Intention-to-Treat Error

The statistical error of analyzing only the people who completed a treatment rather than everyone who was assigned to it—creating a biased picture of the treatment's effectiveness. If 100 people start a drug trial, 30 drop out due to side effects, and the remaining 70 show improvement, analyzing only the 70 completers overstates the drug's effectiveness. The same error pervades business: analyzing only customers who completed onboarding (ignoring those who dropped out), only employees who stayed (ignoring those who quit), or only projects that finished (ignoring those that were abandoned). Survivorship bias is a form of intention-to-treat error.

When to use it

When success metrics seem too good (you may be excluding failures from the analysis); when program evaluation only includes completers; when customer or employee metrics show improvement but the base keeps shrinking; when any analysis might be suffering from survivorship bias.

How it can help

Always analyze based on the FULL population that started, not just those who completed. In customer analysis: include churn in your success metrics—a product that converts 10% of sign-ups but retains 90% may be worse than one that converts 20% but retains 50%. In employee programs: measure everyone who was enrolled, not just those who graduated. In investment: include failed investments in your portfolio returns, not just the winners. The rule: if someone dropped out or failed, they're still in your denominator. Removing them produces an optimistic illusion.