The single most important distinction in measurement. Get it wrong and a team lives in permanent reaction — always responding to yesterday's results, never steering toward tomorrow's outcomes.
▸ Try the interactive toolThis is the single most important distinction in product measurement, and getting it wrong condemns a team to a permanent state of reaction — always responding to yesterday's problems, never steering toward tomorrow's outcomes. Every metric is one of three things: a lagging indicator (a result, visible after the fact), a leading indicator (an early signal that predicts the result), or neither.
A lagging indicator — revenue, churn, retention — tells you what happened. It's accurate but too late to change. A leading indicator — an early-week engagement signal, a usage pattern — moves before the lagging one and predicts it, so you can act while there's still time. The skill is finding leading indicators that genuinely predict your lagging outcomes, then steering by them — because a team that watches only lagging metrics is forever driving by the rear-view mirror.
Lagging — the result (revenue, churn) — accurate but too late to act on.
Leading — an early signal that predicts the lagging metric — actionable in time.
Neither — noise that predicts nothing. Steer by validated leading indicators.
A team that watches only lagging metrics learns of every problem too late to prevent it — it can report outcomes but never steer them.
| The shortcut | What it costs | What it gives you instead |
|---|---|---|
| Only lagging metrics | You learn of problems after they've already happened. | Leading indicators give early warning you can still act on. |
| Permanent reaction | Always fixing yesterday, never steering tomorrow. | Leading indicators let you steer proactively. |
| False leading indicators | An early metric that doesn't actually predict the outcome. | Validating the leading–lagging link ensures real predictive power. |
| Confusing the two | Treating a lagging result as if it were actionable. | Knowing which is which directs effort to what you can still influence. |
For every metric you track, ask: is this a result (lagging), an early predictive signal (leading), or noise? Most teams over-weight lagging metrics without realising.
Name the results that matter — retention, revenue, churn. These are what you ultimately want to move, even though they're too late to act on directly.
For each lagging outcome, find an earlier metric that reliably moves before it. If retention is the lagging outcome, early engagement depth might be the leading signal.
A leading indicator is only useful if it genuinely predicts the lagging one. Check the relationship with data — an unvalidated 'leading' metric is just noise you're trusting.
Watch and act on the validated leading metrics through the period, so you can influence the lagging outcome while there's still time — rather than discovering the miss after the fact.
A team watched its lagging metrics diligently — monthly retention, revenue — and reacted whenever they dipped. But by the time retention dropped, the users were already gone; every response was a post-mortem. They were driving entirely by the rear-view mirror, forever fixing problems that had already cost them.
Identifying a leading indicator changed this. They found that early-week engagement depth reliably predicted whether a cohort would retain — and it was visible weeks before retention itself moved. Steering by that leading signal, they could intervene with at-risk cohorts before they churned, turning a lagging post-mortem into a proactive save. The lagging metric still mattered as the scoreboard; the leading one was the steering wheel.
The deliverable is your metrics classified, with validated leading indicators paired to the lagging outcomes they predict.
| Type | Example | Use |
|---|---|---|
| Lagging | Retention, revenue, churn | The scoreboard — accurate, too late |
| Leading | Early engagement, usage depth | The steering wheel — act in time |
| Neither | Vanity counts | Noise — discard |
Lagging indicators are the scoreboard; leading indicators are the steering wheel. A team with only the scoreboard knows whether it's winning but can't change the outcome — which is the difference between reporting and steering.
The hard part is the validation. It's easy to declare some early metric a 'leading indicator' because it's convenient to watch, but if it doesn't actually predict the lagging outcome, you're steering by a fiction — confidently acting on a signal that means nothing. The genuine skill is finding leading indicators whose predictive link to the lagging outcome is real and checked with data. Done well, this transforms a team's relationship with its metrics: instead of holding monthly post-mortems on results that are already fixed, it watches early signals it can still influence and intervenes in time. This is also the foundation of the advanced-OKR distinction from Module 2 — leading and lagging key results are the same idea, applied to goals.
Condemns you to permanent reaction. Find leading indicators to steer by.
An early metric that doesn't predict the outcome is noise. Validate the link with data.
Treating a result as actionable wastes effort. Know which you can still influence.
Leading indicators serve the lagging outcomes — don't lose sight of the results that matter.
Business outcomes lag; product/feature metrics often lead (Tool 02).
Leading and lagging key results (Module 2, Tool 26) are this distinction applied to goals.
Checking the leading–lagging link is a statistical exercise (Tool 17).
Leading input metrics are the lower branches that drive the lagging North Star (Tool 05).
Pick a lagging outcome you care about (retention, revenue). Brainstorm an earlier metric that might predict it — something visible before the outcome moves.
Then ask the validation question: how would you check that this leading metric actually predicts the lagging one, rather than just feeling like it should?
If you can name a validated leading indicator you'd genuinely act on, you've found your steering wheel — and escaped the rear-view-mirror trap of lagging-only measurement.