Customers & Markets
Use Data to Build a Better Startup Faster
How to choose the one number that matters at your stage, measure it honestly, and turn it into evidence of traction.
The big idea
Croll and Yoskovitz argue that startups should focus on one metric that matters most at a given time, chosen according to their stage and business model. They lay out five stages, empathy, stickiness, virality, revenue and scale, and argue that companies should move through them in order. Good metrics are comparative, understandable, ratios or rates, and capable of changing behavior. Vanity metrics that only go up feel good but hide problems, while cohort analysis and pre-set targets keep founders honest.
Key takeaways
At any stage, one number matters more than the rest for deciding what to do next. Tracking a dozen indicators equally means none drives decisions. The metric changes as the company evolves; it is a spotlight on the current problem, not a permanent scoreboard. When the current metric reaches its target, choose the next one.
A useful metric is comparative, across time periods or user groups; understandable, so people remember and discuss it; a ratio or rate rather than a raw count; and behavior-changing, meaning you would act differently if it moved. Run every number you report through those filters.
Cumulative totals such as all-time users, downloads or page views can never deliver bad news, which is why they are tempting under pressure. Experienced investors discount them, and relying on them lets you avoid the real diagnosis. Replace them with rates and trends, such as active users as a share of total, tracked over time.
Empathy: confirm a real problem people will pay to solve. Stickiness: build something they return to. Virality: grow through users. Revenue: prove the economics work. Scale: grow a proven engine. Skipping stickiness and paying for growth into a product people abandon wastes money, because new users leave as fast as they arrive.
The book covers several business models, including software as a service, e-commerce, marketplaces, media, user-generated content and mobile apps, each with different key numbers. Subscription businesses watch churn and retention; marketplaces watch how often listings actually transact; e-commerce watches repeat purchases. Borrowing another model's metrics leads you to optimize the wrong thing.
Decide in advance what result counts as success, and write it down. People are skilled at rationalizing whatever number comes back, and an experiment with no threshold always seems to have kind of worked. A pre-set target turns a debate into a decision and saves weeks of defending a weak result.
Split each metric by when customers joined, who they are and where they came from. A flat average can hide steady improvement in newer cohorts, or one strong segment carrying several weak ones. A cohort chart showing improvement is persuasive evidence for investors, and the habit protects you from misleading yourself.
The framework
Find a real problem people will pay to solve.
Build something people come back to; retention before growth.
Grow through users bringing in other users.
Show customer value exceeds acquisition cost, repeatably.
Invest to grow a proven engine and expand into new markets.
Apply it this week
Read next
A method for building under uncertainty: test your riskiest assumptions with small experiments and let evidence decide what to build next.
A systematic way to test a business model on one page, measure traction, and sell before you build.
A practical case for OKRs: a simple goal system that turns a founder's priorities into numbers the whole team can grade.
This page is an original summary and commentary prepared by XRAYCLOUD.AI with AI assistance. It is not affiliated with or endorsed by the author or publisher, and it is no substitute for the book — read it for the full argument.