Conversion rates multiply, they do not add. A funnel with five sequential stages, each improved 20%, converts 1.25 = 2.49× better, not 100% better. On a 1,000-lead funnel closing 12.6 deals a month, that is 31.4 deals. The same arithmetic runs backwards: five stages each drifting 10% worse leaves you 59% of the deals.
- Stages multiply. End-to-end conversion is the product of every stage rate — contact × set × show × proposal × close — not their average.
- Five stages, +20% each = 2.49×. Adding the gains would tell you 2.00×. The 24.4% difference is the compounding.
- A ~3× end-to-end result needs about 25% at five stages, not 300% at any one of them.
- Levers that touch the same stage do not multiply. 3× × 2× × 2× = 12× is almost always a double count — including when we are the ones quoting the parts.
- Relative change compounds; percentage points do not. A close rate moving 20% → 25% is +5 points and +25% relative; only the 25% belongs in the chain.
- Decay compounds too. Five stages each 13% worse halves your deal count with nothing visibly broken.
Do conversion rates multiply or add?
They multiply, because the stages are sequential: every lead that fails a stage is gone from the denominator of the next one. If 40% of leads are contacted, 25% of those book an appointment, 70% of those show, 60% of those receive a proposal and 30% of those sign, then the end-to-end rate is 0.40 × 0.25 × 0.70 × 0.60 × 0.30 = 0.0126, or 1.26%. Out of 1,000 leads, 12.6 closed deals.
Improve all five stages by 20% each and the naive answer — add the gains, 5 × 20% = 100%, so twice the deals — is wrong by a quarter. The right answer is 1.2 × 1.2 × 1.2 × 1.2 × 1.2 = 2.4883, a 149% increase. A sales funnel is a chain of multiplications, so improvements to it are exponents, not sums. One unit warning: “close rate went from 20% to 25%” is +5 percentage points and +25% relative, and only the relative figure (1.25) can enter the chain.
How it works
How to compound conversion gains without double-counting them
List every stage
Write out each sequential step from lead to closed deal, with its own numerator and denominator. Five stages is typical: contact, set, show, proposal, close.
Multiply, never add
Multiply the stage rates together to get your end-to-end rate. That product is the only number an improvement can be measured against.
Ledger the levers
Convert each proposed lever into the stage rates it actually moves. Two levers touching the same stage are recorded once in the ledger, not multiplied together.
Run the ceiling test
Re-read every stage rate in the ledger. If any has been pushed above 100%, the stack is double-counted and the headline multiple is wrong.
MAKE MORE SALES.
Pay-Per-Result pricing — We scale sales HARD aligned to your interests, better than anyone else.
The worked calculation: 1,000 leads through five stages
Here it is end to end with substitutable inputs. The baseline rates are round illustrative numbers, not a benchmark — swap in your own and the arithmetic is unchanged.
| Stage | Baseline rate | Volume from 1,000 leads | Rate after +20% relative | Volume after |
|---|---|---|---|---|
| Leads in | — | 1,000 | — | 1,000 |
| Contact rate | 40% | 400 | 48% | 480 |
| Appointment set rate | 25% | 100 | 30% | 144 |
| Show rate | 70% | 70 | 84% | 121.0 |
| Proposal issued rate | 60% | 42 | 72% | 87.1 |
| Close rate | 30% | 12.6 deals | 36% | 31.4 deals |
| End-to-end | 1.26% | 12.6 | 3.14% | 31.4 |
Same leads, same ad spend, same market: 12.6 deals becomes 31.4 because five ordinary 20% improvements sat on top of each other. Nothing there is a claim about anyone’s business — it is 1.25 = 2.4883 applied to a funnel, reproducible in a spreadsheet in two minutes. No single stage in a five-stage funnel has to improve dramatically for the funnel to improve dramatically — that is the honest mechanism behind every large conversion number you have been quoted, and unlike most of them it is checkable.
Want this done for you? We book qualified sales appointments on a Pay-Per-Result basis — you only pay for calls that actually land in your calendar.
How much better does every stage need to get to double, or triple?
If every stage improves by the same relative amount g across n stages, the end-to-end multiple is exactly (1 + g)n. For the five-stage funnel above:
| Uniform gain per stage | End-to-end multiple (5 stages) | What adding the gains would wrongly say | End-to-end rate from 1.26% |
|---|---|---|---|
| +5% | 1.28× | 1.25× | 1.61% |
| +10% | 1.61× | 1.50× | 2.03% |
| +15% | 2.01× | 1.75× | 2.53% |
| +20% | 2.49× | 2.00× | 3.14% |
| +25% | 3.05× | 2.25× | 3.85% |
| +30% | 3.71× | 2.50× | 4.68% |
| +50% | 7.59× (impossible here) | 3.50× | 9.57% |
Read the +25% row twice. Tripling end-to-end conversion across five stages requires roughly 24.6% at each stage (31/5 = 1.246) — contact rate 40% to 50%, show rate 70% to about 87%. Hard, but ordinary. Nobody has to be 300% better at anything. The +50% row is marked impossible on purpose, and that is the constraint most compounding arguments quietly ignore: it would need a show rate of 105%. Compounding is bounded by the ceiling of every stage it runs through, so the exponent stops working long before the spreadsheet does.
Why 3× speed to lead, 2× contact rate and 2× set rate do not make 12×
This is the section we would leave out if this page were marketing. In our own client work we typically see about a 3× lift in conversion when a business still running 2020-era sales operations — manual follow-up, business hours only, one or two touches — moves to 2026 AI-driven operations. Our component figures: fixing speed to lead on its own, around 3×; doubling contact rate, around 2×; doubling appointment set rate, around 2×. These are our own observations across the campaigns we run, not a study — no published sample, no window, and not a guarantee.
Multiply them and you get 3 × 2 × 2 = 12×. We do not claim 12×, and after the arithmetic below you should not accept 12× from anyone. Write the levers into stage rates instead, starting from contact 40% and set 25%:
| Lever, in order applied | Contact rate | Set rate | Running end-to-end multiple |
|---|---|---|---|
| Baseline (2020-era operations) | 40% | 25% | 1.00× |
| Speed to lead fixed (reply in minutes, around the clock) | 80% | 37.5% | 3.00× |
| “Double the contact rate” applied on top | 160% — impossible | — | — |
| Set-rate work that speed did not already do | 80% | 50% | 4.00× |
Row two is the whole explanation. “Speed to lead is worth ~3×” is already an end-to-end number, and on any honest reading of where that 3× comes from it is earned by doubling contact rate (40% → 80%, ×2.0) and lifting set rate by half (25% → 37.5%, ×1.5), because a lead reached while still on your website books more readily than one reached on Thursday. Applying “doubling contact rate, ~2×” on top asks contact rate to double a second time, to 160%, which does not exist. Row three fails, and that failure is the proof of the double count.
The set-rate lever survives only partially. Speed already carried set rate to 37.5%; the genuinely separate work — better qualification, offering a specific time on the call itself, a confirmation sequence — carries it to 50%. That is an extra ×1.33, not an extra ×2. The honest stack is 4.00×, not 12×, and the figure we actually publish, ~3×, sits below even that, because stages run into ceilings and most clients get some of the levers rather than all of them. So, the rule, free to anyone reading including our competitors:
The stage-ledger rule: convert every lever claim into the stage rates it moves, write those rates into one ledger, then multiply the stages once. Never multiply the lever claims.
And its ten-second audit, the ceiling test: if stacking the claims pushes any stage rate above 100%, the stack is double-counted. Run it on every conversion number you are quoted this year. It fails a naive reading of our own numbers, which is exactly why we have written it down.
If we can’t make you money, we don’t deserve yours.
Pay-Per-Result pricing — performance-based alignment.
What the response-time research actually measures, and what it does not
Speed to lead is the one lever here with genuine independent research behind it, so be precise about what that research says — and keep it well away from our own operator figures above.
In The Short Life of Online Sales Leads (Harvard Business Review, March 2011), Oldroyd, McElheran and Elkington audited 2,241 US companies with a web-generated test lead: 23% never responded at all, and the average response time among those who replied within 30 days was 42 hours. A separate study of 1.25 million leads across 29 B2C and 13 B2B US firms found that firms contacting a prospect within an hour were nearly seven times as likely to qualify the lead as those contacting an hour later, and more than 60 times as likely as those waiting 24 hours or more. The earlier InsideSales.com/MIT Lead Response Management study (Oldroyd and Elkington, presented October 2007; three years of data, six companies, over 15,000 leads and 100,000 call attempts) reported that “the odds of contacting a lead if called in 5 minutes versus 30 minutes drop 100 times”, and the odds of qualifying drop 21 times.
Those are odds ratios on contact and qualification, not end-to-end conversion multipliers, and no honest funnel model puts “100×” into a stage. The 100× came from six companies in 2007; the 7× is a likelihood of having a meaningful conversation with a decision maker, not of banking revenue. What they legitimately support is the direction and rough size of the contact-rate effect — the ×2.0 in the ledger above is consistent with them, and it is all we take from them. Where the five-minute framing is right and where it is over-quoted is covered on our speed to lead and the five-minute rule guide.
Which stage has the most compounding headroom left?
A 20% relative gain is worth exactly the same end-to-end wherever in the chain you win it — multiplication cares about neither order nor stage size. What differs is how much room a stage has left: a stage at rate r can improve by at most 100/r before it hits 100%.
| Current stage rate | Maximum remaining gain | What that means |
|---|---|---|
| 15% | 6.67× (+567%) | Almost all the funnel’s headroom lives here |
| 25% | 4.00× (+300%) | Room for several rounds of improvement |
| 40% | 2.50× (+150%) | Room for two or three 25% rounds |
| 60% | 1.67× (+67%) | Two 25% rounds and you are done |
| 70% | 1.43× (+43%) | One 25% round and a bit |
| 85% | 1.18× (+18%) | Not worth a project; spend the effort elsewhere |
| 95% | 1.05× (+5%) | Finished. Any “improvement” here is measurement noise |
For the example funnel the arithmetic ceiling is 1 ÷ 0.0126 = 79.4×, the multiple if every stage hit 100%. Nobody gets there, but it explains why 3× is not a suspicious number: it is 3.8% of the way to a ceiling nobody is near. Compounding does not run out at the funnel level for a long time; it runs out one stage at a time, and the stage that runs out first is the one already performing best. If your show rate is 85% and your contact rate is 25%, stop working on show rate — and a 25% contact rate is usually a coverage problem, which our guide to what breaks in a sales funnel as volume grows takes apart.
Neglect compounds too: what a 10% drift at five stages costs
The exponent is symmetric, and this is the version nobody puts in a deck. Five stages each drifting 10% worse gives 0.95 = 0.590 — you keep 59% of your deals, 12.6 a month falling to 7.4. Five stages each 5% worse gives 0.955 = 0.774, a 23% fall; each about 13% worse halves the deal count outright (0.875 = 0.50).
A funnel does not need a disaster to halve; it needs five ordinary stages each getting about 13% worse, which no single dashboard will flag. That is the price of one quarter of slightly slower replies, slightly fewer follow-ups and slightly weaker confirmations, and it is why stage-level measurement beats watching one blended number. Show rate drifts unnoticed most often, because it degrades between the booking and the call rather than during either — covered separately in our guide to improving sales appointment show rates.
What it costs to run this arithmetic yourself every month
The maths is free; the inputs are not. You need every stage timestamped in one system with boundaries that mean the same thing in March as in September — the definitional work, not the calculation, is what takes days the first time. After that it is a couple of hours a month: pull five counts, deduplicate, compute five ratios and their product, compare against last month.
The real constraint is sample size, and it is arithmetic you can check. At a 40% contact rate, the standard error on a monthly estimate is √(0.4 × 0.6 / n). At 100 leads a month that is 4.9 percentage points — wider than the 8-point move a 20% improvement would produce, so you cannot tell the improvement from noise. You need about 600 leads a month to get that error under 2 points, and roughly double that to compare two months with confidence. Below that, use a rolling three-month window and review quarterly, or you will spend the year chasing randomness.
| Monthly lead volume | What you can actually measure | Sensible approach |
|---|---|---|
| Under 100 | Nothing month to month; 100 leads make ~4.2 proposals, so one extra deal moves close rate ~24 points | Count deals, not rates. Fix the obvious stage by eye |
| 100–600 | Large moves only (roughly 40%+ relative) | Rolling 3-month window, review quarterly |
| 600–2,000 | 20% relative moves per stage, monthly | Monthly stage ledger; one lever at a time |
| Over 2,000 | Sub-10% moves, and by segment | Segment before you average, or the blend hides both |
The other cost is coverage. Most of the contact-rate stage is decided outside business hours, so running the speed-to-lead lever by hand means answering enquiries across 168 hours a week rather than 40 — a staffing question, not an analytics one, and the specific thing our speed to lead automation and AI appointment setting work removes. Read our own commercial model through this page’s lens too: we are paid on booked qualified appointments rather than retainers or seats, which puts our incentive squarely on two stages of five — contact rate and set rate — and does nothing on its own for your show rate, proposal rate or close rate. Those three remaining stages are yours, and on the arithmetic above they carry most of the multiple.
The same discipline applies to numbers we publish about ourselves. Our methodology page defines the 7× average sales lift as trailing three-month closed-deal revenue at month six over the three months before launch, and discloses on the same page that the median is closer to 4×, because a few high performers lift the average. An average and a median that far apart is information, not a footnote.
This page is the arithmetic layer under a wider set of pipeline-stage guides — one per stage from contact rate through to close rate, plus a pipeline-stages hub setting out what each stage is and what it costs when it leaks. Those pages give you a stage rate; this one tells you what improving it is worth.
Frequently asked questions about compounding conversion math
Do conversion rates multiply or add?
They multiply. Sequential funnel stages each operate on the survivors of the previous one, so the end-to-end rate is the product of the stage rates: 40% × 25% × 70% × 60% × 30% = 1.26%. Improving five stages by 20% each gives 1.2^5 = 2.49×, not the 2.00× you get by adding the gains.
Do small improvements really compound into large ones?
Yes, and the exponent is where it comes from. Across five stages, +10% each is 1.61×, +20% each is 2.49× and +25% each is 3.05×. Tripling end-to-end conversion needs roughly 24.6% per stage (3^(1/5) = 1.246), which is why a 3× result rarely requires anything spectacular at any single stage.
Why do the component numbers agencies quote never multiply to their headline?
Because the levers overlap. Fixing speed to lead is how contact rate improves, and contact rate is part of how set rate improves, so multiplying “3× from speed” by “2× from contact rate” counts the same doubling twice — it would put a 40% contact rate at 160%. Use the ceiling test: if stacking the claims pushes any stage above 100%, the stack is double-counted. It fails a naive reading of our own component figures, which is why we publish ~3× rather than 3 × 2 × 2 = 12×.
Should I compound percentage points or percentage change?
Percentage change. A close rate moving from 20% to 25% is +5 percentage points and +25% relative; only 1.25 goes into the multiplication. Compounding percentage points is meaningless because points from different stages are not the same unit.
How much does responding faster really move the contact-rate stage?
The best independent evidence is response-time research, and it measures odds rather than revenue. Harvard Business Review’s 2011 audit of 2,241 US companies found 23% never responded to a web lead at all and the average response time was 42 hours, while a companion study of 1.25 million leads found firms replying within an hour were nearly seven times as likely to qualify a lead as those replying an hour later. The 2007 InsideSales.com/MIT Lead Response Management study reported the odds of contacting a lead drop 100 times between a 5-minute and a 30-minute callback. Treat those as evidence that the contact-rate stage is highly speed-sensitive — not as a multiplier you can drop into a funnel model.
How many leads do I need before these stage rates mean anything?
About 600 a month to detect a 20% relative move in a single stage, because the standard error on a 40% rate at n = 600 is 2.0 percentage points against an 8-point signal. Under 100 leads a month the same funnel produces only about 4.2 proposals, so a single extra deal moves your close rate by roughly 24 percentage points and monthly rates are noise — count deals and use a rolling three-month window instead.
Pay-Per-Result appointments
See if we’re a fit
We book qualified sales appointments for you and you pay on results, not retainers. Our booking page asks a few quick questions so you find out in two minutes whether that model suits your business.
- 50,769+ appointments booked without cold calling.
- Pay-Per-Result pricing — you pay for booked, qualified calls.
- Pick your own time on our live calendar, no phone tag.