A framework for metrics

Most companies want to be "data driven," but there are few concrete, prescriptive guides to what this actually means. This is an opinionated framework derived from my experiences and conversations with other startups.

First, choose the single most important metric for the company. This metric should correspond directly to the success of the company, for whatever definition success means for you in this quarter. For some companies, it's revenue growth, for other it's number of active users, etc. You should make this a growth rate, instead of a target number. Make this number very public - put it on a board somewhere, send it out in daily emails, present it every week, whatever you need to do to make sure people see this number and feel good when it goes up and motivated when it goes down.

Next, recognize that it's very hard to move that "primary" metric directly. Work with your team to put forward roughly 5 metrics that you hypothesize will have a material correlation with that primary metric, but that are more "tangible" and can be impacted more directly. Select these metrics based on what you believe will be the quality of their correlation, not on how easy they are to measure. Look out for metrics that could be negatively correlated with your primary metric, for instance driving down support tickets per customer could increase NPS score. If your primary metric is week-over-week growth in number of active users, some good secondary measures might be number of new visitors per day, conversion rate between new visitors and active users, churn rate of active users, average number of users referred per active user, etc.

For each of those 5-10 secondary metrics, assign them an owner. The owner picks a target number to hit by a target date. Write these down. Also, write down your expectations as to how these secondary metrics will impact the primary one. As an example: "If we can increase number of new visitors to our site by 20% in the next month, we predict that will increase the number of weekly actives by 12% over the same month period."

Now, use those secondary metrics to drive product decisions, and attempt to make the secondary metrics move how you want them to. Run experiments and see how they affect the secondary metrics. Every few weeks, check in with the owners of each secondary metric, review progress towards their goal, and make corrections or resource allocations as appropriate. Also, check that the secondary metrics are impacting the primary metrics as you expect.

Your primary target will rarely change, if ever, but over time you'll end up swapping in and out secondary metrics as you emphasize different aspects of growth or revenue. You will also learn which secondary metrics are meaningful and which are not - maybe pages per visit is the most important indicator of retention, or maybe not. Maybe it's much easier to move one metric than it is to move another, and both have similar impact on your primary goal. Keep an eye out for new things to measure, but make sure the number of secondary metrics you're focused on stays small, and that each metric has targets and people responsible for making sure the metric hits the target.

As your team expands beyond about 40 people, you may be able to add an additional primary metric, dividing the organization into groups that focus on one or the other. Each primary metric should still be broken down into secondary metrics using the framework above. As the organization grows, secondary metrics become a team's primary metric, which they then subdivide into their own secondary metrics, and the framework above gradually starts to look like OKRs.