Zero results are not a KPI
Collapsing every empty page into “search failure” hides the only split that changes the catalogue.
Alltachonaich teaching studio · GB
Typed queries are the last honest log in the product.
Data Flowbase is a small studio for catalogue, product, and analytics people who need in-app search analytics that survive a sceptical merchandising meeting. We train you to read what people actually type — including the ugly strings — rather than decorating a “search success” chart.
Open the flagship lab outline| Query | Hits | Note |
|---|---|---|
| navy parka mens | 41 | Facet already selected “coats” |
| waterproof | 0 | Synonym missing; rain jacket exists |
| size 14 | 18 | Typed into search, not filter |
| gift for mum | 0 | Intent, not SKU language |
| blazer | 27 | Ranked behind “blazer dress” |
Flagship
Query Signal Lab
Nine weeks on diaries, zero-result taxonomies, ranking snapshots, and how to present search motion without inventing a conversion religion. You bring an export. We refuse to average it into a single “findability” score.
Cohorts meet remotely with two optional days in the west of Scotland. Homework is unforgiving if your event names are a mess. That is the point.
Modules, fees, and a real limitation →
From people who sat in the query log
Notes, not slogans
The zero-result clinic still sits on my Friday calendar. I would not roll their ranking snapshot onto a marketplace with 2 million SKUs yet — we are smaller — but the naming work stuck.
I arrived sure “no results” meant the catalogue was thin. After Query Signal Lab I could show merchandising where people were typing sizes into the box because the filter rail was hidden on mobile.
8/10 on our internal wiki. The synonym homework exposed how much of our “search quality” was actually a CMS spelling habit. Slightly humiliating. Useful.
What we actually teach
Search analytics as a reading skill, not a dashboard SKU
Most product teams already drown in funnel tiles. The missing piece is a shared vocabulary for what a typed string is doing: retry loops, facet leakage, synonym theatre, ranking drift after a feed refresh. Those are signals. They are not “engagement.”
You leave with a method you can defend in front of a sceptical ops director in Leeds or a buyer in Edinburgh — not a template that only works in a demo tenant.
How we define in-app search analytics →We will not
Tune Elasticsearch for you, sell a ranking model, or promise a lift in add-to-basket. If your problem is index mapping, hire an engineer. If your problem is that nobody can agree what a “failed search” is, you are in the right studio.
We will
Sit with a fortnight of queries, force the ugly strings onto a whiteboard, and make you pick a sampling rule you can live with when legal asks about identifiers.
47product teams coached since 2018
9 wkslength of Query Signal Lab
312query diaries reviewed last year
£1,840cohort seat, excluding VAT
Field notes
Recent writing
Ranking drift after a catalogue refresh
Feeds move on Tuesday. Nobody timestamps the queries. Here is a snapshot habit that fits a UK merchandising week.