Searching Turns into Finding
OBI is one of the best-known home improvement stores in Germany. In addition to a wide range of products, its business model also includes services and consulting. The product search plays a key role in this: Customers shop online or research products digitally before visiting a brick-and-mortar store.
As an omnichannel retailer, OBI has a strong presence in e-commerce. The store search function, which has been iteratively developed over the past few years, was already very powerful and ran smoothly. However, precise results are no longer enough today. OBI has reimagined its search to accommodate changes in customer behavior: Customers now search in a more descriptive, problem-oriented manner or use different terms for the same product.
The starting point, therefore, was not a “poor search” but an already very good search facing new challenges. A standard solution was insufficient and could not address these challenges.
We worked together to optimize the system: With an AI-powered, hybrid search, we’ve turned “searching” into “finding better.”
About OBI:
0+
Stores in Europe
More than0million
Annual store visitors
More than0million
Hey, OBI customers
Since0
Collaboration Between OBI and neuland
Use AI where it really helps
We didn't replace the existing search with an AI-powered search; instead, we expanded it in areas where traditional search reaches its limits.
The key focus was on long-tail search—that is, search queries that occur relatively rarely and cannot be optimized individually when considered as a whole.
The goal, therefore, was not to build a new search engine, but to optimize the existing one in a targeted manner: maintaining precision where it already works well and adding understanding where limitations previously existed. Instead of rolling out a one-size-fits-all solution, we collaboratively determined where AI would actually add value.
The result is a hybrid search architecture combining classic full-text search for precision and AI-powered vector search for understanding.
What AI Really Improves
The AI component specifically addresses those areas of traditional search where standard methods fall short.
A Better Understanding Instead of Keyword Matches
The search recognizes the intent behind a query and delivers more relevant results—even without an exact keyword match (e.g., "I have mold in my bathroom").
Handling Synonyms
Even so, different terms, phrases, or descriptions will still lead you to the right product. For the query “What can I do to protect myself from burglars?”, the search results show window handle locks, safes, and alarm systems.
Troubleshooting becomes possible
The search works reliably even with vague queries or natural language (e.g., “thing for drilling a hole in the ground” instead of “hand-held drill”).
Multilingual Support
The solution can process multilingual search queries. However, the key to this is selecting appropriate models and training them for the specific linguistic context.
Implementation of the Hybrid Search Architecture
The implementation was specifically tailored to OBI:
neuland’s in-house development instead of a standard solution—full control and flexibility over logic and integration
Various LLM models were tested, compared, and evaluated in the OBI context—rather than relying on a standard solution, a targeted analysis was conducted to determine which model delivers the best results in conjunction with search behavior, product data, and long-tail queries
Training was conducted using OBI’s own data rather than generic datasets
An iterative approach with A/B tests to validate effectiveness and use cases
Products and search queries are now mapped in a multidimensional space (vector). The search finds not only exact matches but also products that are relevant in terms of content—based on their semantic proximity.
What Specifically Has Improved
Fewer "No Results"
The "Product Not Found" rate was significantly reduced while maintaining high precision and relevance.
Improved Visibility in the Long Tail
Vague or descriptive search queries lead to relevant products.
Business Impact
Customers find what they're looking for faster and are more likely to convert.
More results without compromising quality
AI search displays more relevant results without compromising relevance: The overall product click-through rate remains stable.
neuland has a particularly high level of technical search expertise. When it came to AI search, the team brought deep, specialized knowledge to the table, which gave us a huge boost. The neuland team knows what they’re doing and always works with us as equals. That helps us get things done successfully.
Senior Product Owner, OBI Group Holding
Outlook
With AI-powered search, we’ve been able to further optimize a search function that was already very good. However, AI-powered search is not a panacea. It does not solve every problem, such as missing product data. At OBI, however, it has addressed precisely what was previously not scalable: language understanding in the long tail. This opens up future possibilities, such as the expansion of features and further internationalization.
We’ve been working with OBI since 2010. We’re pleased to see how quickly the AI search has become a success, and especially that we’re creating real added value for and with OBI. The use of AI in search will also become increasingly important for Agentic Commerce.
Project Manager, neuland
