ESAB Corporation2024 to 20267 min read
ESAB's site search had been running on the same custom crawler for years. It knew one thing, the product catalog. The first time we plugged Bloomreach in, it threw up on the page, 13 seconds to first paint and every content type crammed into one view with no hierarchy. Nobody was going to use that.
01 / Strategy
Three thousand plus products across all the regions. Over a thousand blogs and videos. Every document in about eight languages, plus news.
The product count is the number everyone quotes back to me, and it is the smallest number on that list.
Training an internal model from scratch was a multi-quarter, six-figure bet with high failure risk. Bloomreach gave us a production-ready engine we could shape around our own content without starting from zero. The decision unlocked the timeline; design got to do the actual work.
I segmented the firehose into four buckets that mapped to how users actually think: Products, ESAB University (articles, videos, courses), Documents (catalogs, manuals, spec sheets), and Other for the long tail. Tabs replaced an unstructured infinite list. Cognitive load dropped instantly.
The sidebar shifts completely depending on which tab you're in. Product filters look nothing like document filters, which look nothing like university filters. The UI bends to the content instead of forcing every type through a single generic mold.
North America has 1,232 products. Europe has 779. That is not one catalog translated twice.
The categories are different too. Gas Equipment exists in North America and not in Europe. What North America calls Guns and Torches is just Torches in Europe. PPE and Safety is 16 items in one region and 47 in the other.
So the sidebar could never be a fixed list of filters with the labels swapped out. It had to render whatever taxonomy that region actually has, off one build. That constraint is the reason the whole surface is data-driven instead of designed once and duplicated per market.
The Other tab has no filters at all. Rather than leave a stale sidebar sitting there, it says so: “No Filters Available. Filters are not available for this content type.”
Building the empty state was more work than leaving the filters up. It is also the only honest version of adaptive filtering. If the sidebar only ever rearranges and never leaves, it is not adapting to the content, it is just moving.
Documents carry a Language filter. It is there because some regions have more than one native language, Canada and parts of Europe being the reason it exists at all. Region is set by geolocation with a manual override, and the region decides which languages are even on the list.
On a technical catalog the spelling variant is a correctness problem rather than a preference. Search “aluminum wire” against a UK-spelled corpus and you get nothing.
The education content filters on Experience Level: Beginner, Intermediate, Advanced, Expert. The library is weighted hard to the front of that, 137 pieces at Beginner against 2 at Expert, and that is on purpose.
Somebody who just bought their first machine and somebody certifying a procedure will type the same words into the same box and want completely different things back. The facet is the interface admitting that out loud.
If you have been on the site before, the ordering is built against what you have actually been looking at. If you have not, you get Popular on a product search, Featured on an unfiltered browse, and Relevance on everything else.
Behavioral merchandising on a welding catalog, which is not where most people expect to find it.
Sitting on top of search is E.V.A., ESAB's LLM-powered AI chat assistant, available site-wide for users who want to skip search entirely and just ask a question. Off-topic guardrails, prompt injection protection, and a content scope tightly bounded to what the knowledge base can answer. EVA flags missing topics; humans decide what to build next.
I worked side by side with the senior front-end dev through the entire process. Not handing off comps and hoping for the best, collaborating on what was buildable, what would perform, and where to cut. The 13-to-1-second improvement came out of those conversations as much as it did the design.
02 / Results
Thousands of conversations per quarter. Well above three quarters of sessions resolved without escalation. Search is live in 16 regions, translated to whatever language geolocation sets.
03 / My role
Owned the UX and product design for both surfaces, from the IA decision on the four content buckets through the dynamic filter logic, the EVA conversation patterns, the off-topic and prompt injection guardrails, and the analytics taxonomy that classifies unresolved sessions into actionable signal. Partnered closely with the senior front-end dev on what was buildable and where to cut, and with the engineering team on the Bloomreach integration, the multilingual rollout, and the model evaluation loop.
The largest unresolved bucket is missing content, things that simply aren't in the knowledge base yet, like spare-parts catalogs and consumable selection guides. The next is out-of-scope by design: pricing, inventory, distributor contacts. Things EVA was never built to handle. Those aren't failures, they're the next sprint's roadmap.
04 / Gallery