~/learn/Product Metrics · Metric foundations
Input, output & proxy metrics
Output metrics show results; input metrics are the levers; proxies stand in for slow outcomes.
why this matters
Interviewers probe whether you know which metrics a team can actually move this quarter.
Output metrics are the results you ultimately care about: revenue, retention, bookings. They're lagging — by the time they move, the cause happened weeks ago. Input metrics are the actions teams can directly influence: items listed, search success rate, delivery time. Amazon famously focuses teams on inputs (selection, price, speed) because outputs follow.
A proxy metric is a fast-moving stand-in for a slow outcome. 90-day retention takes 90 days to measure, so a team might use 'number of active days in week 1' — but only after showing that week-1 activity actually predicts 90-day retention.
The chain
inputs (levers) → outputs (results) listing quality bookings delivery speed repeat orders search success revenue proxy = fast metric proven to predict a slow output
Interview trap
A proxy that isn't validated can mislead you. 'Time on site' looks like engagement, but if people stay longer because search is broken, it's measuring frustration.
Pro tip
When you propose a proxy in an interview, say how you'd validate it: 'I'd check historically that users with 3+ active days in week 1 retain much better at day 90.'
key takeaways
- Outputs lag; inputs are levers.
- Proxies stand in for slow outcomes.
- Validate a proxy before trusting it.
check your understanding
A team uses 'week-1 active days' instead of 90-day retention. What must be true for that to work?