Agent Flowcore trains product, data, and search teams to treat the in-app search box as a system: queries typed, results shown, empty states endured, and sessions abandoned. We work from logs, not slogans.
Cohorts meet in Bangkok Yai and online across Southeast Asia. The work is slow on purpose — taxonomy before dashboards, instrumentation before opinions.
Eleven weeks on a sandbox catalogue and anonymised query dumps. You leave with a working taxonomy, a zero-result forensic method, and a ranking report a product lead can actually use in a Tuesday review.
The first fortnight is unglamorous: event contracts, session stitching, and the difference between a search impression and a result click. Teams that skip this step later argue about numbers that never meant the same thing.
We do not promise a lift in conversion. We promise that you will know which queries are starving, which ranks are decorative, and which “successful searches” are just people giving up on page one.
Each session opens with a log slice, not a slide deck. Instructors facilitate; they do not perform. You annotate queries in pairs, defend a cluster name, and only then look at a chart.
Homework is short and dated: instrument one event correctly, write a one-page zero-result note, or challenge a ranking feature that looks important and is not. Office hours run on Bangkok time, early enough for Singapore and late enough for Jakarta.
Most in-app search dashboards start with “search CTR.” We start with what people typed. A taxonomy that separates navigational SKU lookups from exploratory browsing stops you averaging two different jobs into one cheerful percentage.
A zero-result page is a conversation. We teach you to read those queries as missing inventory, missing synonyms, missing language coverage — especially Thai script and mixed EN/TH tokens — rather than as a tracking gap to ignore.
Position one always looks gifted. The Lab spends a full week on click models, result card design, and why a “ranking win” can be a layout change in disguise. You will leave slightly harder to impress.
These are not campaign quotes. They refer to modules that actually exist on this site.
“The zero-result forensics week finally gave our grocery app a vocabulary. We still argue about synonyms, but we argue about the same list.”
Instrumentation day assumed more SQL than our PM pair had. We caught up, though the first lab felt steep. After that, session stitching clicked.
★★★★☆ · 8.7/10
Thai query normalisation workshop was the piece our Jakarta playbook never covered. Short, specific, no filler.
We reply within two working days. Enrolment is by conversation, not a cart.
