pm-metrics
Делает ревью продуктовых метрик — тренды, аномалии, root causes и рекомендации к действиям. Включает декомпозицию North Star (L1/L2), диагностику retention-кривых, анализ воронки, разбор A/B-экспериментов, проверку соответствия OKR и фреймворк атрибуции аномалий. User-invoked only — do NOT auto-trigger. Triggers on /pm-metrics, "обзор метрик", "разбор воронки", "анализ удержания", "ретеншн", "A/B результаты", "review metrics", "DAU analysis", "retention analysis", "funnel analysis", "metric anomaly".
serejaris · AI review on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add serejaris/personal-corp-os --skill pm-metricsAbout this skill
name: pm-metrics description: Делает ревью продуктовых метрик — тренды, аномалии, root causes и рекомендации к действиям. Включает декомпозицию North Star (L1/L2), диагностику retention-кривых, анализ воронки, разбор A/B-экспериментов, проверку соответствия OKR и фреймворк атрибуции аномалий. User-invoked only — do NOT auto-trigger. Triggers on /pm-metrics, "обзор метрик", "разбор воронки", "анализ удержания", "ретеншн", "A/B результаты", "review metrics", "DAU analysis", "retention analysis", "funnel analysis", "metric anomaly".
pm-metrics — Product metrics review
Part of the Personal Corp framework — running a one-person business through AI agents. Systematically review product metrics, identify trend changes, locate root causes, output action recommendations. Includes North Star decomposition, retention diagnostics, funnel methodology, and A/B experiment reading.
Inputs
| Field | Required | Notes |
|---|---|---|
| Metric data | yes | Excel / CSV / pasted table / verbal description |
| Cycle | no | Weekly / monthly / quarterly review; default weekly |
| Focus | no | Full review / single-metric anomaly / experiment readout |
| Business context | no | Releases, campaigns, incidents in the period |
Mode: full data → complete review; single-metric change → focused anomaly analysis.
Step 1 — Data integrity check
- Confirm time coverage (current vs comparison period)
- Confirm metric coverage (which North Star / L1 / L2 are present)
- Flag missing critical data
Step 2 — North Star metric system
Decomposition: North Star → L1 → L2.
L1 dimensions:
- User growth: DAU/WAU/MAU, new, returning
- User engagement: core action frequency, session length, feature reach
- User retention: D1 / D7 / D30
- Conversion efficiency: signup → activation → paid step-by-step rates
- Business value: paid rate, ARPU, LTV
- Satisfaction: NPS, complaint rate, ratings
North Star selection guide:
| Product type | Recommended NSM | Typical L1 |
|---|---|---|
| Social / community | Weekly active posters | DAU/MAU ratio, interactions per user, D7 retention |
| Tools / productivity | Weekly users completing core task | Task completion rate, frequency, feature reach |
| E-commerce | Weekly transacting users | GMV, AOV, repeat rate, conversion |
| Content / media | Weekly content-consumption time | Time per user, completion rate, return rate |
| SaaS / B2B | Weekly active teams | Team penetration, feature depth, renewal rate |
Step 3 — Growth metric analysis
Definitions:
- DAU: distinct users with valid action that day
- WAU: distinct users active ≥ 1 day in 7
- MAU: distinct users active ≥ 1 day in 30
- DAU/MAU ratio (stickiness): > 0.5 very high, 0.3-0.5 high, 0.2-0.3 medium, < 0.2 low
User segmentation:
| Type | Definition | Focus |
|---|---|---|
| New | First-time user | Channel quality, activation rate |
| Active retained | Active in both periods | Depth, feature reach |
| Returning | Inactive last period, active this | Return reason, secondary retention |
| Churned | Active last period, inactive this | Churn cause, win-back potential |
| Dormant | Inactive multiple periods | Possibly permanent loss |
Growth identity: This-period MAU = prev-period retained + new + returning − churned
Step 4 — Retention analysis
Definitions:
- D1: % of new users who return on day 2
- D7: % of new users who return on day 8
- D30: % of new users who return on day 31
Retention benchmarks:
| Product type | D1 | D7 | D30 | Note |
|---|---|---|---|---|
| Social / messaging | > 70% | > 50% | > 35% | High-frequency essential |
| Tools | > 40% | > 25% | > 15% | "Use and leave" pattern |
| Content / news | > 35% | > 20% | > 10% | Many alternatives, lower retention |
| E-commerce | > 25% | > 15% | > 8% | Low-frequency, watch repeat rate instead |
| Games | > 40% | > 20% | > 10% | High variance by genre |
| SaaS / B2B | > 60% | > 45% | > 30% | High switching cost, higher baseline |
Retention-curve diagnosis:
- Steep drop (D1 → D7 loses > 60%): activation experience broken — users didn't find value
- Slow decay (D7 → D30 keeps falling, doesn't level): no long-term hook
- L-shape (levels off after D7): healthy, core user base formed
- Bounce-back (sudden uptick on a specific day): cyclical use pattern (e.g. weekday-only)
Retention segmentation:
- By channel: organic vs paid retention gap
- By behavior: completed activation vs not
- By cohort month: compare month-over-month curves to gauge product improvement
Step 5 — Conversion funnel analysis
Funnel construction:
- Define start and end points (e.g. homepage visit → payment success)
- Split into key intermediate steps (each step = a user decision point)
- Per-step rate = arriving at next / arriving at this
Funnel framework:
| Step | Action | Output |
|---|---|---|
| Draw | List steps + rates | Full funnel view |
| Identify bottleneck | Find lowest-rate step | Optimization focus |
| Benchmark | Compare history / industry / competitor | Gap quantification |
| Segment | By channel / device / user type | Locate problem cohort |
| Hypothesize | Why is the bottleneck there? | Optimization direction |
| Experiment | Propose A/B test | Action plan |
Common funnels:
- Acquisition: impression → click → install/signup → activation
- Activation: signup → onboarding done → core action first-trigger
- Payment: browse → cart → order → pay success
- Sharing: trigger → share click → recipient open → recipient conversion
Step 6 — A/B experiment readout
| Dimension | Standard | Note |
|---|---|---|
| Statistical significance | p < 0.05 | p > 0.05 → inconclusive, don't decide |
| Effect size | Lift > MDE | Significant but tiny lift may not be worth it |
| Sample size | Reaches pre-set N | "Significant" without N is unreliable |
| Duration | Covers ≥ 1-2 full we |