Funnel math turns a revenue target into the number of wins, opportunities, qualified leads, leads and visits that target implies. You divide the target by the value of one won customer to get the wins you need. Then you work backward one stage at a time, dividing each required count by the conversion rate into that stage. Multiply forward at the end to check the result.
The whole method fits on one line:
Required count at a stage = required count at the next stage ÷ the conversion rate between them, written as a decimal.
The answer is a planning scenario. It tells you what has to be true if your assumptions hold. It is not a forecast, and it carries no measured certainty, because every rate in it is an assumption until you replace it with your own records.
One label marks evidence boundaries on this page. Recommendation means a rounding rule, workflow or check that CoreAEX prescribes and no source specifies. Every number in the worked example is hypothetical. Untagged text is arithmetic, ordinary explanation, or a conclusion following from something already labelled.
Name the target and the stages first
Start with a target whose unit is named. "Grow revenue" cannot be divided. "$600,000 in new annual recurring revenue (ARR) next year" can. New ARR, total revenue, bookings and recognised revenue are different measures, so pick one and keep it for the whole calculation.
Next, define the value of one won customer in the same unit. If the target is new ARR, the divisor is the expected first-year ARR per won customer, not lifetime value and not total contract value.
Then list the stages your team actually records. A contact, an account, a lead, a marketing-qualified lead (MQL), a sales-qualified lead (SQL), a qualified opportunity and a closed-won deal are different records or decisions. Some are people, some are companies, and some are judgements a team makes about a record. The model only works if each stage in it matches a stage someone can count in your CRM. This site's definitions of MQL, SQL and win rate are a reasonable starting vocabulary.
Your stages may not be leads at all. Forrester, an analyst firm, describes its B2B Revenue Waterfall as a move away from lead-centric demand processes toward opportunities with connected buying groups. If your team counts buying groups or opportunities rather than individual leads, the arithmetic does not change. You divide through the stages you record instead of the ones in this example.
A worked example, backward and forward
The company, target and rates below are hypothetical. They show the mechanics, not a benchmark, and no rate here should be read as typical for B2B SaaS.
Inputs: a target of $600,000 in new ARR; $30,000 of first-year ARR per won customer; opportunity-to-win 25%; SQL-to-opportunity 40%; lead-to-SQL 20%; visit-to-lead 2%.
| Stage | Calculation | Result |
|---|---|---|
| Wins | $600,000 new ARR ÷ $30,000 per won customer | 20 won customers |
| Qualified opportunities | 20 wins ÷ 0.25 | 80 opportunities |
| SQLs | 80 opportunities ÷ 0.40 | 200 SQLs |
| Leads | 200 SQLs ÷ 0.20 | 1,000 leads |
| Relevant visits | 1,000 leads ÷ 0.02 | 50,000 visits |
Now check it forward. 50,000 visits × 0.02 = 1,000 leads. 1,000 × 0.20 = 200 SQLs. 200 × 0.40 = 80 opportunities. 80 × 0.25 = 20 wins. 20 × $30,000 = $600,000. The forward pass lands on the target, so the backward pass has no arithmetic error.
Recommendation When a division does not come out even, round the required count up to the next whole number at every stage. You cannot win two-thirds of a customer, and rounding down would leave the plan just short of the target. Every division in this example comes out even, so the rule changes nothing here. It matters in the range below.
Pipeline is not revenue
The example contains a second number that is easy to misread. Eighty opportunities at $30,000 each add up to $2.4 million. That is unweighted potential contract value, and it only holds under the example's assumption that every deal is the same size. It is not revenue, and it is not a measured pipeline value.
Apply the 25% win rate and the expected value is $600,000: the target. An opportunity is a deal that may still be won or lost, so a pipeline figure is an expectation that carries a probability. Reporting $2.4 million as though it were money coming in would overstate the plan by a factor of four.
Change one rate, then show a range
Change only one assumption and watch what moves. If visit-to-lead falls from 2% to 1%, the required visits become 1,000 ÷ 0.01 = 100,000. Wins, opportunities, SQLs and leads stay exactly where they were, because they sit downstream of that rate. Halving the top-of-funnel rate doubles the traffic requirement and changes nothing else.
Real plans rarely move one rate at a time. When several rates change together, recalculate the whole scenario from the target down, because the changes multiply. The table below keeps the target and deal value fixed and moves all four rates a little in each direction. The scenarios are hypothetical, and counts are rounded up at each stage.
| Scenario | Win | SQL to opp. | Lead to SQL | Visit to lead | Opportunities | SQLs | Leads | Visits |
|---|---|---|---|---|---|---|---|---|
| Conservative | 20% | 30% | 15% | 1% | 100 | 334 | 2,227 | 222,700 |
| Base | 25% | 40% | 20% | 2% | 80 | 200 | 1,000 | 50,000 |
| Optimistic | 30% | 50% | 25% | 3% | 67 | 134 | 536 | 17,867 |
Moderate changes to each rate move the visit requirement from under 18,000 to over 220,000, roughly a twelve-fold spread. Rounding up at each stage adds a little at the top: unrounded, the two ends are about 17,778 and 222,222.
Recommendation Present the output as a range with the assumption behind each end, not as one figure. A single target looks decisive, but it hides which rate the plan depends on most. A range shows the team which assumption to test first.
Mistakes that quietly break the model
Most funnel-math errors are not arithmetic errors. They come from feeding the calculation numbers that do not mean what the model assumes.
- Percent instead of decimal. Dividing 1,000 leads by 2 instead of 0.02 gives 500 visits, not 50,000. Convert every rate to a decimal before you divide.
- Visits treated as people. Analytics tools usually count sessions or visits. One person can visit many times, so a visit-to-lead rate is not a person-to-lead rate. Use the same unit on both sides of the rate.
- Rates from different cohorts. A win rate from last year's enterprise deals and a lead-to-SQL rate from this quarter's self-serve signups describe different populations. Chained together, they describe nobody.
- Pipeline reported as revenue. Unweighted opportunity value is not revenue, as the example shows. Name the measure every time a value appears.
- Every visit treated as qualified. A visit-to-lead rate measured on relevant traffic does not apply to all traffic. If the rate came from visitors in your market, the 50,000 visits are relevant visits, not any visits.
- An all-channel rate used as an SEO target. If the rates were measured across every channel, 50,000 is an all-channel requirement. It is not an organic-search traffic target.
- Win rates that ignore the sales cycle. Open deals have not reached an outcome yet. Counting them as losses makes a recent win rate look worse than it will turn out to be.
Recommendation Before you multiply anything, check that every rate in the chain shares one cohort, time window, market, channel scope and set of stage definitions. If one rate fails that check, mark it as an assumption in the plan rather than presenting it as measured.
Turn the output into a channel plan
The base case says the plan needs 1,000 leads. It does not say where they come from. A channel plan splits that requirement across sources, and each source gets its own rates.
Suppose a team expects 400 of the 1,000 leads from organic search, 350 from paid, and 250 from events and partners. Each channel then needs its own backward calculation, using that channel's visit-to-lead rate. If organic search converts visits to leads at a different rate from paid traffic, the two traffic requirements will differ even for the same lead count. The split and the rates here are hypothetical; the point is that each channel inherits only its share of the target.
Recommendation Write each channel's share, rate and source on one line of the plan, so anyone reading it can see which numbers are measured and which are hopes. Then add the channel requirements back up and confirm they still reach the lead total.
For the organic share, you also need to know which searches could bring the right visitors. Competitor Page Scout, a free CoreAEX tool, can suggest relevant competitor pages and keyword opportunities for your market, and the Scout guide explains how to read its report. Scout does not calculate your conversion rates, the traffic a target needs, or expected pipeline. Its search volumes are third-party estimates of monthly searches, not a promise of visits, leads or customers. Use it to find where the organic share might come from, then run that share through your own rates.
Replace assumptions with your own cohort data
The example's rates are placeholders. The model becomes useful once each one is replaced with a rate you measured.
- Pick one cohort. For example, every lead created in one quarter, in one market, from one channel or from all channels together. Keep that choice for every rate.
- Wait for it to mature. Give the cohort long enough to pass through your typical sales cycle before you measure the later stages. A cohort measured too early still has open opportunities in it.
- Count each stage in the cohort. Leads, SQLs, opportunities, wins. Each rate is the later count divided by the earlier one, within that cohort.
- Calculate win rate among closed outcomes only. Closed-won divided by closed-won plus closed-lost, with open deals left out. The measurement page sets out this rule and the conversion-lag adjustment that goes with it.
- Keep the counts beside the rates. A 25% win rate from 8 closed deals and a 25% win rate from 400 are not equally reliable. Show both numbers.
Once the plan runs, the next question is whether its assumptions match what actually happened. That runs in the other direction in time. To test your assumptions against deals you have already won and lost, trace your past deals backward with reverse funnel marketing.
Next steps and limitations
To run this on your own numbers:
- Write the target with its unit and period.
- Set the first-year value of one won customer in the same unit.
- List the stages your CRM records, in order.
- Divide backward, rounding each count up.
- Multiply forward to check.
- Build a conservative and an optimistic scenario next to the base case.
- Mark which rates are measured and which are assumed.
What this page does not claim: that any rate in the examples is typical for B2B SaaS; that the output is a forecast; that more traffic in one channel causes more customers in the same proportion; or that keyword volume converts into visits. The model assumes identical deal values and a steady conversion rate at each stage, and real funnels have neither. Treat the result as a plan to test, not a number to defend.
Sources. The buying-group framing is Forrester's own description of its B2B Revenue Waterfall (page modified August 3, 2026, reviewed November 6, 2026); it is an analyst framework for organising demand processes, not a study, and it is cited here only to note that some teams count opportunities and buying groups rather than leads. The definitions of MQL and SQL and the closed-outcome win-rate rule are owned on this site by Measuring Content-Market Fit. Statements about Competitor Page Scout were checked against the live tool and its guide on November 6, 2026. All inputs, outputs and scenarios in the worked examples are hypothetical, and the arithmetic is CoreAEX's own.
About the author
Zarko Zivkovic is the founder of CoreAEX, building technical SEO, AEO, and AI-visibility systems for B2B SaaS companies. Connect on LinkedIn.