Glossary
Conversion rate
Conversions divided by a base such as sessions, users or clicks. Tools pick different bases, so one campaign can show a different conversion rate in each tool.
Also called: CVR
Everyone agrees what goes on top. The argument is always underneath.
Divide by sessions and a returning visitor who browses three times before buying counts as three, so the rate drops. Divide by users and they count once, so the rate rises — and now your figure depends on how well you identify returning people, which is an identity resolution question, not a marketing one. Divide by clicks and you are measuring the ad platform’s count, which will not match your site’s. Divide by new users only and you get a different number again.
None of those is wrong. They answer different questions. But they are all called “conversion rate”, and tools pick their default without announcing it.
Two consequences worth holding on to.
A rate can move with no change in behaviour. Change the session timeout, improve returning-visitor identification, or start filtering bot traffic, and the denominator shifts. The rate moves. Nothing about the marketing changed.
Rate and volume can move in opposite directions, and usually should. Broaden targeting and you reach more people who were less likely to buy: conversions go up, rate goes down. Cutting the campaign to protect the rate is optimising the metric rather than the business. A rate is only interpretable next to the volume it came from — which is why it is a diagnostic, not a target.
Do not confuse with
Close enough to get mixed up, different enough that the mix-up costs something.
- Conversion An outcome you chose to count, such as a purchase, a lead form or a trial start. Each team sets its own definition, so two reports can count different events.
- Session A group of one visitor's interactions treated as a single visit. Each tool applies its own timeout and reset rules, so session counts differ between tools.
- Customer lifetime value The total revenue or margin a customer brings over the whole relationship. Most CLV figures are predictions from past purchases, so check the model behind them.