· Quick Take · 2 min read
Your Primary Metric: The One Number That Drives Everything
Like ML's loss function, your product needs one metric to optimize while others serve as guardrails. Choose wisely.
Choosing the right primary metric (aka the North Star metric) for your project is critical. It can guide you in the right or the wrong direction.
In Machine Learning, we have many metrics, but there is one primary metric called the “loss function.” The whole training process is only to optimize that one thing. You can care about many things, but you should pick one primary metric. Everything else becomes a constraint or a guardrail.
Same idea for your product or your team. You will mostly optimize for your primary metric. It might be revenue, profit, time saved, tasks completed, or success rate. The key is: one number leads, the others protect.
A common mistake in software-as-a-service is chasing “user engagement time”, the minutes a user stays in your app. That’s great for content sites, where the goal is to keep people reading or watching. But most SaaS tools are for getting a job done. If you improve your PDF converter and make it faster, then what will happen? You may see engagement time going down, because users get their job done faster. Time saving makes them happier, and they’ll come back. A better primary metric here might be to maximize percentage of job completions or minimize time to first one.
Another famous example is Wells Fargo, where upper management pushed hard on account-opening targets. It turned out employees found a way to hit the number by opening unauthorized accounts. The metric was easy to game and not tied to real customer value. When a number becomes the target, people will game it. So pick a metric that aligns with the company’s value.
So think hard about the primary metric for your product or your team. Then commit. Don’t swap your north star every month; only change it when your strategy truly changes, and communicate clearly when you do.
