The Three Kinds of Metrics Every Product Actually Has
The hardest part about metrics is not collecting them. It is interpreting them correctly. When a number moves, teams often jump straight to conclusions without asking what kind of signal the metric actually represents.
A while back, my team brought a metric into a review that looked genuinely encouraging. Checkout engagement was up 12%. We had been iterating on the experience, so the lift was plausible, and the instinct was to take it as a clean win. But in my experience, any metric that feels too “headline ready” deserves one more step of rigor. We asked a simple question as a team: is engagement up because checkout is easier, or because customers are getting stuck and clicking more? “engagement” in this case was checkout button taps per session. Useful, but it left a lot unsaid about whether customers were actually succeeding.
So we did what we usually do when a metric could be telling multiple stories. We decomposed it into three signals using a framework I now apply to most reviews. First, an input metric: the percentage of users who actually saw the new checkout flow. That confirmed the change was truly in market. Next, an output metric: purchase completion rate, which answered the real question of whether customers finished the transaction. Finally, a health metric: payment retry rate, capturing how often the system forced a second attempt. We found that users were tapping more because some payments were failing or requiring retries, not because the checkout experience had improved. So the engagement metric went up, but part of that lift was friction. Put side by side, the interpretation got sharper. Exposure was rising and taps were higher, but completion rate had barely moved and retries were trending up. The lift was not fake, it was mixed. Some of it looked like improved interaction, and some of it looked like friction.
That moment reinforced why I like this framework. Most metrics fall into one of three categories: Input, Output, or Health. Input metrics tell you whether the team built and delivered something. Output metrics tell you whether users achieved the outcome. Health metrics tell you whether the system delivered the experience reliably and sustainably. When those categories get mixed together, dashboards can create confidence without clarity. Now when a metric shows up in a review, the first question I ask is simple: is this an input, an output, or a health signal? Very often that classification explains the number better than the number itself.