- A metric is anything you measure; a KPI is the small set that decides success.
- Most teams have 30 metrics and 0 KPIs. Pick 3 to 5.
- Separate input/output/outcome and leading/lagging; choose one North Star, kill vanity metrics.
Walk into the average product team's review and you'll see a slide with thirty numbers on it. Conversion rate, weekly active users, NPS, support tickets opened, support tickets closed, churn, expansion revenue, page load time, error rate, cohort retention, feature adoption. The list keeps going. The team will spend an hour walking through movements in each one. By the end, nobody can tell you what the team is actually trying to achieve, because the act of measuring thirty things has crowded out the question of which two or three actually matter.
This is the metric-vs-KPI distinction, and it sounds like a vocabulary problem but isn't. It's a discipline problem. A metric is anything you measure. A KPI, a Key Performance Indicator, is one of the small handful of measurements that actually tell you whether the business is winning. The first list is theoretically unbounded. The second list, done properly, is three to five items long. Most teams have the first list. Very few have the second.
The definitions, sharp.
Metric: anything quantified. A measurement of any behaviour, output, or state. Useful for diagnosis, debugging and curiosity. Not, in itself, a basis for decisions.
KPI: a metric the team has elevated to key. It means that if this number doesn't move in the right direction, the strategy is failing, and the team will change what they're doing. The promotion from metric to KPI is the act of attaching a decision to a number. A metric becomes a KPI when not moving it would cost someone their job.
By this definition, most "KPIs" on most dashboards aren't KPIs. They're metrics in a dashboard tile. Nothing changes when they move, because the team has no agreed action to take if they go up, down, or sideways. That's the test. If the number went the wrong way for two months and nobody changed plan, it wasn't a KPI.
The metric pyramid.
Input · output · outcome.
Within the pyramid, metrics divide into three kinds, and the kinds matter, because teams almost always over-index on the easiest one to measure. Input metrics measure the work the team does: releases shipped, tickets closed, features delivered. Output metrics measure what the product does: page views, sessions, transactions. Outcome metrics measure what changes for the user or the business: revenue, retention, time saved, behaviour shifted.
Outcome metrics are the only ones that matter as KPIs. Input and output metrics are diagnostic. They're useful for understanding why the outcome moved, but not the things a team should be optimising directly. A team that ships ten features a sprint (input) which generate a million page views (output) but no improvement in retention or revenue (outcome) is busy, not effective. The risk of an input-heavy dashboard is that it shows green when the strategy is failing.
KPI selection: 3 to 5 max.
The hard work is choosing which three to five metrics get promoted to KPI status. The discipline is partly mathematical and partly political. Mathematically: the KPIs should together capture whether the product is delivering its strategic promise. If you can't say what your strategy is in two sentences, you can't select KPIs. Politically: agreeing the KPIs forces leadership to commit publicly to what the product is for, which is uncomfortable, which is why most teams have thirty metrics instead.
A defensible KPI set for an early-stage SaaS team might be: weekly active users (engagement), week-4 retention (stickiness), net revenue retention (commercial health), customer-effort score (experience quality). Four KPIs. Each ties to a different facet of the business model. Each has a clear direction of travel. Each, if it moves the wrong way for a quarter, will force the team to change plan.
Leading vs lagging.
The other distinction worth getting right: lagging indicators tell you what already happened (revenue, churn, NPS). They're authoritative but slow. Leading indicators predict where the lagging metric is heading (activation rate, week-1 retention, demo bookings). They're noisier but actionable. A good KPI set mixes both. All lagging and you'll find out you're failing after it's too late. All leading and you'll chase signals that don't show up in the books.
The North Star Metric.
Sean Ellis popularised the North Star Metric as the single number that best captures whether the product is delivering its core value to the user. Amplitude's playbook on it is the standard reference. The North Star sits above the KPIs, one metric rather than three to five, and its job is to align the whole company. Spotify's was hours of music listened to. Airbnb's was nights booked. Facebook's was monthly active users.
The trap with North Star is that everyone wants one, but the metric people choose is usually a vanity proxy. A North Star Metric should: capture the core value the product delivers; correlate strongly with revenue over time; be something the team can move with its actions; and be measurable cheaply and quickly. "Brand awareness" fails on every criterion. "Active users" fails three out of four. The honest North Star is harder to find than the marketing version suggests.
Vanity metrics — Ries was right.
Eric Ries's distinction between vanity and actionable metrics (from N°19 · Lean Startup) cuts the same surface from a different angle. A vanity metric goes up over time but can't tell you why or what to do about it: cumulative signups, total downloads, app store rank. An actionable metric is tied to a specific behaviour change and a specific cohort: activation rate of this week's new users, compared to last week's.
The simple field test: would a 10% movement in this metric, week-over-week, cause the team to change anything? If yes, it's actionable. If no, it's vanity, no matter how nicely it grows.
Dashboard fatigue.
The natural endpoint of metrics-without-discipline is dashboard fatigue. The team has access to forty charts. They look at none of them. The morning ritual of opening the dashboard becomes a five-minute scroll and a vague feeling that things are mostly fine. When a number is genuinely moving, nobody catches it because the signal is buried in the noise of thirty-nine other numbers wobbling around their baselines.
The fix isn't better dashboards. It's fewer numbers, with clearer thresholds. If your KPI dashboard has more than five tiles, you don't have a KPI dashboard. You have a metrics dashboard you're calling a KPI dashboard.
How to retire metrics.
Most teams add metrics. Few retire them. The result is the dashboard sprawl described above. Every quarter another metric gets added because a stakeholder wanted visibility, and nothing gets removed because that's politically awkward. Retiring metrics is its own discipline. Two rules help.
First: every new metric requires retiring an old one. One in, one out. Forces the team to confront which existing metric is least valuable. Second: metrics get an explicit review every quarter. For each one, what decision did this metric drive in the last 90 days? If the answer is "none," it's diagnostic at best and clutter at worst. Move it off the main dashboard. The metric still exists in the data warehouse, available when needed. It's just not in the team's daily field of view.
Common KPI failure modes.
- Too many. Ten "KPIs" is metrics-in-disguise. Cut to five.
- All output, no outcome. Page views, sessions, transactions. All measuring whether the product is being used, none measuring whether using it changed anything.
- All lagging. Revenue and churn only. You'll find out you're failing one quarter too late to do anything about it.
- Owned by no one. A KPI without a name attached is a wish. Each KPI needs a single owner who'll be held accountable for it next quarter.
- Goodhart's Law. "When a measure becomes a target, it ceases to be a good measure." Choose KPIs that are hard to game without delivering the underlying value.
How we use this at Product Pieces.
When we run a metrics review with a team, the first artefact we produce is the pyramid. North Star at the top, three to five KPIs in the middle, the rest of the metrics, twenty or thirty, at the base. The exercise of forcing the team to choose which metrics get promoted, and which get demoted, is most of the value. Leadership teams almost always disagree about which metric matters most. Surfacing that disagreement is the work — naming the strategy in the form of a single number you can rally behind.
The second artefact is the retirement list. Every metric currently on the dashboard that didn't drive a decision in the last quarter. Sometimes that's twelve metrics. Removing them isn't theatre. It's the act of focusing the team on what actually matters.
Next issue: N°26, The CPO Piece. The most senior product seat, and what it actually decides.