The maths
CPA = CPC ÷ conversion rate
Uplift = visitors × (target rate − current rate)
The middle line is why conversion rate deserves more attention than it usually receives. Cost per acquisition is click price divided by conversion rate, so the two levers are mathematically equivalent. Improving conversion rate from 1% to 2% has the same effect on CPA as negotiating your click price in half, and it is almost always the easier of the two.
The third line is the one to take into a meeting. It converts a percentage improvement into a number of additional conversions from traffic you are already paying for, which is a far more persuasive way to argue for landing page work than quoting a rate.
Reading it properly
Decide what the denominator is before you compare anything. Sessions, users, clicks and landing page views are all defensible choices, and they produce materially different rates from identical data. Most disagreements about conversion rate that I have watched turn out to be disagreements about the denominator.
Expect the rate to vary enormously by traffic source, device and intent. Branded search converts far better than cold display, and always will. A single site-wide conversion rate averages away exactly the information you would use to act.
Common mistakes
- Changing the denominator mid-comparison. Switching between sessions and users makes performance appear to change when nothing has.
- Averaging across sources. Blending branded search with prospecting display hides both the thing that is working and the thing that is not.
- Declaring a test winner too early. Small samples produce large apparent differences. Stopping a test the moment it looks good is how teams ship changes that do nothing.
- Optimising the page while ignoring the promise. If the ad set an expectation the page does not meet, the fix belongs in the ad.
Embed this calculator
Free to use on your own site, for your team or your students. The link inside is the only attribution required, and I will not add tracking to it later.
Questions people actually ask
What is a good conversion rate?
It varies so much by industry, traffic source, device and price point that a single benchmark would mislead you. The comparison that matters is your own rate over time, segmented by source. A branded search visitor and a cold display visitor should never be judged against the same standard.
Should I use sessions or users as the denominator?
Either, provided you are consistent and say which. Sessions suit campaign analysis because a visit maps to a click. Users suit business reporting because a person maps to a customer. Problems begin when the two are mixed inside one comparison.
How do I improve conversion rate?
Start with message match between ad and landing page, then page speed, then the friction in your form or checkout, then clarity of the offer. In my experience the largest gains usually come from removing steps and matching expectations, not from redesigns.
How much traffic do I need to test a change?
Enough that the difference you are looking for is larger than the noise. Smaller expected improvements need considerably more traffic. If a test looks conclusive within hours, it almost certainly is not.
Why is my conversion rate falling while traffic grows?
Usually because the additional traffic is less qualified. Scaling reaches people with less existing intent, so the rate falls even though total conversions rise. Check whether conversions in absolute terms went up before treating it as a problem.
Go deeper: How I run performance marketing · Measurement & analytics · Field notes