Correlation Is a Clue, Not a Verdict
Evaluate an observed relationship by checking timing, alternatives, mechanism, and study design.
Do not turn an association into a cause without a design. Timing The cause must happen before the outcome. If timing is ambiguous, the relationship may be reverse causation or shared movement. Alternatives Look for common causes. Customer size, tenure, channel, geography, and plan type often explain both the behavior you observe and the outcome you care about. Comparability A comparison is stronger when groups are similar before the exposure. Randomization is one way to create comparability; matching and adjustment are weaker but useful when done honestly. Mechanism A causal story should explain how the action changes behavior. Without a…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in