Product Analyst

Responsibilities: 

1. Ownership

   Be the analytics counterpart for a product domain 1:1 chat, content packs, negotiations, live streaming — you become the person PMs and engineers come to for any number in that area, and the person accountable when it's wrong.

2. Experiments

   Run A/B tests on product features, end to end Turn a product hypothesis into a test: primary and secondary metrics, the minimum effect worth detecting, and the audience. Confirm the population split and assignment events are genuinely correct before trusting the readout. Monitor it while it runs, then deliver a clear ship / don't-ship call and the reasoning behind it. This is a large part of the job, not an occasional project.

3. Pipelines

   Build and maintain the datasets behind your domain Write production transformations as SQL and YAML pipeline config, with freshness, uniqueness and validity checks. Your work goes through code review like any engineer's.

4. Events

   Specify instrumentation before it's built Write the BI event spec for new features across Android, iOS and Web; then verify after release that events actually fire the way the spec said, and chase the gaps when they don't.

5. Dashboards

   Own dashboards and the models under them

   Including retiring the ones nobody opens. A dashboard with no reader is a cost, not an asset.

6. Analysis

   Investigate what moved and why Revenue and conversion anomalies, feature performance, behavioural questions from PMs. Present findings to people who will act on them.

7. Incidents Size production incidents under time pressure When something breaks, you're asked how many users were affected and for how long — quickly, and precisely enough to defend the number when engineering disagrees with it.
8. Cost Keep query and pipeline cost honest Partitioning, clustering, and knowing why a query scans what it scans. Warehouse capacity is shared and finite; an expensive pipeline is everyone's problem.

Requirements:

• Advanced SQL, including window functions and multi-source joins — and the ability to reason about what a query costs, not only whether it returns.
• A strong experimentation background. Hypothesis testing, confidence intervals, power and minimum detectable effect — and the failure modes: peeking at a running test, multiple comparisons, hunting for a segment where the result works. You should also be the kind of person who checks that the experiment itself is sound before trusting its numbers.
• 3+ years as a product or data analyst on consumer products, with real exposure to both mobile and web.
• Critical use of AI analytics tooling. We use an internal Claude-based layer over BigQuery heavily. We want someone who checks what it produces before passing it on.
• Comfort shipping code: git, merge requests, reading and reviewing someone else's SQL. You'll write production pipeline code, not only notebooks and queries.
• Working knowledge of product event tracking: what a well-specified event looks like, and why the same event behaves differently across platforms.
• English at B1 or above. You'll write specifications that engineers build from and present findings to product stakeholders, so clear written English matters more than fluent speech.

Strong advantage

• BigQuery specifically — partitioned tables, clustering, slot-based pricing.
• Airflow, dbt, or another orchestrated transformation framework.
• Having been the analyst on a production incident, not only in the post-mortem.

Nice to have

• Python with the usual statistical libraries — pandas, scipy, stats models.
• A BI tool such as Looker, Tableau, or Power BI — having built the models and dashboards, not only read them. We use Looker today but are moving off it, so the particular tool matters much less than having done the work in one.
• Data quality tooling such as Soda or Great Expectations.
• Experience in social, live-streaming, gaming or another engagement-driven consumer product.