Take a store doing 100,000 sessions a month at a 2 percent conversion rate. That is 2,000 orders. The instinct is to treat the 2 percent as the problem and go hunting for a sitewide fix - new theme, faster pages, another popup. But 2 percent is an average of five different survival rates stacked on top of each other, and averages hide exactly the thing you need to see.
Break the same 100,000 sessions into stages and the picture changes. Say 47,000 reach a product view, 4,200 add to cart, 1,600 begin checkout, and 2,000 finally purchase (paid traffic and returning carts refill the bottom). Now do the arithmetic step by step. Product view to add to cart is running at about 9 percent, which is fine. But add to cart to begin checkout is running at 38 percent, and the drop from cart to checkout is where roughly 2,600 intent-rich shoppers vanish every month. That is not a conversion-rate problem. That is one leaking step, and it is worth far more than a new theme.
That is what ecommerce funnel analysis does. It stops you from optimizing the average and points you at the single step where the money is actually leaking out.
Funnel analysis is a sequence, not a number
An ecommerce funnel is an ordered set of events, and the whole method depends on those events being tracked cleanly. Google's own ecommerce framework defines the canonical sequence: view_item_list, view_item, add_to_cart, begin_checkout, add_shipping_info, add_payment_info, purchase. Each of those is a recommended GA4 event with a defined shape, and each transition between two of them is a place where shoppers either continue or quit.
Most day-to-day analysis collapses that into five decision points:
- Session to product view - did the visitor engage with a product at all?
- Product view to add to cart - did the product page earn intent?
- Add to cart to begin checkout - did the cart survive contact with reality?
- Begin checkout to add payment - did the checkout form itself hold them?
- Payment to purchase - did anything break at the last step?
The reason to keep these separate is simple: a store can have a perfectly normal add-to-cart rate and a catastrophic checkout, and the blended conversion rate will look merely "a bit low." Only the stage view tells you which step to fix.
What "normal" looks like at each step
You cannot read a drop-off without a reference for what a healthy transition looks like. The ranges below are rough industry orientation, not targets - treat them as illustrative, and treat your own trailing three-month numbers as the real benchmark.
| Funnel step | Illustrative healthy range | What a bad number usually means |
|---|---|---|
| Session to product view | ~45-50% | Wrong traffic, or a homepage that hides products |
| Product view to add to cart | ~8-10% | Weak PDP: price, images, or a buried cart button |
| Add to cart to begin checkout | ~30-35% | Surprise costs, or a cart nobody can find again |
| Checkout to completed purchase | ~45-55% | Friction, forced accounts, thin payment options |
Across the whole checkout process, documented cart abandonment averages close to 70 percent - so a large drop between cart and purchase is normal, and the analytical job is to separate the normal portion from the fixable portion. The point of the table is not to hit the numbers. It is to spot the one row where you sit far below the range, because that row is your leak.
Reading the drop-offs: find the one step that is off
With the funnel laid out and the benchmarks next to it, the read is quick. Do not look for the step with the biggest absolute drop - the biggest drop is almost always the top of the funnel, where you have the most people. Look for the step that is furthest below its own benchmark. A 4 percent product-view-to-cart rate against a 9 percent norm is a bigger, more recoverable problem than a 40 percent checkout completion against a 50 percent norm, even though the checkout loses more raw sessions.
This is also where funnel analysis and a proper conversion rate optimization audit meet: the funnel tells you which step to investigate, and the audit tells you why that step is failing before anyone proposes an A/B test on it. Measuring the leak comes first; guessing at fixes comes never.
Two real leaks, and what they were worth
Benchmarks tell you where to look. The reason to look is that leaks are almost always concrete and boring once you find them. Two examples from delivered analytics audits, anonymized:
In a high-average-order-value Shopify store (around $460 per order), the funnel showed a healthy add-to-cart rate and then a wall: 63 percent of add-to-cart sessions abandoned at the shipping step. The cause was not psychological. Shipping rates were hidden behind a three-field calculator that most shoppers never completed, so the true cost only appeared late and felt like a surprise. Modeling a modest ~15 percent recovery of that abandoned intent was worth roughly $8,000 a month at that order value. One step, one cause, a five-figure annual number.
In the same store, the top of the funnel had its own leak: product view to add to cart ran at about 4 percent, less than half a healthy rate. The reason showed up only when the funnel was segmented by device. Around 73 percent of product views happened on mobile, and the mobile fold was occupied by a 130-option currency selector and a first-paint email popup - the add-to-cart button was pushed below two obstacles before anyone could act on intent.
Building the funnel in GA4
You do not need a data warehouse to start. GA4's Funnel exploration report builds the whole thing from the standard ecommerce events. Open the Explore section, choose Funnel exploration, and add your steps in order: view_item, add_to_cart, begin_checkout, add_payment_info, purchase. Set the funnel to closed if you only want strictly sequential journeys, or open if you want to count people who jump in mid-way.
The step that makes or breaks the analysis is the breakdown. Add device category as a breakdown dimension and the single funnel splits into mobile, desktop, and tablet - which is how the currency-selector leak above went from invisible to obvious. Traffic source is the other breakdown worth running, because a funnel that is healthy on email and broken on paid social is a targeting or landing problem wearing a conversion-rate costume.
All of this rests on the events being correct, which is why funnel work and Shopify conversion tracking that survives checkout changes are the same project from two angles. If begin_checkout or add_payment_info never fire, GA4 will draw you a funnel with a step missing and you will misread the leak entirely.
The funnel-analysis checklist
Run these in order. Each one is a gate: do not move to the next until the current one is clean.
- Confirm the events fire once and in order. Watch
view_item,add_to_cart,begin_checkout,add_payment_info, andpurchasein GA4 DebugView on a real test purchase. No duplicates, no missing steps. - Build the closed funnel in Funnel exploration with those five steps.
- Calculate step-to-step completion, not just overall. Write the percentage next to each transition.
- Compare each step to its benchmark and circle the one furthest below range. That is your leak, regardless of raw volume.
- Segment the leaking step by device category and traffic source. The cause usually hides in one segment.
- Quantify the recovery in revenue, not percentage points: abandoned sessions at that step, times a realistic recovery rate, times average order value.
- Fix that one step, then re-measure the same funnel before touching anything else.
If you would rather have the funnel built, read, and quantified for you - with the leak named and the recovery modeled in real money - that is the core of a product analytics engagement, and you can see the shape of that work in a sample tracking audit report before committing to a full analytics audit of your own store.
The discipline is the whole point. A conversion rate is a verdict with no defendant. A funnel is a list of suspects with one of them clearly guilty. Break the journey into steps, find the step that is off, put a dollar figure on it, and fix that step first. Everything else on the optimization list can wait until you know where the money is actually leaking.
