A large retailer used Augmeta to connect paid traffic, unavailable products and a mobile conversion decline into one explanation, and a practical change to test.
Between 7% and 9% of a large retailer's product-page visits, up to 282,000 sessions a week at the peak, landed on items that could not be bought, because automated bidding was buying clicks on items that had just gone out of stock and the retailer's own category pages showed stale availability. The first fix, fresher availability on those category pages, addressed about 30% of affected sessions and was live two days after the investigation started.
Between 7% and 9% of the retailer’s product-page visits were landing on items that were fully unavailable: 282,000 sessions a week at the peak. The rate had been stable for months, no alarm fired on it, and it was owned by merchandising as a chronic inventory problem.
Meanwhile the metric leadership actually watched, mobile-web add-to-cart rate, was falling, and was being investigated as a page-design problem. The two were never the same conversation. The unavailable-page rate was an inventory metric, the add-to-cart decline was a product metric, and the automated bidding that connected them was a marketing system. Three functions, three dashboards, no shared view, and no anomaly in any single one of them. It is the first of the five places in our guide to finding revenue leaks on a large retail website.
The Operator reframed the question, from "why is the unavailable-page rate high" to "why is mobile-web add-to-cart falling," and followed it across traffic sources, product availability and device behaviour.
The chain:
Most of the affected traffic came from two places: paid search, and the retailer's own category pages, whose availability information lagged the product pages. That second share was the part the retailer controlled directly.
Two of the retailer's own product managers independently reproduced the analysis from their own tools.
A parallel analysis suggested the same paid channel explained an entire year-over-year decline in desktop cart-to-checkout. Before it went anywhere, a check of the channel’s history found that two paid channels had swapped a large volume of visits between them in a single month: a reporting reclassification, not a real shift in media. The analysis was parked pending validation with the media team rather than delivered. A conclusion that would have been wrong for a taxonomy reason never reached the customer.
The work split into two tracks.
Track 1: the retailer's own pages. The team shipped fresher inventory information on category pages, so shoppers stop being routed from the retailer's own pages to items they cannot buy. It addressed about 30% of affected sessions, and it was live two days after the investigation started.
Track 2: the paid media feed. The feed driving automated bidding was updated with current inventory, so the bidding system stops buying clicks on items that have just gone out of stock. [status]
A persistent issue that three functions had each half-explained now had one shared explanation and two targeted changes. Issues like this are why KPI war rooms happen. The conversion and revenue effect was still awaiting measurement when this was written, and it is not presented here as a closed loop. What is demonstrated is the diagnosis, the speed of the first fix, and a stated model of what is recoverable: those sessions converting at the site average instead of at zero.
See how AI Operators run retail and ecommerce KPIs.
Why did nobody catch the leak earlier?
The unavailable-page rate had been stable for months, so no alarm fired on it, and it was owned by merchandising as a chronic inventory problem. The decline leadership watched, mobile-web add-to-cart rate, was being investigated as a page-design problem. Three functions had three dashboards and no shared view.
What caused it?
Automated bidding bought clicks on items that had just gone out of stock, those clicks landed on product pages with nothing to buy, and mobile-web visitors recovered from that dead end far less often than desktop visitors did.
Has the revenue effect been measured?
Not yet at the time of writing. The conversion and revenue effect was still awaiting measurement, and the case study does not present it as a closed loop. What it demonstrates is the diagnosis, the speed of the first fix and a stated model of what is recoverable: those sessions converting at the site average instead of at zero.
Bring one revenue-critical KPI. We'll show you the causal chain in days, not quarters.