CRO Diary · a conversion specialist's working notebook entries reviewed weekly
CRO Diary

More sales from the traffic you already have

Notes

Reading Your Funnel in GA4 Without Losing Your Mind

GA4 already tracks a nine event path from product list to purchase. The sanity trick is ignoring most of it: one funnel exploration, five steps, closed mode, compared by device. Here is my exact setup.

From the diary · CRO Diary  ·  July 28, 2026  ·  5 min read  ·  AI-drafted from sourced data

A hand sketching a funnel diagram in a paper notebook with a laptop showing an analytics dashboard blurred in the background
Photo via Openverse (CC0)

I have watched more than one store owner open GA4, click through three menus, land on a report full of event names in snake_case, and close the laptop with the expression of someone who has just been handed tax forms in a foreign language. I sympathize. GA4 is a database wearing a dashboard costume. But somewhere inside it there is one view that pays for all the suffering, the funnel exploration, and this note is the exact way I set it up so it produces a diagnosis instead of a headache.

The funnel is already there

The part almost nobody tells beginners: if your store runs a standard GA4 ecommerce integration, the shopping path is already being recorded as a chain of named events. The schema goes view_item_list, select_item, view_item, add_to_cart, view_cart, begin_checkout, add_shipping_info, add_payment_info, purchase. Nine checkpoints from a product appearing in a list to money changing hands, each carrying an items array and a value.

You do not need new tracking, new tools, or a consultant to draw this. You need one exploration and a few deliberate choices. The choices are where sanity is won or lost.

My setup, choice by choice

Five steps, not ten. GA4’s funnel exploration allows up to 10 steps, and using all of them is how you get a chart that resembles a staircase in fog. I build with five: view_item, add_to_cart, begin_checkout, add_payment_info, purchase. Each gap between those steps maps to one thing I can actually fix: the product page, the cart-to-checkout handoff, the checkout form itself, and the final payment moment. A drop I cannot act on is decoration, so the extra events stay out.

Closed, not open. GA4 offers open funnels, where users can enter at any step, and closed, where they must start at step one. For diagnosis I always choose closed. It answers a clean question: of the people who saw a product, how many advanced through each gate in order? Open funnels flatter the middle steps with side entrances and turn every drop into a maybe.

Broken down by device. One dimension, device category, applied as a breakdown. As I wrote in the mobile-first note, the benchmarks put three quarters of traffic on phones converting far below desktop, and the funnel is where that abstraction becomes a location. The step where the mobile line sags below the desktop line, relative to its own start, is usually the audit’s destination. GA4 also allows up to four segments if you want to compare, say, paid against organic, but I add those on the second pass, not the first.

Elapsed time, on. A toggle most people never find. It shows average time between steps, and it changes readings. A long pause between add_to_cart and begin_checkout is not always a leak; for considered purchases it is often just deliberation. A long pause between begin_checkout and add_payment_info, though, means the form is making people think, and forms should never make people think.

Next action, for the exits. For any step with an ugly drop, the next action breakdown shows what users did instead of advancing. Did they bounce, go back to the collection page, or wander to the shipping policy? That last pattern is one of my favorite tells: people hunting for the shipping answer mid checkout are announcing which surprise made them leave.

THE FIVE-STEP FUNNEL I ACTUALLY BUILD GA4 FUNNEL EXPLORATION, CLOSED MODE, BY DEVICE 1 · view_item 2 · add_to_cart 3 · begin_checkout 4 · add_payment_info 5 · purchase ANCHOR: 4.6% OF SESSIONS ADD TO CART ANCHOR: 45% OF CHECKOUTS COMPLETE BAR WIDTHS ILLUSTRATIVE · ANCHORS: LITTLEDATA, 2,800 SITES (SHOPIFY AVG) EVENTS: DEVELOPERS.GOOGLE.COM GA4 ECOMMERCE SCHEMA
Five gates, each mapped to one fixable thing. Anything more detailed goes on the second pass, after this view has pointed somewhere.

Reading it without benchmarks is a mood

A funnel on its own tells you that numbers get smaller from left to right, which you knew. The reading starts when each step is held against an external anchor, and the public ones I use come from Littledata’s benchmark of 2,800 ecommerce sites: average Shopify add to cart rate of 4.6% of sessions, checkout completion averaging 45%, and overall conversion at 1.4%. The best 10% of stores complete more than 66% of checkouts, so there is real headroom above the average.

The method is comparative, not absolute. If your add to cart rate is near benchmark but your checkout completion is 30% against an average of 45%, the store has a checkout problem and the product pages are innocent. If add to cart sits at 2% while completion is fine, the product page or the traffic quality is the suspect and rebuilding checkout would be a month of wasted effort. One funnel, two anchors, and the argument about what to fix first usually ends there.

The funnel does not tell you what is wrong. It tells you where to look, which in an audit is nine tenths of the job.

The sanity rules

Three habits keep this view honest. First, date ranges long enough to mean something; on a small store, a week of funnel data is an anecdote with gridlines, so I default to 28 days minimum. Second, accept that GA4 undercounts. Consent banners and ad blockers thin the data, and the funnel counts users completing sequences, not orders, so the proportions are trustworthy while the absolute numbers will never match your platform dashboard. Verify revenue against the platform, read shapes in GA4. Third, resist rebuilding the exploration weekly. The funnel is a thermometer, and a thermometer you keep redesigning tells you nothing about the patient.

Set up once, screenshot monthly, compare against the last one. The month the shape changes at a specific step, you have a lead worth an afternoon. That is the entire practice, and it costs nothing but the twenty minutes of setup this note just walked through.

Questions I get about this

How do I build an ecommerce funnel in GA4?

Open Explore, choose the Funnel exploration template, and define steps from the standard ecommerce events: view_item, add_to_cart, begin_checkout, add_payment_info, purchase. GA4 allows up to 10 steps, shows completion and abandonment at each one, and lets you break the funnel down by a dimension like device category.

Should my GA4 funnel be open or closed?

Closed, for diagnosis. A closed funnel only counts users who enter at step one and move through in order, so each drop is a real reading of that transition. Open funnels let people enter mid path, which is more flattering and much harder to interpret.

What is a normal drop between funnel steps?

Public benchmarks give rough anchors rather than rules. In Littledata's 2,800 site benchmark the average Shopify add to cart rate is 4.6% of sessions, checkout completion averages 45%, and overall conversion is 1.4%. If your funnel deviates far from one of those anchors at one specific step, that step is where the audit starts.

Why do my GA4 funnel numbers look different from my Shopify dashboard?

Different counting. GA4 funnels count users completing event sequences within the exploration's timeframe, while platform dashboards typically count sessions or orders directly. Consent banners, ad blockers, and tagging gaps also thin GA4's view of reality. Treat GA4 funnels as a map of proportions, not a ledger of absolute truth, and verify order counts against the platform.

Sources & data

  1. Google Analytics Help, GA4 Funnel explorations: steps, open vs closed funnels, elapsed time, next action (accessed July 2026)
  2. Google Analytics developer documentation, Measure ecommerce: recommended event sequence from view_item_list to purchase (accessed July 2026)
  3. Littledata, add to cart and checkout completion benchmarks from 2,800 ecommerce sites (2023 benchmark, accessed July 2026)
Cite this entry: CRO Diary (2026). “Reading Your Funnel in GA4 Without Losing Your Mind.” https://crodiary.com/notes/reading-your-ga4-funnel/