Setting goals is hard. Actually, setting is not the hard part, tracking them continuously is. But before tracking them, it's important to understand the distinction between key results and key performance indicators.
I recently learned a great analogy for thinking about key results and key performance indicators from the great Fabian Zeller.
Key Performance Indicators (KPIs)
Often confused with OKRs or key results, key performance indicators are long-term metrics. They show and tell us how we're doing. Not where we're going.
You can think of them as a car dashboard, you can see how fast you're going, how much fuel you have, how much time you have left.
This translate within software engineering to examples like the following.
- Uptime percentage (e.g. "99.9% uptime")
- Average response time (e.g. "API response < 250ms")
- Number of active users per month
Key Results
Key results themselves are about the path to the goal. They give direction in a given timeframe. They are temporary and based on key performance indicators.
You can think of them as a route on a map, they tell you how far you've traveled and where you're going next.
Oppositely to a route on a map, it's essential to remember that actually achieving the goal is not required. The main purpose of defining the pin on the map is to provide direction throughout a specific timeframe. Once that period ends, the goal is archived and you start fresh, regardless of whether you fully reached the goal or not.
This would usually translate within software engineering to examples like the following.
- Increase monthly active users by 20% until Q1 2026
- Reduce average response time by 10% until Q3 2029
- Reduce bug count in production by 10% until Q1 2029
But here's the catch: Most of these are output metrics, which we have little control over.
Input Metrics
Another mistake we do when setting key results, is by setting key results based on output metrics, like I just mentioned. Most often it's super hard to estimate the output metrics, as you can barely control them. However, what you can control is the input metrics. Essentially key results are bets to invest in input metrics to achieve the output metrics.
So instead of setting a key result like "Increase monthly active users by 20% until Q1 2026", you can set a key result like "Increase % of new users completing onboarding by 20% until Q1 2026". Changing the onboarding completion rate is much easier to control than the monthly active users.
The following is a list of input and output metrics.
| Input Metric | Output Metric |
|---|---|
| % of new users completing onboarding | MAU (Monthly Active Users) |
| # of notifications sent | DAU/MAU ratio (stickiness) |
| Avg sessions per user per week | Retention |
| Feature adoption rate | Conversion rate |
| Load time / crash rate | Revenue or ARPU |
Adaptability
Another thing to keep in mind is, that if you are not able to adapt key results, you already created obsolete ones. They will always change, the same way as your environment, your team, and your product changes. Be able to delete them, as soon as they are not relevant anymore, else they become a weird burden nobody wants to carry.