What Metrics Should Product Managers Actually Track?
Track three tiers of metrics—unit economics, engagement velocity, and health signals—to avoid metric theater and catch problems early.
The Core Answer
Most product teams track 15+ metrics and understand none of them. The mechanism that works is a three-tier system: unit economics (revenue per customer, cost to acquire, lifetime value), engagement velocity (frequency, depth, growth rate), and health signals (churn drivers, feature adoption, sentiment). Each tier answers a different question about product-market fit. Unit economics tell you if the business model works. Engagement velocity tells you if the product is sticky. Health signals tell you which parts are breaking. If you’re tracking churn but not the feature adoption that predicts churn, you’re managing by coincidence, not causation.
Tier 1: Unit Economics—Does the Business Model Work?
These are the metrics that determine if your company survives. Every product manager must know:
Revenue per customer (MRR/ARR per user or account). This is your northstar financial signal. If it’s declining, your product is being commoditized or customers are finding workarounds. If it’s flat while usage grows, you’re leaving money on the table. Track by cohort (customers acquired in Q1 2024 vs. Q2 2024) to spot unit economics degradation early.
Customer acquisition cost (CAC) and payback period. How much do you spend to acquire a customer? How long until they generate enough revenue to cover that cost? If payback is 18+ months and your product has 6-month average lifetime, you’re in a death spiral. CAC varies by channel; track separately for sales-driven vs. self-serve cohorts.
Churn rate and churn drivers. Churn is lagging (you already lost them), so it’s useless for weekly decisioning. What predicts churn? Low usage? Specific feature gaps? Failed implementation? Identify the leading indicators. “5 customers churned” is noise. “Customers with zero logins in 30 days have 60% 90-day churn” is actionable.
Gross margin by segment. If you sell to enterprises and SMBs at the same price, you’re probably losing money on SMBs. Calculate gross margin by customer size, industry, use case. This drives prioritization: if SMBs are 40% revenue but 10% margin, you either raise prices or deprioritize their feature requests.
Tier 2: Engagement Velocity—Is the Product Sticky?
Unit economics answer “Does the math work?” Engagement answers “Do customers actually want this?”
Weekly active users (WAU) and monthly active users (MAU). Raw counts are noise; retention is signal. Track the cohort retention curve: of users acquired in week 1, how many come back in week 2, 4, 12? If 50% don’t return after the first week, your onboarding is broken. If retention drops 30% in month 3, you hit a paywall problem or the product lost novelty.
Feature adoption and depth. When you ship a new feature, what % of users touch it? Of those, how many use it more than once? If 80% of users never touch your new intelligent recommendations feature, it’s not a priority anymore—backlog it and invest elsewhere. Adoption is the fastest leading indicator of product-market fit for specific features.
Time in app and interaction depth. “Time spent” is only meaningful if paired with action type. Spending 60 minutes reading a report is different from spending 60 minutes creating a new object. Track actions per session and session frequency. If actions are declining while time is stable, users are clicking around; if both decline together, engagement is really dying.
Activation rate (users who complete key action in first session). If you get 1000 signups and 40 complete setup, your activation is 4%. Unless you’re a competitor replacement, sub-5% activation means the product doesn’t communicate value. This is your early signal that positioning or onboarding is broken—before you get churn data.
Tier 3: Health Signals—What’s Breaking?
These are the canaries that tell you problems before they become disasters.
Feature-specific churn correlation. If users who use Feature X have 20% 90-day churn and Feature X non-users have 40% churn, Feature X is core to retention. If the inverse is true, Feature X is distracting. Segment churn by cohort behavior: what do churned users have in common?
Sentiment and NPS. NPS is lagging and noisy (many factors influence it), but NPS comments are gold. Are users saying “I don’t have time for this” (product-market fit problem) or “This doesn’t do X” (feature problem)? One requires a pivot; the other requires execution. Use CSAT on specific flows (post-checkout, post-setup, post-feature-launch) to catch problems early.
Support volume by category. If support tickets for Feature X doubled, something broke. If ticket complexity is rising (each resolution takes longer), the product is getting harder to use. Track tickets as a leading indicator of friction, not just a cost center.
Infrastructure and platform health. Uptime, latency, error rates, database size growth. If p99 latency is creeping toward your SLA, scale before it breaks. If database is 80% of capacity, plan migration now. These aren’t “nice to track”—they’re non-negotiable.
The Measurement Model: What to Watch Weekly
Create a dashboard with three sections:
| Unit Economics | Engagement | Health |
|---|---|---|
| MRR, churn rate, CAC payback | MAU, cohort retention, feature adoption | NPS comments, support backlog, latency p99 |
| Updated: Monthly | Updated: Weekly | Updated: Daily |
| Driver: CFO + Product | Driver: Product + Data | Driver: DevOps + Data |
Weekly rhythm: Review engagement and health daily (10 min). Review unit economics monthly (30 min). This prevents metric overload while keeping you connected to what matters.
Common Mistakes: Vanity Metrics and Misalignment
Mistake 1: Tracking outputs instead of outcomes. “We shipped 12 features” (output). “Three features had adoption above 30%, two caused churn to rise” (outcome). Track what changed in customer behavior, not what you shipped.
Mistake 2: Metrics without owners. If no one is responsible for churn, churn creeps up. If retention is “the team’s metric,” it’s no one’s. Assign each metric an owner (e.g., onboarding designer owns activation rate, platform engineer owns latency).
Mistake 3: Optimizing a metric that doesn’t predict business outcome. Feature requests per user is not a health signal. Revenue impact per feature request is. DAU growth looks good until churn spikes because engagement is hollow.
Mistake 4: Ignoring cohort composition changes. If your average deal size fell 40%, unit economics look terrible—but it’s because you shifted to self-serve. Track the same customer cohort over time or segment explicitly.
How to Apply This
Week 1: Audit your dashboard. List every metric you currently track. For each, write: (1) Why do we track this? (2) What decision does it drive? (3) Is it a leading or lagging indicator? Anything without a clear “driver” gets deleted.
Week 2: Build the three-tier dashboard. Start with the 10-12 metrics above. Implement or expose them in your BI tool. Focus on cohort-level metrics first (retention curve, churn driver analysis); daily aggregates are less useful.
Week 3: Run weekly reviews. 15-minute standup where a data analyst walks the team through engagement and health. What’s moving? What’s not? Are there surprises? This rhythm creates muscle memory.
Week 4: Connect metrics to decisions. When churn ticks up, ask “Which cohort? Which feature set? Which segment?” Don’t react to top-line metrics. Dig into the story. This is how metrics become useful instead of decorative.
The Bottom Line
Metrics are a navigation system, not a report card. The three-tier framework keeps you focused on what matters: unit economics prove the model works, engagement proves customers stick around, health signals warn you before things break. Most teams drown in metrics because they track everything and understand nothing. Start with 10-12. Use them for a month. Learn what they mean. Only then add more. This discipline prevents metric theater and ensures every number on your dashboard drives a real business decision.