The calculation
A conversion rate is the percentage of people who complete a specific action out of everyone who had the chance to. If 500 people visit a signup page and 25 sign up, the conversion rate is 5%.
conversion rate = (people who did the thing / people who could have) x 100 25 / 500 x 100 = 5%The arithmetic is trivial. Everything difficult is in the two definitions: which action, and which people. Change either and the number changes without anything real changing, which is how the same product ends up described as converting at 2% and at 40%.
Defining the two halves
| Denominator | Action | Typical rate |
|---|---|---|
| Everyone who visited the site | Made a purchase | 1–3% |
| Visited the pricing page | Started a trial | 5–15% |
| Started a trial | Became a paying customer | 10–25% |
| Booked a demo | Became a paying customer | 20–40% |
Write both halves down explicitly next to every rate you track. "5% of pricing-page visitors start a trial" is a metric. "Our conversion rate is 5%" is a number waiting to be misunderstood.
Reading a funnel
One rate hides where the problem is. A funnel splits the journey into steps so the worst step is visible.
Visited site 10,000Viewed pricing 2,000 (20% of visitors)Started signup 400 (20% of pricing views)Completed signup 120 (30% of starts) <- the leakUsed it once 96 (80% of completions)Seventy per cent of people who begin signing up abandon partway. That's the bottleneck, and it's usually mundane: too many fields, a card required before any value is shown, or an error message that doesn't say what went wrong. Improving the top of that funnel sends more people into the same leak.
How much data you need
Small numbers produce large, meaningless swings. With forty visitors and two conversions, one additional signup moves the rate by 2.5 percentage points — and that movement is noise.
- Under a hundred people through a step, treat the rate as an anecdote
- Compare like periods — a weekday against a weekend tells you about days, not about your page
- Change one thing at a time, or you won't know which one did it
- When two variants are close, they're the same; only act on differences large enough to survive noise
At low traffic, watching people use the page teaches you more than the rate does. Five recorded sessions or five observed users will find the broken step that a percentage only tells you exists.
Common mistakes
- Quoting a rate without saying what it's a rate of
- Comparing to a benchmark with a different denominator
- Optimising the top of the funnel while the leak is at the bottom
- Declaring a winner from twenty conversions
- Counting a signup as success when the product's success is a second visit
Key takeaways
- Conversion rate is completions divided by opportunities — the definitions carry all the meaning
- Always state both halves alongside the number
- Funnels reveal the worst step; fix that one first
- Under a hundred observations, the number is an anecdote
Try it yourself
Write your funnel out as five steps with real numbers from the past month. Find the step with the largest drop. That step is your next piece of work, and it's probably not the one you assumed.