Product analytics investigator
Skill sisodiabhumca/agent-skills/skills/product-analytics-investigator
Production-Ready Agent Skills : product analytics, growth experiments, CRM, research synthesis, postmortems, data contracts, SaaS spend, compliance, architecture maps, and LLM eval and many more.
npx -y skills add sisodiabhumca/agent-skills --skill product-analytics-investigatorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when a PM, data PM, or analyst needs to investigate product metrics in Amplitude or Mixpanel — diagnosing drops in activation, retention, or funnel conversion, or attributing changes to releases, segments, or experiments. Pulls events, runs cohort/funnel analysis, and writes a root-cause memo.
SKILL.md
1.9 KB, 397 tokens by cl100k_base, as published. Nobody here has run it
Product Analytics Investigator
When to invoke
- "Why did activation drop last week?"
- "Investigate the checkout funnel for mobile users."
- "Compare retention before/after the v4.2 release."
Inputs needed
- Metric or funnel under investigation (event names + ordering).
- Time window (default: last 14 days vs prior 14 days).
- Segments to slice by (platform, plan, country, cohort).
- Hypotheses (optional) — release dates, experiments, marketing pushes.
If any are missing, ask the user once before running.
Workflow
- Frame — restate the metric, window, baseline, and segments.
- Pull data — call
investigate.py(Amplitude/Mixpanel API or warehouse). - Decompose — break the change by segment, surface, and step.
- Correlate — overlay releases, experiments, incidents.
- Report — produce a memo: TL;DR, evidence, top 3 hypotheses ranked by likelihood, recommended next steps.
Output format
## TL;DR
<one sentence: what changed, by how much, who is affected>
## Evidence
- Funnel step deltas (table)
- Top contributing segments
- Timeline overlay with releases/experiments
## Hypotheses (ranked)
1. <hypothesis> — supporting evidence — confidence
2. ...
## Recommended next steps
- [ ] Owner — Action — ETA
Guardrails
- Never claim causation from correlation alone. Surface confounders.
- Always show absolute counts alongside percentages.
- Flag low-sample segments (n < 100) explicitly.
Reference code
See investigate.py for a runnable funnel + cohort comparison against Amplitude or Mixpanel APIs (or a CSV export).