has declared interest; if they abandon before payment, the cause is almost always operational — unexpected shipping costs, forced account creation, or a slow mobile form. Those are fixable in a sprint, not a quarter.
Typical user retention across a 7-step e-commerce funnel
Notice where the steepest relative drops happen: landing → product (58% loss) and product → add-to-cart (74% loss of those who viewed). Late-funnel stages lose smaller percentages but each lost user is far more valuable, because acquisition cost is already sunk.
Benchmarks: where the average leak sits
Benchmarks aren't targets — they're sanity checks. If your checkout-start to purchase conversion is 45% and the vertical median is 65%, you've found your leak without needing another query. If you're already at 70%, the marginal point is much harder to win and you should test elsewhere.
The table below shows typical step-conversion ranges across three common online-retail verticals. Pay attention to add-to-cart and checkout-completion: those are the two stages where vertical norms diverge most and where most measurable wins live.
Median step-conversion rates by vertical (mobile + desktop blended)
| Vertical | View → Add to cart | Add to cart → Checkout start | Checkout start → Purchase | Overall (landing → purchase) |
|---|---|---|---|---|
| Apparel & accessories | 8-12% | 55-65% | 60-70% | 1.8-2.6% |
| Beauty & personal care | 10-14% | 60-70% | 65-75% | 2.4-3.2% |
| Consumer electronics | 5-8% | 45-55% | 55-65% | 0.9-1.6% |
| Home & furniture | 6-9% | 50-60% | 50-60% | 1.1-1.8% |
| Food & beverage (DTC) | 12-16% | 65-75% | 70-80% | 2.8-3.6% |
If a beauty store sees a 45% checkout-start-to-purchase rate, that's roughly 20 percentage points below median — almost always a payment or shipping issue. If an electronics store sees the same 45%, it's bang on average and the leak is somewhere else.
Turning findings into experiments
Once you've ranked the leaks, the next move is hypothesis generation — and this is where most teams stall. A drop-off number tells you where, not why. To get to a testable hypothesis you need behavioural data: session replays, form analytics, and exit-survey responses for the failing stage.
Prioritise tests by expected revenue impact, not by drop-off size alone. A 5-point lift on a stage that processes €40k/month of intent is worth more than a 15-point lift on a stage handling €8k. The formula is simple: stage traffic × current value × expected lift. Rank, then test top-down.
The 90-day rule
If a drop-off persists for 90 days across multiple segments, it's almost always a structural problem (broken form, payment method gap, mobile bug) — not a copy or design issue. Audit the page mechanically before designing a test.
Frequently asked questions
Funnel analytics is the broader practice of measuring user flow through a defined sequence of steps. Drop-off analysis is one specific output of that practice — ranking the stage-by-stage losses to find where the biggest leak is. You do funnel analytics to enable drop-off analysis.
Five to eight steps is the sweet spot for e-commerce. Fewer and you miss the transitions where friction actually lives (e.g. shipping → payment); more and your stage sample sizes get too small to draw conclusions from. Map the steps the user actually experiences, not the technical pageviews.
Aim for at least 1,000 users entering the funnel per segment per week. Below that, week-to-week variance can swamp the signal and you'll chase noise. If volume is lower, widen the analysis window to a full month before drawing conclusions.
Both, ranked separately. Relative drop-off (% lost at this step) tells you where friction is highest. Absolute users lost tells you where revenue impact is highest. The stages that appear in the top three of both lists are your priority targets.
Mobile checkout typically converts 30-50% lower than desktop because of smaller forms, slower autocomplete, and intent-mixing (people browse on mobile, buy on desktop). Don't compare them like-for-like — analyse each as its own funnel with its own benchmark.
Continuously at a dashboard level, and deeply once a quarter. A weekly glance catches sudden regressions (a deploy broke checkout); a quarterly deep-dive catches slow drifts and lets you re-prioritise the test roadmap based on which stages have improved and which haven't.
For online retail, blended site conversion sits around 1.5-3% depending on vertical and traffic mix. But aggregate conversion is a vanity number — the useful target is each step's conversion against its vertical benchmark. A store can have 2% overall conversion and still have a fixable 20-point gap on one stage.
Yes, but you'll be flying blind for the first few weeks. Importing historical analytics (e.g. backfilling GA4 data) gives you immediate baselines and seasonality context, which is the difference between guessing your problem area on day one and knowing it.
Compare against three reference points: your own historical baseline for that step, your segment-level benchmark, and the vertical norm. A leak shows up as a meaningful gap (typically 10+ percentage points) against at least two of those three references. One-off gaps are usually noise.
At minimum, a funnel-capable analytics tool (GA4, an integrated CRO platform, or product analytics) plus session-replay for the failing stages. Heatmaps help on landing and product pages; form analytics is essential for checkout. The diagnostic itself is platform-agnostic — what matters is segmentable funnel data.
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.