how to calculate funnel conversion rate
The short version
- Overall funnel rate: (people who completed the last step ÷ people who entered) × 100
- One step's rate: (people who reached that step ÷ people who reached the step before it) × 100, never divided by the total number of entrants
- Mixing those two formulas up is the single most common calculation mistake, and it makes every step after the first one look worse than it is
- Someone who arrived mid-funnel and gets counted as if they walked the whole path inflates step one and quietly wrongs every rate under it
- A rate built from a small sample moves on its own. Getting the arithmetic right does not make a noisy number meaningful
A funnel conversion rate is the percentage of people who took the next step, out of the people who could have. That's the whole definition. The part that trips people up is that "could have" means two different things depending on which rate you're calculating, and most guides to this quietly use both without saying so.
There are two formulas here, not one. Getting them right takes five minutes. Knowing why the number moved is a different job, and a longer one.
The two numbers people call "conversion rate"
The overall funnel rate answers "of everyone who entered, how many made it all the way through?" You divide by the number of people who entered the funnel, full stop.
A step's conversion rate answers a narrower question: "of the people who reached this step, how many went on to the next one?" You divide by the people who reached that step, not by everyone who ever entered the funnel.
Confuse the two and a five-step funnel makes every step after the first look artificially bad, because you're dividing a small number by a much bigger one that most of those people were never going to reach anyway. A step converting at 70% of the people who actually saw it can look like it's converting at 12% if you divide by total entrants instead. Neither number is wrong, exactly. They're answers to different questions, and only one of them tells you whether that specific step is working.
The formula for the whole funnel
Overall conversion rate = (people who completed the final step ÷ people who entered) × 100
Say a software trial funnel looks like this over a month:
| Step | People who reached it |
|---|---|
Landed on /pricing |
4,000 |
Started /signup |
900 |
| Activated (used the product once) | 240 |
Overall rate: 240 ÷ 4,000 × 100 = 6%. Six in every hundred people who saw the pricing page ended up as an activated trial. That's the number you'd quote if someone asked "how's the trial funnel doing overall".
The formula for one step
Step conversion rate = (people who reached this step ÷ people who reached the step before it) × 100
Using the same funnel:
| Step | Formula | Rate |
|---|---|---|
| Pricing → Signup | 900 ÷ 4,000 × 100 | 22.5% |
| Signup → Activated | 240 ÷ 900 × 100 | 26.7% |
Notice the second step's rate (26.7%) is higher than the overall rate (6%), even though it comes later in the funnel. That's not a contradiction. The overall rate is diluted by everyone lost earlier; the step rate only counts the people who were still in the running when that step happened. Both numbers are correct. They just answer different questions, which is the entire reason to keep the two formulas separate rather than reaching for whichever one is easiest to compute.
The mistake that quietly wrecks the number: mid-funnel entrants
The formulas above assume every person in "people who reached this step" actually walked the path from the step before it. In practice, that's often not true.
Someone lands on /signup directly from a shared link, skipping /pricing entirely. Someone comes back a week later on their phone and picks up mid-flow. If your tool counts these people as if they entered at step one, your top-of-funnel number is inflated, and because every step rate downstream divides by that inflated number, every single percentage in the report comes out lower than it should.
This is arithmetic, not diagnosis: it happens whether or not anything is wrong with your funnel. A tool that doesn't separate "reached this step having started at the top" from "reached this step, full stop" is handing you a number that answers a question you didn't ask. Tiny Funnel excludes mid-funnel entrants by default when it calculates a step's rate, with a toggle to bring them back for a different question ("how many people did this step regardless of where they came in"). Whatever you're using, find out which one it does before you trust the percentage in front of you, because it changes every number under it.
A quick-reference table
For a funnel with entrants at the top and completions at each stage, here's the arithmetic laid out so you can run it against your own numbers:
| You want | Divide | By | Multiply by |
|---|---|---|---|
| Overall funnel rate | Completed the last step | Entered the funnel | 100 |
| One step's rate | Reached this step | Reached the step immediately before it | 100 |
| A rate excluding mid-funnel arrivals | Reached this step, entered at the top | Entered the funnel at the top | 100 |
If you're doing this by hand from raw event data rather than a tool built to draw the funnel for you, the step immediately before is the number to hold onto. It's the one people forget, reaching for total entrants because it's the number they already have open.
Why the same formula can still mislead you
Getting the arithmetic right doesn't make a small number meaningful. If nine people reached your checkout step this week and three finished, your rate is 33.3%. If a tenth person had finished, it would jump to 40%. Nothing about your product changed between those two worlds; one person's decision moved the percentage by nearly seven points.
That's not a flaw in either formula above. It's what happens when you divide by a small denominator, and no amount of correct arithmetic fixes it. Avinash Kaushik's long-standing case for computing statistical significance before acting on a conversion number makes the same point about web analytics generally: a rate is an estimate, not a fact, until the sample behind it is big enough to trust. The practical fix here is simpler than running a significance test: widen the date range before you widen the funnel, and don't read meaning into a rate built from a couple of dozen people.
Once you've got the number right and you're confident it's built from enough people to mean something, the next question is why it moved, which is a different job from calculating it. How to read a funnel drop-off covers that: how to tell whether a fall is the page or the traffic, and when a number that looks bad is actually just noise.
Frequently asked questions
What counts as a "conversion" when calculating a funnel rate?
Whatever the last step in the funnel you've defined is. It doesn't have to be a purchase: it can be a completed signup, an activated trial, or a course enrolment. The formula doesn't care what the step means, only that you've decided in advance what the funnel's endpoint is.
Do I divide by total entrants, or by the step before it?
Depends which question you're asking. Total entrants gives you the overall funnel rate. The step immediately before gives you that one step's own conversion rate. Dividing a single step by total entrants is the mistake that makes every step after the first look worse than it actually is.
How does a mid-funnel entrant change the calculation?
It inflates whichever step they're first counted at, which lowers every rate calculated downstream of that step, because those rates all divide by a number that's now bigger than it should be. Check whether your tool counts arrivals from mid-funnel by default before you trust a step rate.
What's a good funnel conversion rate?
There isn't a universal one, and a number from someone else's funnel with different steps, traffic and pricing can't tell you whether yours is healthy. How to read a funnel drop-off has the fuller answer on benchmarks, and the short version is: compare your own funnel to itself last month.
Do I need an analytics tool to calculate this, or can I do it by hand?
The arithmetic is simple enough to do by hand once you have the raw counts. What a tool actually saves you is getting those counts right in the first place: separating mid-funnel entrants from the people who walked the whole path, matching a returning visitor across devices, and stitching together a backend event like a completed payment with the browser session that started it. Recording the steps by clicking through your own site is how those counts get produced without writing event-tracking code first.
Two formulas, one habit
Overall rate divides by everyone who entered. A step's rate divides by the people who reached it. Keep those separate, take mid-funnel entrants out before you trust a step number, and don't read a verdict into a rate built from a dozen people.
That's the whole job of getting the number right. Working out what to do about it is a separate, longer conversation, and it's the one a funnel analytics tool that opens on the answer is actually for: the steps recorded by clicking through your own flow, the rate calculated correctly by default, and the next question, whether it's the page or the traffic, one filter away.