Site Search & Discovery
On most large catalogues, a third of revenue walks through the search box and a tenth of those searches return nothing. We tune the box until it answers.
The problem
Search is the highest-intent surface on your store and the least maintained
Search returns results ranked by whatever your platform shipped with, not by what converts. Synonyms are incomplete, attribute filters are inconsistent, and nobody owns the zero-result report, so the same failed queries repeat every week.
The users affected are the ones closest to buying. They typed exactly what they wanted, which means every failure is a sale you had already earned and then lost at the last step.
What you actually get
Every engagement ships these. No line items you cannot point at.
Relevance rebuild
Ranking, boosting and tie-breaking rules rewritten against your own conversion data rather than platform defaults.
Zero-result rescue
Every failing query mapped to a synonym, a redirect or a catalogue gap, with a weekly loop so new ones get caught.
Facet and filter design
Filters that match how your buyers actually narrow down, built on attributes that are complete enough to be trusted.
Search analytics layer
A dashboard for search revenue, exit rate and query coverage, so the box has an owner and a number.
Typical outcomes
+0%
Median lift in search-led conversion rate
−0%
Reduction in zero-result queries
0.0×
Revenue per search session after tuning
How we run it, week by week
Query audit
We pull 90 days of query logs, cluster them by intent, and rank failures by lost revenue.
Relevance build
Synonyms, boosts and rules written and staged against a frozen test set of real queries.
Merchandising rules
Category and campaign overrides handed to your team with a simple editing surface.
Weekly tuning loop
New failing queries triaged every week. Search stops being a project and becomes a habit.
Stack we work in
Platform-agnostic. We work in whatever you already own and tell you honestly when it is the constraint.
Questions buyers ask us
Usually not. In roughly four out of five audits the platform is capable and the configuration is the problem. If we do recommend a change, you get the reasoning and the cost of both options before anyone commits.