Hobbycraft Adapts to Shifting Search Behavior With Bloomreach

Through Bloomreach’s Limitless Lift Program, Hobbycraft ran a structured POC to measure the impact of Bloomreach’s Loomi Search+ feature. The team tested configurations like LLM precision mode to improve relevance for tail and torso queries and category learn-to-rank to dynamically adjust rankings based on merchandising goals. While revenue per visitor (RPV) was already strong, they wanted to continue experimenting to uncover additional opportunities for optimization.

Here’s how Loomi AI supported Hobbycraft on a granular level:

• Search that understands intent, not just keywords. When “pipe cleaner flowers” surged following a TikTok trend and generated over 1,300 searches on Hobbycraft’s site, Loomi Search+ surfaced both exact and contextual product matches. Previously, the team would have had to spot the trend themselves, pull the right products, and push them into results before the moment passed; however, Loomi AI removed that complicated workflow entirely.
• Serving every type of shopper — from “seekers” to browsers. Hobbycraft’s customers range from people who know exactly what they want (aka “seekers”) to those browsing for creative inspiration without a specific product in mind. Loomi AI is built to handle both, returning precise, relevant results for specific queries while casting a wider net for broader, more exploratory ones. Testing different precision mode configurations helped the team find a setup that served the full spectrum of their shopper base without alienating any one group. 
• Smarter category rankings. The team stopped manually configuring rules and used category learn-to-rank, which bases product rankings on real shopper behavior and conversion signals. This made category pages more responsive to what shoppers were actually engaging with and buying.

How It Worked

1. Structured A/B testing from the start. The POC was designed to produce clear, comparable results and give the team a reliable baseline to measure impact across the full funnel — RPV, conversion rate, ATC, and average order value (AOV). 
2. Starting with the queries that mattered most. Testing focused first on long-tail and specific category queries — the areas where a table stakes search approach was the most exposed. More specific queries delivered some of the biggest uplifts, demonstrating how Loomi AI translates natural-language intent into more relevant results.
3. Finding the right balance across query types. Rather than settling on a single setup, the team tested multiple precision mode configurations to understand how each performed across different query types and work toward a global setup. 
4. Turning search insights into action. Hobbycraft worked closely with Bloomreach’s CSM and professional services teams to interpret results and refine its approach. 
5. Building a foundation for the future. With the POC complete, Hobbycraft’s ambition goes beyond locking in these initial gains. The team plans to keep testing on a consistent basis, explore features they haven’t yet leveraged (e.g., vector temperature), and continue expanding what the Loomi AI platform can do for their business.