How a Fashion Retailer With 1.2 Million Orders a Year Lifted Conversion by 7% and Recovered €790,000 in Returns
A large online fashion retailer. Some of our clients prefer not to be named publicly, so we keep it that way.
The challenge: customers don't follow a funnel
Most analytics setups assume a purchase journey that looks something like: land on a page, view a product, add to basket, check out. That's not how fashion shoppers behave.
This client processes around 1.2 million online orders per year with an average order value of roughly €90. Their customers compare products across categories, remove and re-add items to their basket multiple times, and often return to the site over several days before committing. Last-click attribution was telling them almost nothing about what was actually driving those purchases.
Mapping the full journey
Instead of relying on session-level conversion data, Stormly connected the full sequence: session events, basket updates, product interactions, and final Order IDs.
This made it possible to ask a different question. Not just “which sessions converted?” but “which behaviors, earlier in the session, predict whether someone will buy?” That shift in framing is where the real findings came from.
The signals that changed everything
Three behavioral patterns emerged that had strong predictive value for conversion.
Users who interacted with the product comparison feature converted at 2.8x the rate of users who didn't. Not slightly more. Nearly three times as often. The comparison tool was buried in the product page UI and most visitors never found it.
A second signal: customers who visited the accessories category at least twice within the first 14 minutes of their session were significantly more likely to complete a purchase. This wasn't something anyone had looked for. It surfaced from the behavioral data automatically.
A third pattern: customers who viewed delivery information before reaching checkout had noticeably lower drop-off rates at the checkout step. Many visitors were apparently abandoning because they didn't know what to expect on delivery costs and timing.
From insight to A/B test
Based on these findings, Stormly automatically suggested a set of A/B tests. The priorities: make the comparison feature more visible on product pages, and reduce friction around delivery information in the checkout flow.
After those changes were implemented: conversion rate increased by 7%. For a store doing 1.2 million orders a year at €90 AOV, that's a meaningful number on its own, without touching any ad spend or acquisition strategy.
Speed: the AI Agent
In most organisations, this kind of analysis takes days. Someone has to build the reports, interpret the output, decide what to test, and write it up for the team. That process usually creates a dependency on whoever owns the data stack.
Stormly's AI Agent works differently. It runs continuously in the background, identifies anomalies and opportunities without being asked, and sends recommendations with context already included: what changed, why it matters, and what should be tested next.
For this client, that translated to 6 to 15 hours saved per person per week across the product and marketing teams, while also accelerating how fast they could act on findings.
Stormly also recently released an MCP connector, which lets teams connect Stormly directly to Claude, ChatGPT, and other AI agents. Analysis that used to require a data team can now be accessed conversationally from the tools teams already use.
Returns: the hidden profit driver
Conversion rate gets most of the attention. Returns often get treated as a logistics problem. For this client, they were a significant margin leak with a very clear product analytics angle.
At the start of the engagement, 27% of all orders were being returned. One product category was responsible for nearly 18% of all returns across the store. High-value baskets had a 35% higher return likelihood than average, meaning the customers generating the most revenue were also the ones most likely to reverse it.
Finding the root causes
Because Stormly connects the full flow from purchase through return request to return outcome, it was possible to trace preventable returns to their source in the product experience.
A large share came down to three issues: unclear product descriptions, incorrect size expectations from product images and copy, and incomplete buying guidance on product detail pages. These weren't fulfilment problems. They were information problems that product and content teams could fix directly.
The result
After improving sizing guides and product detail pages for the affected categories, preventable returns dropped from 15% to 11%.
The annual recovery from that reduction alone: approximately €790,000. That figure doesn't include the reduction in logistics handling, customer service load, or the operational cost of processing returned inventory.
Combined with the 7% conversion lift, the total impact came from product analytics work, not from changing any acquisition channel or increasing ad spend.
See what Stormly finds in your store.
Book a Demo