Marketing Funnel Calculator

What this tool does

CM8-373 models the funnel from first contact to customer. It works out the conversion rate into every stage, shows where people are lost, compares each rate against what you expected, and works backwards from a customer target to the traffic and leads it would actually require.

Everything runs inside this single file — no account, no upload, no network request of any kind.

What one row is

One row is one stage, in one segment, in one period. Stages are numbered one to six and the tool reads them in that order, so the names are yours entirely — "visitors" and "reach" and "impressions" all work as stage one.

The tool derives each stage's rate from the highest-numbered stage below it in the same segment and period. That means you can skip stages: a funnel of visitors, enquiries and customers is stages one, three and six, and the rates still work.

Choosing the stages

Four to six stages is right for almost everybody. Fewer and you cannot see where the problem is; more and you are measuring steps nobody can influence separately.

The test for whether a stage earns its place: could somebody own it, and could they change it? "Visitors who scrolled past the fold" is a measurement, not a stage. "Enquiries" is a stage, because a person can be made responsible for the number of them.

Define each stage so it cannot drift

Write the definition in the notes and keep it there. This matters more than the arithmetic, because the most common failure of a funnel is not a bad rate — it is a stage that quietly changes meaning.

"Qualified" is the usual offender. If it means "a salesperson liked the look of it", the rate into it will move whenever the sales team is busy, and no amount of analysis will find out why. If it means "the lead answered the budget and timing questions", it is a fact and the rate means something.

Rates, not counts

Counts tell you how big the funnel is; rates tell you how well it works. Only rates can be improved, and only rates can be compared between periods when volume is moving.

Record what you expected each rate to be, from your own history or a comparable funnel. A rate on its own is almost impossible to judge — 7.5% of readers becoming enquiries could be excellent or terrible — and the comparison against your own expectation is far more useful than any published benchmark.

Finding the stage worth fixing

Three different answers, and they are usually different stages:

• The lowest rate — almost always the first step, because that is where the least interested people are. Rarely the right target.
• The biggest count lost — also usually the first step, for the same reason.
• The rate furthest below what you expected — this is the one to work on. It is the stage that used to work better, or works worse than a comparable funnel, and it is the one where something is actually broken.

Working backwards from a target

The most useful output of this tool, because it converts a target somebody set into a traffic number somebody has to deliver. Needing forty percent more customers means needing forty percent more of everything, and that is usually the moment a target gets renegotiated.

The last column shows the alternative: what one stage's rate would have to reach if nothing else changed. Where that comes out above 100% the tool says so, because the target cannot be hit through that stage alone.

Both assume rates hold as volume rises. They will not — extra traffic is less interested than the traffic you already have — so treat the traffic figure as a floor.

Where the leverage is

A funnel multiplies, so an improvement anywhere flows all the way through. A tenth more at any single stage is a tenth more customers, whether it happens at the top or the bottom.

That makes late stages far better value than they look. Improving a 34% close rate to 37% is a small change to one number and the same effect as a tenth more traffic — which usually costs money, takes months, and dilutes quality.

The tile prices this: what a tenth better at your worst step would be worth in customers, and in gross profit if you set a value per customer.

Segment, or the average describes nobody

Paid and organic traffic behave completely differently, and a blended funnel averages them into a shape that describes neither. In the sample, organic enquiries qualify at more than twice the rate of paid ones — invisible in the blended view, and the single most important fact in the whole model.

Record the same stages for each segment and the comparison chart draws itself. Where a segment converts much better late in the funnel, the answer is usually to buy more of that traffic rather than to fix the funnel at all.

The timing problem

The customers counted this month came from traffic that arrived weeks or months ago. In a short cycle this hardly matters; in a long one it makes every rate wrong, in both directions — growing traffic makes conversion look worse than it is, and falling traffic makes it look better.

Two workarounds, both imperfect: use a period longer than your sales cycle, or offset the stages so each is counted in the period the people actually arrived. Whichever you choose, say so in the notes and keep doing it the same way.

What a funnel model is not

• It is not a cohort. It counts flows through stages in a period, not individuals followed through.
• It assumes people move in order. Real buyers skip stages, and some arrive at stage four having never appeared at stage one — a referral, an inbound call. Where that is common, the top-of-funnel rate is meaningless and the tool will flag counts that rise between stages.
• It has no view on quality. Doubling enquiries by removing every question from the form will improve one rate and destroy the next.
• It says nothing about value. Fewer, larger customers can be worth more than the target.

Printing and sharing

The Report tab prints the tiles, charts and both tables with a title block you fill in. The working-backwards table is the one for a planning meeting: it turns a customer target into the traffic somebody has to actually produce.

Saving your work

Stages are held in this browser, on this computer, and stay there between visits. Use the backup button to write a JSON file you control.

Accuracy & disclaimer

Every count here is one you entered, and the rates are arithmetic on them. The tool has no view on whether your stages are the right stages or whether a rate is good, and it cannot tell you that a stage definition changed halfway through the period.