How HSE Uses Elasticsearch to Boost CTR 4% and Lift Satisfaction 8%

HSE operates at a scale few retailers match: 43,000 packages shipped daily from four logistics centers, a reach of 46 million households across three countries, and a customer base that arrives through fundamentally different channels. TV viewers—who often skew older and navigate the internet differently than casual online shoppers—search with product numbers and creator names. Visitors arriving from the app or social commerce channel use general language and browsing patterns. Serving all of them well from a single search platform is a genuine engineering challenge.

The previous solution, which Peter Strasser, Software Developer at HSE, described as a “black box,” made that challenge worse. Its internals were opaque, making it difficult to understand why a given search term returned a particular set of products. Fine-tuning the experience for different customer entry points was limited. When a TV viewer searched by product number and got irrelevant results, or when a casual browser’s general query surfaced the wrong category, the team had few levers to pull.

HSE selected Elasticsearch and deployed it on Elastic Cloud in AWS, choosing the platform for its flexibility and the control its APIs provided. Elasticsearch exposed filters, aggregations, boosts, and sorting as composable building blocks, letting the team craft fine-grained queries tailored to specific customer contexts—whether a search originated from a creator lookup, a brand browse, a product description, or a direct product number. The team also integrated Elastic’s AI and machine learning features, including generative AI and large language models to extract semantic meaning from customer queries, complementing traditional keyword matching and returning results that align with intent rather than just literal terms.

HSE piloted the deployment first on its social commerce app, HSE Hello, where the team built deep familiarity with the Elastic APIs before tackling the main website. That sequence paid off: when the team rebuilt the website search platform, they had a working mental model that accelerated development. Elastic Cloud’s SaaS architecture also eliminated the burden of managing the virtual machines their previous vendor required, freeing engineers to focus on search quality rather than infrastructure.

An A/B test comparing the old search platform with Elasticsearch on the main website showed a measurable shift: the share of customers dissatisfied with search results fell by 8%, while click-through rates on search result pages rose by 4%. For a site at HSE’s scale, a 4% CTR gain on search represents a significant revenue impact. Looking ahead, HSE plans to extend AI-powered search into user-personalized results and image—and potentially video—search, using Elastic’s open architecture to integrate best-in-class models as the technology evolves.