What is a conversion rate, and how do I work out mine?

Where the definition comes from

No one body owns the term. The definitions here are Google's, because its advertising and analytics tools print the figure. They do not all divide by the same thing.

Google Analytics calls the action a key event. Its page on key events says a key event measures an action that is particularly important to the success of your business. Its list of metrics then gives two rates. This is the first.

The second is the session key event rate: the sessions in which a key event happened, divided by all sessions. A session is one visit, and one person can make several. So Google alone has three bottom numbers: ad interactions, sessions and users. "My conversion rate is 3%" says nothing until it says 3% of what.

How to work it out, step by step

1. Pick one step. Name the action, such as signed up or paid, and who had the chance to take it, such as everyone who visited the page.
2. Pick one period. The same start and end for both counts.
3. Count who did it. Count people, not actions, where one person can do the thing twice.
4. Count who had the chance.
5. Divide the first by the second, and multiply by 100.

An example, in round numbers, for one month:

The same product is at 3%, 10% and 0.3% at once, which is why a rate has to name its step. The 200 ad clicks are among the 1,000 visits, and convert at their own rate.

Why a rate on a handful of visits means little

An example. 20 people visit and 1 signs up: 5%. One more signup makes it 10%. One fewer makes it 0%. A single person moved the rate by 5 points. Now say 2,000 visit and 100 sign up. That is also 5%, and one person moves it by 0.05 of a point. The two percentages match, and the second is the one to trust.

The people who came this week are a sample of the people who could have come. The Office for National Statistics describes the same problem in its own surveys:

That page is about surveys, not websites, and the step from one to the other is this page's own. No fixed number of visits makes a rate safe. The more people behind it, the less one person can move it. Until then, write the two counts, "1 of 20", and not the percentage.

The mistakes a new founder makes with it

• One rate for everything. A visitor from an advert and a visitor from a search are not the same people. Work the rate out for each source, or a good source hides behind a poor one.
• Comparing your rate with a stranger's. Their step, their source, their price and their product are not yours. The Google pages above define the rate and name no figure as good.
• Counting one person twice. Google Ads Help says its rate can go over 100% when more than one conversion is counted for an interaction.
• Chasing the rate and not the money. Making signup free and easy can lift the visit to signup rate and bring nobody who pays. People sign up but nobody pays is that case.

Where it matters

• Nobody signs up. That is the visit to signup rate. Nobody is signing up takes the causes in order.
• Signups that never pay. That is the signup to payment rate, and the price is one thing that moves it. Judge a price change on the money it brings in, not on the rate alone.
• Paid ads. On Google Ads or Meta Ads the rate turns a cost for each click into a cost for each signup. Say, as an example, a click costs $1 and 2 of every 100 clicks sign up: 100 clicks cost $100 and bring 2 signups, so a signup costs $50. Carry that through to payment and you have your customer acquisition cost.

How Sami uses it

Sami works from the numbers you give it and the records AskSami keeps. Signups, clicks and drop offs come from a small snippet in your product, and payments from your own payment webhook, each tied to its source where the snippet recorded one. A payment with no recorded source is kept as unknown, never guessed.

So when you ask about a step, Sami answers from those counts. Where it has no data for a step it says so, and it never invents a figure.