Key Startup Metrics

Translation and notes

A readable adaptation of the article's core ideas, with the original source clearly credited.

Metrics are for seeing and choosing, not for filling a dashboard. This free adaptation of Tom Blomfield's YC guide starts with a practical question: if the team could see only a few numbers today, which ones would change the next decision? A large number without a definition, time window, and owner is an invitation to tell a story rather than evidence.

For many subscription businesses, revenue, burn, and runway are a useful base. Revenue shows what customers pay for; burn is the net cash consumed in a month; runway is the time the current cash balance can support the current burn. Put those three numbers at the top of the operating update and keep their definitions stable. Switching between gross and net can create apparent progress where none exists.

Then inspect retention by cohort. Of the customers who started in one month, how many are still active or paying later? A single average can hide a weak new cohort, so segment by start date and customer type. An early consumer product may watch active users; a B2B product may need recurring revenue and active accounts. The right metric follows the business model and the decision it must support.

Every metric needs an owner, a data source, and a reporting period. If producing a number requires stitching together several manual files, first create a small repeatable definition; a complex dashboard cannot repair unreliable data. Keep one explanatory measure beside the outcome measure. More signups without activation, or higher transaction value without more revenue, can be a change in appearance rather than health.

Avoid both hiding a bad number and tracking hundreds of numbers. A visible weak metric brings the conversation back to the constraint and makes runway manageable. Too many metrics let each meeting select the number that fits its preferred story. Limit the weekly review to three questions: what changed, why did it change, and what decision or experiment will make the next change clearer?

At month end, compare each metric with the decision it was meant to support. Drop or demote metrics that never change a decision. When a number is disputed, document its definition and source and assign someone to repair the data. Startup health does not come from more charts; it comes from moving limited resources toward the right problem before runway is gone.

This is a free adaptation, not a word-for-word republication of the original article.