Increased average monthly classification accuracy from ~85% to 95%+.
Significantly reduced merchant tickets tied to inaccurate and misleading classifications.
Freed up their team to drive more merchant growth.
Fruugo is a UK-based marketplace on a mission to create a truly global platform where retailers can connect with shoppers worldwide. With automatic translation across more than 25 languages, support for over 30 major currencies, and a dedicated account manager assigned to every seller, Fruugo removes the friction that typically keeps retailers from selling internationally.
But as Fruugo's merchant base grew, one part of the onboarding journey kept slowing things down: product classification.
"Product classification was a sticking point during onboarding for merchants with large catalogues," said Calum McDonald, Product Owner (catalogue) at Fruugo. "Manual classification proved ineffective at scale, which meant their integration timeframes took longer than expected and delayed their launch on our platform."
The downstream effects went well beyond onboarding. Inaccurate classification quietly degraded the customer experience — surfacing the wrong products in sort, filter, and search results — and, critically, hurt advertising effectiveness for merchants who depended on Fruugo to amplify their reach.
Manual classification proved ineffective at scale, which meant their integration timeframes took longer than expected and delayed their launch on our platform. ”
It was also costing the Fruugo team time they wanted back. "Our Integration Specialists spent longer helping merchants understand the Google Shopping taxonomy, validating classifications within catalogues, and in some cases manually updating feeds," McDonald said. "Our Account Managers spent more time dealing with classification-related queries, spending less time on performance and growth."
Fruugo first engaged Feedonomics in late 2022 and ran a proof of concept in early 2023 before committing to a longer-term agreement.
"Feedonomics were able to demonstrate the effectiveness of their FeedAI solution by providing highly accurate classifications against a range of datasets," McDonald said. "We were also encouraged by their collaborative approach to improving outcomes by analysing and sharing insights about possible root causes within our catalogue data — merchant behaviour, keyword stuffing, additional datapoints, and so on."
A new model, a new standard for accuracy.
Since deploying the new FeedAI model, the change in Fruugo's classification quality has been substantial — and measurable.
"Accuracy has continued to improve, with the average monthly accuracy increasing from around 85% to 95%-plus since deploying the new model," McDonald said.
Recent weekly reports tell an even stronger story. Across a recent five-week stretch, FeedAI processed nearly 7 million product records for Fruugo while maintaining accuracy between 98.44% and 99.74% — consistently north of 99% for most of the window.
Accuracy has continued to improve, with the average monthly accuracy increasing from around 85% to 95%-plus since deploying the new model. ”
Just as important as the headline numbers is how Fruugo and Feedonomics define success together. "We validate through QC, with the methodology agreed with the Feedonomics team, so we know we're aligned on what good looks like," McDonald said. "In terms of success for us — happy merchants. Our merchants used to raise lots of queries relating to inaccurate and misleading classification results, and we have seen a significant reduction in tickets since the new model was deployed."
A partnership that flexes with the business.
The day-to-day relationship has been a meaningful part of the value Fruugo gets from Feedonomics.
"Feedonomics have been great to work with," McDonald said. "The team has scaled up and down according to our needs, and the flexibility has been appreciated. Most notably, the team transitioned to a UK/Europe-based Account Manager to make sure we could maintain a close relationship as we investigated accuracy issues and established a new way of working."
That confidence in classification accuracy has changed how Fruugo's internal teams spend their time. "It allows our teams to move from focusing on getting merchants integrated effectively to growing their business on our platform," McDonald said.
Looking ahead: new languages, agentic commerce.
With English-language classification running smoothly, Fruugo is already looking at how to extend the same approach to the rest of its global marketplace.
"We'd like to explore classification for non-English language products as we grow our non-EN merchant base," McDonald said. "A proof of concept is already being explored."
Further out, Fruugo sees high-quality classification as the foundation for what's next in commerce.
"We’re looking to use accurate classifications to improve the quality of our catalogue data," McDonald said, "to capitalize on agentic commerce initiatives."
We’re looking to use accurate classifications to improve the quality of our catalogue data to capitalize on agentic commerce initiatives. ”
