analytical · standard
Leading vs lagging indicators: PM interview answer
What is the difference between a leading and lagging indicator? Give an example.
This is a proxy question. The interviewer is not testing your vocabulary; they are checking whether you build dashboards that give early warning or ones that only explain what already went wrong.
What the terms mean
A lagging indicator measures an outcome that has already occurred: revenue, churn rate, NPS score. The data is precise but arrives too late to adjust anything within the cycle that produced it.
A leading indicator measures a behavior that, based on empirical evidence, reliably predicts that lagging outcome before it occurs. The key phrase is “based on evidence.” A metric is not a leading indicator because you hope it matters. It earns that label after you validate the relationship.
The causal chain is what separates a measurement strategy from metric soup. A PM must be able to state: “When users do X within the first N days, they are Y times more likely to retain at month three.” Without the hypothesis and the validation, you are tracking noise.
The canonical examples
Facebook found that connecting with seven friends in ten days predicted long-term retention strongly enough to anchor onboarding. That metric was not declared a leading indicator by intuition; it was validated against retention data first, then promoted to a north star that shaped product decisions.
Slack’s internal signal was 2,000 messages sent within a team. Teams that hit that threshold had near-zero churn. So Slack built onboarding flows explicitly designed to reach it. The leading indicator (messages sent in week one) predicted the lagging outcome (churn at month three) with enough confidence to bet engineering time on.
Twitter used “follows made in first session” as a leading indicator for 30-day retention. Same logic: identify the early behavior that predicts the downstream outcome, validate the relationship, then optimize for the behavior.
Structure a strong answer
strong
"A lagging indicator measures an outcome after it happens: churn, revenue, NPS. A leading indicator measures a behavior that predicts that outcome before it occurs, and the word 'predicts' is doing real work here: it means the relationship has been tested, not assumed. For Slack, message volume in the first week predicted whether a team would churn in month three. That's not a guess. Slack validated it, which is why they built onboarding to hit the 2,000-message threshold. The practical difference: monthly churn tells you which customers are already gone; messages-per-day in week one tells you which new teams are at risk so you can act before they cancel. In terms of where each belongs: lagging indicators belong at the business or company level to confirm strategy is working; leading indicators belong at the team level for weekly decisions and OKR Key Results, because they are the only ones you can actually move within a quarter."
weak
"Revenue is lagging, so we should track leading indicators like NPS and feature adoption." This fails on three counts: NPS is itself a lagging indicator: it measures satisfaction after the fact. Feature adoption is only a leading indicator if you have tested its correlation to a specific outcome. And nothing here shows a causal chain. Interviewers hear this constantly; it signals a PM who reads dashboards rather than designs them.
The 2026 angle: AI and agent products
Most prep resources miss this entirely. AI and agent products have a longer, less visible causal chain than classical SaaS. When your product is an agent doing autonomous work, the lagging indicator (did the user’s underlying business outcome improve?) can be weeks downstream.
This forces more deliberate leading signal selection. For an AI agent: task completion rate, time-to-first-useful-output, and handoff or escalation rate are your leading indicators. Downstream outcomes (support tickets deflected, revenue influenced) are lagging. A PM who can name that distinction and defend why task completion rate predicts value delivery, rather than defaulting to engagement proxies, is operating at a meaningfully higher level than someone who recites the Facebook example and stops.
This connects directly to the 2026 PM mandate: viability requires leading indicators that predict actual value delivery, not just usage activity.
Where each type belongs
- Lagging indicators: company or business-unit OKRs, board reporting, strategy confirmation
- Leading indicators: team-level Key Results, sprint decisions, onboarding optimization, at-risk user detection
- The test for any proposed leading indicator: can you articulate the causal hypothesis? Have you validated it against data? Can the team actually influence it? If the answer to any of those is no, it is not yet a leading indicator.
For adjacent interview questions, see how to measure success for a new AI product and what to do when two metrics conflict.