Every product team has a dashboard. Very few teams have metrics.
The difference: a dashboard is a collection of numbers you can look at. A metric is a number that changes what you do. Most teams are drowning in the first and starving for the second.
This post is about picking the small set of numbers worth watching, understanding what they actually tell you, and avoiding the traps that make smart people draw wrong conclusions.
Start With the Value Question
Before choosing metrics, answer one question: when a user gets value from your product, what does that look like in the data?
For Slack, it's messages sent within a team. For a project management tool, maybe it's tasks created and completed by multiple collaborators. For an e-commerce product, it's purchases. Not logins. Not page views. The moment of actual value.
This is your north star metric, or at least the raw material for one. Everything else in your metric stack should connect to it. If you can't articulate the value moment, no framework will save you, because you'll be optimizing proxies for something you haven't defined.
A good north star has three properties:
- •It reflects value delivered to users, not value extracted from them
- •It's a leading indicator of revenue, not revenue itself
- •Teams can actually influence it through product work
The Core Metric Stack
Beyond the north star, most products need a small set covering the user lifecycle. The AARRR framing (acquisition, activation, retention, referral, revenue) is old but still useful. Here's the practical version:
Acquisition
How many new users show up, and from where. PMs often over-index here because the numbers are big and go up. Careful: acquisition is mostly marketing's lever. Your product lever starts at the next step.
Activation
The percentage of new users who reach the value moment. This is the highest-leverage metric for most product teams, and usually the most neglected. A 10% improvement in activation compounds through everything downstream.
Define it precisely: which actions, within what time window. "Created a project and invited a teammate within 7 days" is an activation definition. "Engaged with the product" is not.
Retention
Do people come back? Retention is the closest thing product management has to a truth serum. You can buy acquisition and fake engagement, but you cannot fake people repeatedly choosing to return.
Look at retention as cohort curves, not a single number. A curve that flattens means you have a core of retained users and a real product. A curve that decays to zero means you have a leaky bucket, and pouring more acquisition into it is setting money on fire.
Revenue
Conversion rate, ARPU, expansion, churn. How much depends on your role; a growth PM lives here, a platform PM barely touches it. But every PM should know how their surface connects to money, because that's the language the business speaks.
Vanity Metrics and Other Lies
Some numbers exist mainly to feel good. Learn to spot them:
- •Cumulative anything. "10 million total signups" can only go up. It tells you nothing about now.
- •Registered users. Counts everyone who ever showed up, including the 90% who left.
- •Page views and time on site, unless attention is literally your business model. For most products, more time in-app can mean confusion, not engagement.
- •Averages hiding distributions. "Average session: 8 minutes" might be most users at 30 seconds and a few at hours. Look at medians and percentiles.
The test for any metric: if this number moved, would we do something differently? If the answer is no, it's decoration.
Ratios Beat Raw Counts
Raw counts grow whenever your user base grows, which makes everything look like progress. Ratios tell the truth:
- •Not "5,000 users activated" but "38% of signups activated"
- •Not "20,000 weekly actives" but "DAU/MAU of 0.25"
- •Not "$2M ARR" but "net revenue retention of 104%"
When a stakeholder shows you an up-and-to-the-right raw count, your first question should be "what's the denominator?" You'll be surprised how often the ratio underneath is flat or declining.
Leading vs. Lagging
Revenue and churn are lagging indicators. By the time they move, the cause is months old. Good metric stacks pair every lagging metric with a leading one you can act on:
- •Churn is lagging. Declining usage frequency in the prior 60 days is leading.
- •Revenue is lagging. Activation rate and trial conversion are leading.
- •NPS is lagging (and noisy). Support ticket themes and feature adoption are leading.
If your quarterly goals are all lagging metrics, you're steering by looking out the rear window. This matters directly for OKRs: key results should mostly be leading indicators you can move within the quarter.
The Traps That Get Smart PMs
Correlation dressed up as causation
Users who use Feature X retain better! So drive everyone to Feature X! Except power users both use more features and retain better, and Feature X caused nothing. Before acting on a correlation, run the experiment. That's what A/B testing is for.
Goodhart's law
When a measure becomes a target, it stops being a good measure. Target "tickets closed" and tickets get closed prematurely. Always pair a target metric with a guardrail metric that catches the gaming: tickets closed and reopen rate.
Segmentation blindness
Your topline can be flat while enterprise users are thriving and SMB users are churning. Any interesting metric movement should immediately trigger the question "for whom?" Cut by segment, plan, geography, and cohort before concluding anything.
Metric drift
The definition of "active user" quietly changed in Q2, and now every historical comparison is wrong. Document metric definitions somewhere permanent, and treat definition changes like API changes: versioned and announced.
What to Actually Watch Weekly
A working setup for most PMs:
- •One north star you check weekly and can explain to anyone
- •Three to five input metrics (activation, a couple of engagement/retention numbers, one revenue number) that feed it
- •Guardrails (performance, error rates, support volume) you glance at for anomalies
- •Everything else on demand, when a specific question arises
If your weekly review takes more than fifteen minutes of looking at numbers, you're watching too many.
One more habit worth stealing: write down what you expect before you look. "I think activation will be flat this week because we shipped nothing relevant." When the number surprises you, that's your model of the product being corrected, which is the entire point of looking. PMs who check dashboards without predictions learn to narrate numbers. PMs who predict first learn how their product actually works. The difference compounds over a year into something that looks a lot like product intuition, because that's what it is.
Metrics in Interviews and On the Job
We see thousands of PM job postings, and "data-driven" appears in most of them. Interviewers test it with questions like "what's the north star for product X?" or "activation dropped 10% overnight, walk me through your diagnosis." The PMs who do well aren't the ones who name-drop tools. They're the ones who reason clearly: define the value moment, pick ratios over counts, segment before concluding.
That reasoning is learnable, and it's the same skill you'll use every week on the job.
Looking for a role where you can put it to work? Browse current PM openings at productmanagerjobboard.com.