Money leaks quietly
Your funnel is simple to draw: a visitor lands, views a product, adds to cart, starts checkout, pays, and buys. Every step loses some people — that’s normal. What isn’t normal is a step that suddenly loses more people than yesterday, for a reason you can’t see. That’s a leak, and leaks don’t announce themselves. Revenue just quietly comes in lighter.
Finding the leak is a three-part job: which step, which segment, and what in the store caused it. The chart gives you the first. The other two are why this is one of the first jobs stores hand to e-commerce autopilot: the answer lives in the analytics and the store at the same time.
The questions every owner asks
- Where exactly are people dropping off?
- Is checkout broken — or broken only for some people?
- Is one browser or device failing silently?
- Did yesterday’s deploy hurt conversion?
- Did a payment provider start declining cards?
A funnel chart shows you the drop. It almost never shows you the segment where the drop concentrates — and that segment is where the money is.
The leak is usually in a segment, not the average
Averages hide leaks. Here’s a real signal from a live analytics property: over the last 28 days, desktop sessions engaged at 34% while mobile engaged at 61%. If you only watched the blended number, you’d never notice that one surface was behaving completely differently from the other. Now imagine that gap is a checkout failure instead of an engagement difference — invisible in the average, obvious the moment you split by device.
The answer an owner actually wants sounds like this:
Cart abandonment rose 14% yesterday because Apple Pay failed for Safari users in Germany after the 3pm deploy. Every other segment is normal.
That single sentence turns a vague “conversion is down” into a specific, fixable bug — often thousands of euros of recovered revenue.
The six steps, and what a leak at each one usually means
GA4 measures an e-commerce funnel as six events. Read it as six steps, and each drop points somewhere different:
| Step | Event | A drop here usually means |
|---|---|---|
| 1 | View item | Traffic quality: the campaign or query is sending the wrong people |
| 2 | Add to cart | The product page: price, missing answers, no reviews, out of stock |
| 3 | Begin checkout | Cart friction: shipping estimate, minimums, forced accounts |
| 4 | Add shipping info | Shipping cost or delivery time surprise, address validation |
| 5 | Add payment info | Payment methods missing for that country or device, wallet failures |
| 6 | Purchase | Gateway declines, 3-D Secure failures, a broken confirmation |
Two rules of thumb. A step showing zero users does not mean nobody got there; it means the shop is not sending that event, and fixing the tracking comes before any CRO work. And the step with the biggest drop is rarely the step with the problem’s cause: a payment-step drop that only affects one country is a payment-method gap, not a checkout bug.
Find the biggest drop, then the worst segment
The tool for this is the GA4 checkout funnel read as decision-ready rows: users per step, conversion and drop-off between steps, each step’s share of the first, the previous period alongside for comparison, and the biggest drop named. With a breakdown by device, country, traffic source or landing page it also returns the funnel per segment and names the worst-converting segment among those with enough users to trust, so a segment of nine visitors cannot masquerade as a crisis.
In datavessel this is one question, in chat or as a scheduled run:
“Show me the checkout funnel for the last 14 days against the 14 before, broken down by device. Which step lost the most, and where?”
The answer is not a graph you interpret. It is the step, the segment, and the size of the gap, which is exactly the sentence from the section above, produced on demand.
Find the leak by asking
datavessel reads your funnel across GA4 and your store and lets you slice the drop by the dimensions that matter, in plain language:
- “Where in the funnel did drop-off increase versus last week?”
- “Break checkout completion down by device, browser, and country.”
- “Did conversion change after yesterday’s deploy?”
Instead of a graph you interpret, you get the segment that’s bleeding. And scheduled to run each morning, it catches a broken payment path the day it breaks — not at the end of the month when the revenue report finally looks off.
Check the cause in the store, not the chart
Once the step and segment are named, the diagnosis needs data the analytics tool does not have. Because the store, Merchant Center, the ad accounts and the reviews sit in the same connector as GA4, the follow-up questions are one conversation:
- Stock. Are the products on the losing pages in stock, at variant level? A sold-out bestseller is the most common “conversion problem” that is nothing of the kind.
- Price. Merchant Center’s price insights give your price against Google’s market benchmark and the predicted change in conversions if you moved it. A product ten percent above market with a low conversion rate is a pricing leak, not a page leak.
- The product page. What the storefront actually renders for that product: description, images, the questions reviews keep asking that the page does not answer.
- Traffic. Which campaigns sent the segment there, and whether they are still sending it to a page that cannot convert. The pairing with ads is in connect Meta Ads to ChatGPT or Claude and stop ads to out-of-stock products.
Fix it with an approval, then keep watching
The fix follows the cause: a rewritten description, a price change, a corrected shipping note, a paused ad set, a promotion. Each arrives as a proposal with the exact change and the reasoning, on web or mobile, and nothing lands until you approve it. The decision goes into a ledger, so next quarter’s “why did we change this” has an answer.
Then take it off your calendar. A conversion alert agent runs on a schedule and fires when traffic is stable but conversion drops beyond a threshold, which is the signature of a checkout or site issue, and posts the numbers to Slack. The funnel read itself runs weekly. What still abandons after the fixes is handled by abandoned cart automation, and the whole loop, funnel to fix to recovery, is what an AI CRO agent does.
Frequently asked questions
What is a good checkout conversion rate?
Benchmarks vary so much by category, price point and traffic mix that the only useful comparison is your own previous period, which is why the funnel read always returns it. A step that converts worse than it did last month, for one segment, is the signal; the industry average is not.
Why does my funnel show zero users at a step?
The shop is not sending that GA4 event. Add-shipping and add-payment are the ones most often missing. Fix the tracking first; a funnel with a silent step cannot locate a leak.
Which segments should I break the funnel down by?
Device first, then country, then traffic source and landing page. Device and country catch payment and shipping failures; source and landing page catch traffic-quality problems that look like page problems.
Can this find a failing payment provider?
Yes, when the failure is segment-shaped: a drop at the payment step concentrated in one device or country over a short window. It cannot read the gateway’s logs; it shows you where and when to look.
Does it work for Shopify, Shopware and WooCommerce?
Yes. The funnel comes from GA4 for all three. The store-side checks and the fixes use each platform’s tools: inventory and products on all three, CMS and landing pages on Shopware, WordPress content on WooCommerce.
The takeaway
A funnel chart tells you people are leaving. It rarely tells you which people, where, and why now — and those three answers are the whole game. The leaks that cost the most are the ones hiding inside an average. Split the funnel, ask why, and stop the bleed while it’s still small.
Watch an agent find the leaking step, name the segment and propose the fix: see it live →
Leave a Reply