Leading and Lagging Indicators
Date: 2026-08-16
Lagging indicators tell you what happened; leading ones tell you early, and less reliably. Every leading indicator is a bet that a correlation will hold — and the bet gets worse the moment anyone starts optimising the indicator.
What it is
A lagging indicator measures the outcome you care about, after it’s happened. Revenue, orders, retention, margin.
A leading indicator measures something earlier in the chain that predicts it. Add-to-cart rate, email signups, first-week repeat visits.
leading lagging
add_to_cart rate ──▶ checkout entry ──▶ orders ──▶ revenue ──▶ LTV
fast, noisy, slow, meaningful,
partial unactionable in time
The trade
| Leading | Lagging | |
|---|---|---|
| Available | Immediately | Weeks or months |
| Sensitive enough to test on | Usually | Often not — Metric Sensitivity |
| Certainly matters | No | Yes |
| Gameable | Very | Less |
You can act on a leading indicator and can’t be sure it matters. You can be sure a lagging indicator matters and can’t act in time.
In plain terms: you’re choosing between a number you trust that arrives too late, and a number that arrives now that might be measuring nothing.
The predictive link is an assumption
The whole value rests on the correlation holding. It usually holds until someone targets the indicator, at which point they find a route to the indicator that bypasses the outcome.
- Add-to-cart rate rises if the button is ambiguous, and orders don’t follow
- Email signups rise with an aggressive popup, and the addresses are worse
- Engagement rises when the site is confusing — Engagement Metrics
That’s Goodhart’s law: a measure that becomes a target stops being a good measure. The mechanism is specific — the correlation was produced by the old behaviour, and targeting the measure creates new behaviour the correlation was never estimated on.
Validate the link periodically rather than assuming it. Do users who add to cart still convert at the historical rate? If the relationship has weakened, the indicator has stopped working and nobody will have noticed.
Choosing a leading indicator
Three conditions, all necessary:
- Causally upstream, not merely correlated. Add-to-cart is a step on the path to purchase. “Visited the blog” correlates with buying because both correlate with intent — targeting it does nothing
- Sensitive enough to move detectably at your traffic — Metric Sensitivity
- Hard to game without doing the real thing. The best leading indicators are ones where the only route to improving them is the outcome you wanted
Condition 3 is the one that gets skipped and the one that matters most.
In a testing programme
The standard arrangement:
- Primary metric: the most sensitive thing plausibly implying the business outcome — usually conversion rate — Overall Evaluation Criterion
- Secondary: revenue per visitor, reported with its interval, not decision-eligible
- Guardrails: the lagging outcomes that must not get worse — returns, refunds, complaints — Guardrail Metrics
- Holdout Groups for the genuinely lagging effects a two-week test can’t see
That last one is how you check the leading indicators weren’t lying, cumulatively, over a year.
Where it bites in retail
The longest lags are the most important numbers:
- Repeat purchase — months for a considered category — Repeat Purchase Rate
- Returns — weeks after the order, and they reverse revenue you already counted
- Lifetime value — a year or more, and always partly modelled — Customer Lifetime Value
A change that lifts first-purchase conversion while raising returns or suppressing repeat purchase is a loss recorded as a win. The lagging indicators are the only thing that catches it, and they arrive after the decision has been made and celebrated.