Flagship course

Query Signal Lab

A nine-week cohort for people who own in-app search and need a method for diaries, empty pages, and ranking drift. Informational fee: £1,840 excluding VAT per seat. There is no payment on this site.

Team discussion at a bright office table

What you leave able to do

  • Write a sampling rule for query logs that legal and analytics can both live with.
  • Split zero-result traffic into stock, language, and interface causes.
  • Take a ranking snapshot before and after a catalogue feed, without treating NDCG as a personality test.
  • Present five ugly queries in a merchandising meeting without sliding into “sentiment.”
  • Keep a weekly diary that still works when event names change in the next release.
Portrait of instructor Priya Bell

Instructor

Priya Bell spent eleven years on catalogue search at a Manchester marketplace, then two years advising NHS digital catalogues on findability language. She teaches from Alltachonaich and remotely. She will not debug your cluster.

Modules

01

Diary hygiene

Event names, session glue, and why “search_submit” that also fires on filter chips will ruin every chart you later trust.

02

Ugly string hour

You read raw queries aloud. Typos, sizes, gift language, competitor names. The room is not allowed to laugh them off as noise.

03

Zero-result taxonomy

Empty pages get three buckets, not one KPI. You practise arguing the bucket with a sceptical merchandiser.

04

Clustering without over-merge

Stemming that hides “boot” versus “boots” is not clever. We set a merge rule you can explain to a buyer.

05

Ranking snapshots

A before/after of the same 40 queries around a feed refresh. Drift is a date, not a vibe.

06

Facet leakage

Typed colour, size, and price that already exist as filters. You measure how often the box is doing the rail’s job.

07

Synonym theatre

When a synonym file is covering a broken attribute, and when it is actually language people use in Scotland versus the South East.

08

Privacy-shaped sampling

Queries that look like names, postcodes, or order IDs. A redaction habit that does not empty the diary.

09

The merchandising hour

You present five queries and one recommendation. The cohort is the hostile room. Priya scores clarity, not charm.

Fee (informational)

£1,840 excluding VAT per seat for the nine-week lab. Materials and two office hours included. Travel to Alltachonaich for optional studio days is yours. Enrolment is arranged by email — this page does not take payment.

Reviews of this lab

Module 06 on facet leakage paid for the seat. Our mobile rail was collapsed by default; people were typing “size 12” because they could not see the chips. We still argue about ranking, but that argument is now dated.

Sian Adeyemi, product manager, Cardiff

★★★★☆

Rated 4/5 in our learning log. Homework volume is real. If your export is a weekly CSV with no session id, week two hurts. Priya will not invent the id for you.

Platform-style note · internal L&D, insurance comparison

Questions we actually get

Do you teach Elasticsearch, OpenSearch, or vector search?

No. Query Signal Lab is analytics of in-app search behaviour. We do not configure analysers, embeddings, or relevance engineering. If that is what you need, this course will feel incomplete — that is a real limitation, not a teaser.

What export do I need?

A fortnight of query strings with timestamp, result count, and some session glue. Hash user ids. If you only have aggregated “top 50 searches,” you can still sit in, but the diary modules will be thinner.

Is it useful for B2B or only retail?

We have run the lab with parts catalogues, job boards, and a university module finder. The ugly strings change. The sampling argument does not.

What if I cannot attend the optional Scotland days?

Remote is the default. Studio days are extra conversation, not a gated exam.