Conversion rate
The percent of people who did the thing you wanted, out of everyone who had the chance. Completions divided by eligible people.
See it
What it is
Completions divided by everyone who had the chance, times one hundred. The arithmetic is trivial and the argument is always about the denominator: the same 400 signups is a 4% rate over sessions, 6% over unique users, and 22% over people who actually reached the pricing page. None of those is wrong, but two teams quoting different ones will spend a month disagreeing about nothing.
Use a rate whenever you are comparing across time, channels, or test variants where the volume differs. Raw counts fall apart the moment traffic moves: 'signups doubled' means nothing if a campaign tripled the visitors. Write the denominator into the metric name, so it lives as 'checkout conversion rate (of users who added to cart)' rather than a bare percentage.
Gotcha: small denominators produce nonsense. Two out of five is 40% and tells you exactly nothing, so suppress the number below a floor instead of shipping it to a dashboard. And watch for Simpson's paradox: the overall rate can drop in a week where every single segment improved, purely because cheaper traffic changed the mix.
Ask AI for it
Add a conversion rate metric for [action] to this dashboard. Define the denominator explicitly as unique users who reached [entry point] in the period, and print that definition under the number. Show the percentage to one decimal, the raw numerator over denominator, a 95% Wilson score confidence interval, and the change versus the previous equal-length period. If the denominator is under 100, render 'not enough data' instead of a percentage. Do not average daily rates together; recompute from summed counts over the whole range.