Conversion funnel analysis: how to find where your store loses shoppers
A conversion funnel is the path from landing on your store to placing an order. Funnel analysis measures how many shoppers survive each step, so you can see which one loses the most. This guide covers the stages, how to run the analysis properly, and the leaks we find most often — with figures from sessions we have actually read, rather than the illustrative numbers these guides usually carry.
What a conversion funnel actually measures
Every shopper who buys from you passes through the same handful of steps. Most shoppers do not finish. The funnel is simply a count of how many are left at each step, and the analysis is the search for the step where the count falls hardest.
The useful part is not the total conversion rate. It is the shape — one step that loses far more than the steps around it is a defect. A funnel that declines evenly is not broken; it is a store whose offer, price or traffic is the constraint, and no amount of button-testing will move it.
The ecommerce funnel, stage by stage
Landing
They arrive — from an ad, a search, a link. Most funnel analyses start counting here, which is why most of them miss the biggest leak.
Product view
They look at something specific. This is the first moment you know what they came for.
Add to cart
The first real commitment. A drop here is usually about the product page — price, photos, sizing, trust.
Checkout started
They have decided to buy. Anything lost after this point was lost by your checkout, not your product.
Payment
Card details entered. Drops here are mechanical — a declined card, a validation error, a form that fights a phone keyboard.
Purchase
The only step that pays for the other five.
What we see in real sessions
The third figure is the one worth sitting with. About half of all sessions end on the page they started on — those shoppers never entered the funnel at all. A funnel analysis that begins counting at “product view” is measuring the half who already stayed, and is structurally blind to the larger half who left immediately.
The second figure changes how you should read every other number. When roughly two in three sessions happen on a phone, a desktop-only check of your own checkout is not a test of your checkout. Most of the mechanical leaks below only exist on mobile, which is exactly why they survive so long — the people who could fix them keep looking at the version that works.
These are aggregate figures across every store we record, and none of them identify a store. They are small numbers by the standards of a large analytics vendor, and we would rather publish a real one than an impressive invented one.
How to run a funnel analysis
- Define the steps before you look at the data. Deciding what counts as a step after seeing the numbers is how you end up with a funnel that confirms whatever you already believed.
- Count sessions, not events. One shopper reloading checkout four times is one shopper, not four. Event counts flatter the steps people repeat.
- Work in rates, not totals. Step-to-step survival is the only number that compares across traffic volumes.
- Find the worst step, then segment it. Device, country, browser, new versus returning. A step that is average overall is frequently catastrophic for one of these and fine for the rest.
- Check the difference is real before you act. A step with 40 sessions behind it can move ten points on noise alone.
- Then go and look. Everything above tells you where. None of it tells you why, and the why is the only part you can fix.
The five leaks we find most often
The cost that appears late
Shipping, tax or a fee that shows up at the last step. The shopper did the maths on the product page and your checkout changed the answer. This is the most common checkout leak we see, and it looks identical in the data to a payment failure — the difference is only visible in the session.
The form that fights the phone
A field that opens the wrong keyboard, a postcode validator that rejects a valid format, an autofill that fills the wrong box. On desktop it is invisible. On mobile — where most of your traffic is — it ends the session.
The answer that isn't on the page
Sizing, delivery time, returns. The shopper scrolls, hunts, and leaves. It reads as a product-page problem, but the fix is usually one line of copy, not a redesign.
The error nobody sees
A failed request, a coupon that silently does nothing, a button that does not respond on one browser. Your analytics record it as a normal exit, because from the server's point of view nothing happened at all.
The step that is fine everywhere except one segment
The overall number looks healthy because the majority pulls it up. One device, one country, one browser is failing badly, and the average hides it. This is why a funnel with no segmentation is a funnel with no findings.
Why the numbers rarely give you the reason
Analytics is a counting instrument. It records that a step was not reached — never what was on the screen when the shopper gave up. Two shoppers who both abandon at checkout can be leaving for opposite reasons, and in the data they are identical.
This is the honest limit of funnel analysis on its own, and it is worth stating plainly because most guides on this topic do not: the funnel is a very good way of finding where to look, and no way at all of finding out why. The why is in the behaviour — the hesitation, the scroll back up, the field filled three times.
Common questions
What is a conversion funnel?
A conversion funnel is the sequence of steps a shopper takes from arriving on your store to completing a purchase — typically landing, product view, add to cart, checkout, payment, purchase. Funnel analysis measures how many people make it from each step to the next, so you can see which step loses the most people.
What is a good conversion funnel drop-off rate?
There is no single benchmark worth chasing, because the number depends on your traffic mix, price point and category. A more useful question is whether one step drops far harder than the others, and whether that step behaves differently for one segment — a phone, a country, a browser. A relative gap tells you where to look; an industry average does not.
How is funnel analysis different from conversion rate optimisation?
Funnel analysis tells you where you lose people. Conversion rate optimisation is what you do about it. The analysis is diagnosis, the optimisation is treatment — and doing the second without the first is how stores end up A/B testing a button on a page that was never the problem.
Why does my analytics show the drop but not the reason?
Analytics counts events. It can tell you that 60% of shoppers left at checkout, but not that a shipping cost appeared, or that a postcode field rejected a valid entry. The reason lives in what the shopper actually did between the two events, which is behaviour, not a count.
How many sessions do I need before funnel analysis is useful?
Enough that a difference between steps is bigger than the noise. In practice a low-traffic store can read a badly broken step within days, because a genuine break shows up as a near-total drop rather than a few percentage points. Small differences on small traffic are where people fool themselves.
Can I do funnel analysis in Google Analytics?
You can build the funnel and see the drop-offs, yes. What you cannot see is why — GA4 records that a step was not reached, not what happened on the screen before the shopper gave up. That gap is the reason most funnel reports end in a guess.
See where your own store leaks
We walk your live store the way a shopper would and send you what we find — the step that loses the most, why, and the cheapest thing to change first. Free, no account, no code.
Run a free store audit