Yujun Zhou

Work

Decisions. Evidence. Working systems.

Each item opens with the problem, my contribution, and what changed. The label on the card says what you are looking at: a case study written from industry work, a working prototype, or a study design with nothing built yet.

Industry cases

Measurement work inside a team

Written from work I did inside a company. Employers and internal products are described by category rather than by name, and confidential data and results are omitted — what is here is my contribution and the method.

Independent builds

Products I built and then studied

Built end to end by me. What varies is how much of each one has been measured rather than shipped.

Case studyIndependent build

ChurchMap

Built a location-first directory, then reframed the problem from crawl volume to which fields a first visit cannot do without.

Independent builderRead the case →
Case studyIndependent build

Debrief

Built a voice-to-review product, and defined the first honest metric for it rather than adding features.

Independent builderRead the case →

In the lab

Scoped, not built

Study designs with stated stop conditions. Nothing here has results, and the pages say so at the top rather than in a footnote.

Study designIndependent build

Creative Evidence Lab

A study design that keeps label quality, predictive value, and incremental impact as three separate claims.

Independent · Study designRead the case →
Study designIndependent build

Analytics Skeptic

A prompt that reviews analyses the way a skeptical data scientist would, plus the design for testing whether it works.

Independent · Study designRead the case →

Earlier research

Peer-reviewed work

Forecasting and policy evaluation from my PhD. The recurring question — whether a model that fits is a model someone can act on — is the same one I work on now.

Full list →