01Business problem
Type “dishwashing liquid” into the webshop and the top results came back as chicken. The search engine was not broken. The product data behind it was incomplete and getting worse, because a large part of the catalogue changes regularly and enriching it by hand could never keep up.
02Intervention
Together with another consultancy, we did not recommend a smarter search engine. We went to the data itself and set up a live test in which part of the visitors got the improved product data, so the business could see the difference before deciding anything.
03Role of data & AI
Image models read the product photos for labels, packaging text and brand. A language model wrote the descriptions, keywords and categories that search needs. The value was already there, in the data.
04Adoption
The improved data went live for a share of visitors first, measured against the rest. E-commerce and category teams saw the effect on their own KPIs rather than on a slide.
05Measurable impact
In a 5-week live test in one market, the visitors who got the enriched product data placed 6% more orders and generated 5% more revenue per visitor than the control group on the old data, measured over the same period.