Search like an experiment, not a shop
Why scientists could not find the right reagent in a fast-growing catalogue, and the research that turned search into a three-step guided funnel.
The one-minute version
- Scientists search by experiment, not by product. Everyone in the sample narrowed the same way: target, then application, then species. The site asked them to do it the other way round.
- The job has two speeds. Months of planning, then urgent reordering when an experiment fails.
- Trust depends on seniority. Senior researchers look for peer citations, junior researchers for visual confirmation. Neither trusts marketing language.
A catalogue growing faster than its search
Abcam sells reagents to scientists whose experiments can take months and cost five figures. In earlier research I had noticed that, as the catalogue expanded, people seemed less and less likely to find the right product for their experiment.
What the data already showed
- Google Analytics: sessions were dropping off at the search results page.
- Google Analytics and Hotjar heatmaps: fewer than 4% of visitors interacted with filters.
- A UX survey in early 2020: search was the area users were most concerned about.
Why it mattered to the business
Harder discovery meant inefficient searching and lost conversion, with satisfaction and customer lifetime value likely to follow.
Area of the site users were most concerned about
UX survey, early 2020. Number of respondents naming each area.
Deciding what to learn before how to learn it
The analytics told us where people left. They could not tell us why. Before choosing methods I ran hypothesis workshops to agree on what kind of evidence would answer the design problem.
Two hypotheses came out of the workshops
- Users feel overwhelmed by the number of search results.
- Users follow a search pattern, because their first search term is almost always what scientists call a target.
Three kinds of evidence, and the method chosen for each
Method: 40 remote ethnographic interviews.
Method: observation inside the same 40 sessions.
Method: diary study, 12 participants.
Who took part
The 40 interviews were split across all of Abcam’s digital personas and included both users and non-users. The diary study had 12 participants worldwide, two for each main persona.
My role
I led the research: I framed the problem from the existing data, ran the hypothesis workshops, designed and conducted the studies, and synthesised the findings. The search redesign and the taxonomy work that followed were delivered by the wider product team.
What I found
The research confirmed the pattern hypothesis and changed the problem. This was not a search interface issue. It was a mismatch between the site and the way scientists work.
1. Scientists search by experiment, not by product
Every researcher in the sample followed the same three-step mental model. The platform asked them to search like shoppers: type a term, receive a long list, then filter. Almost nobody filtered.
↑ The order scientists design an experiment in, and the order they wanted to search in
2. Months of planning, then hours to reorder
The job has two speeds. Scientists plan experiments months ahead, then need products urgently when an experiment fails and they have to reorder. A search built for browsing fails the urgent case.
3. Trust signals differ by seniority
Senior researchers trust peer citations. Junior researchers trust visual confirmation. Both distrust marketing language, and both respond to structured peer evidence: citations, validation data and lab of origin.
“Finding the right reagent wasn’t just frustrating. It could cost researchers time, grant money, and experiment success.”
UX research report, 2020From insight to decision
One principle came out of the synthesis: make the platform mirror the experiment, not the catalogue. Three decisions followed from it.
| Evidence | What it meant | What the team decided |
|---|---|---|
| Everyone narrowed by target, application, species | The open text box put the hardest work after the results, where fewer than 4% did it | A three-step guided funnel replaced the open-text default, so people narrow by intent before they see results |
| Filters did not match how scientists think | Filter logic followed the catalogue, not the experiment | Filter logic was rebuilt around target, application and species, which required taxonomy work across the catalogue |
| The job has two speeds, and marketing language is distrusted | Features that add browsing or promotion add friction in the urgent case | Two planned features were dropped: an AI-style “smart search” experiment and a marketing-led recommendation widget |
The new search, step by step
The design adds a step on purpose. It is useful friction: one extra choice at the start, in exchange for a short, relevant list at the end. The filters offered change with the type of product being searched for.
Impact
Measured 90 days after the redesigned search launched.
How to read these numbers
They belong to the redesign the whole team shipped. The research set the direction, the three-step model and the roadmap choices. Design, engineering and taxonomy work delivered it.
Beyond the launch
The study design became the template for three later foundational studies at Abcam.
“Nass built clarity where there was ambiguity. His research roadmap at Abcam gave teams a shared direction and confidence in their decisions.”
Dan Edmeades, Head of UX at AbcamLimits, and what I took from it
Limits I kept in view
- The findings are qualitative. Forty interviews show how scientists think; they do not size how many behave each way.
- The diary study had two participants per persona, enough to see context, not to compare personas.
- The guided funnel adds a step. That trade-off needed watching in real lab contexts, not only in sessions.
What I took from it
- Agree on the type of evidence first. The hypothesis workshops made the choice of methods easy to defend.
- Reframing the problem did more than any single finding. Once the team saw a workflow mismatch, the design direction followed.
- Research earns its place when it removes work from a roadmap as well as adding to it.
Want the full story?
I am happy to walk through the method, the evidence behind each finding, and what I would do differently. nassim.seridji@gmail.com