Confidence interval
A confidence interval is the range your true result probably sits in, given that you only measured a sample. If an A/B test reports a 12% lift with a 95% confidence interval of 4% to 20%, it's saying the real lift is likely somewhere in that band, not exactly 12%. The wider the band, the less you actually know.
For a founder this is the antidote to over-reading a single number. A flattering point estimate with a band that crosses zero means the effect might be nothing at all. Smaller samples give wider intervals, which is why early results swing wildly and shouldn't drive big bets. The practical habit: never quote a conversion lift or metric change without the interval, and treat any result whose band includes 'no change' as unproven. It pairs with statistical significance, the interval shows the size and uncertainty, significance shows whether it's real.