Module 2 · Designing the Engine · Chapter 8 of 32
The Mathematical Model
The two halves of a recurring-revenue business obey different math: acquisition multiplies and retention compounds. Learn to run GRR, NRR, and LTV, build a Growth Formula for a motion, and stress-test it before you fund it.
9 min read
A VP once asked me to double next year’s new bookings by doubling the outbound team. It sounded like arithmetic: twice the reps, twice the sends, twice the pipeline, twice the ARR. I pulled the funnel into a sheet on the call and showed him the number barely moved. Doubling the sends dropped the reply rate, and a win is not one dial you turn: it is a dozen small conversions multiplied together, so shaving a few points off any of them costs far more than the drop looks like it should. On the same screen, the renewal book he had never asked me to model was compounding in the other direction, quietly, on its own math. That call is the reason this chapter exists.
The metrics spine in the previous chapter (/course/metrics-spine/) gave you the letters: VM for volume, CR for conversion, Δt for time. This chapter is the arithmetic that runs on them. The first thing to know, straight from Revenue Architecture, is that the two halves of the Bowtie do not use the same arithmetic at all.
Two engines, two kinds of math
The left side of the Bowtie, acquisition, is a product. You multiply every conversion rate in the funnel together, and the output is how many deals you win. The right side, retention and expansion, is a sum. You add up the ARR a customer produces year after year, and the output is what that customer is worth over its life. Revenue Architecture names them by their Greek operators: acquisition is a Pi, a product; retention and expansion is a Sigma, a sum.
Commits = Π CR(n) = CR1 × CR2 × … × CRn
Every stage of the funnel is a fraction. Multiply them all and you get the win count. Drop one term and the whole product drops with it.
LTV = Σ ARR(t) for t = 1 … Lifetime
Lifetime value is the ARR the customer pays each year, summed across the years they stay. Time, not frequency, is the driver.
That difference is not a curiosity. It decides which lever actually moves the business, and it explains why the same headcount cut can be smart on one side and ruinous on the other.
| Acquisition (Π, open loop) | Retention & Expansion (Σ, closed loop) | |
|---|---|---|
| The math | Multiply every conversion rate together | Sum the ARR across each year of the life |
| How it behaves | Small simultaneous drops cascade; the loop runs away | Small gains compound; the loop self-corrects |
| Measured in | Units: leads, opps, wins | Revenue: MRR, ARR, LTV |
| What drives it | Frequency: the number of actions | Time: the duration of the relationship |
| The goal | Go fast: compress the cycle to days | Go long: extend the lifetime to years |
Nonlinearity is the whole game
Multiplying is not adding, and the gap between them is where hypergrowth and the 2022 crash both came from. A linear relationship is y = x · n. A nonlinear one is y = x^n, and a small change in x produces a wildly disproportionate change in y.
y = x^n (compounding) ≠ y = x · n (adding)
A 15% edge applied once is 15%. Applied six times over, compounded, 1.15 to the sixth power is 2.3x, not 90% more. The exponent is where the growth lives.
Acquisition looks linear and is not. Wins equal qualified opportunities times win rate, which reads like one multiplication. But win rate is itself the product of every meeting’s conversion rate, so a twelve-step deal is opps × CR^12. Revenue Architecture works the example: 20 qualified opps where every meeting converts at 85% yields 20 × 0.85^12 ≈ 2.85 wins, and at a $40,000 ACV that is about $114,000. Coach each conversation five points higher, to 90%, and 20 × 0.90^12 ≈ 5.6 wins, about $226,000. A five-point lift per step nearly doubled the outcome. That is compound impact, and it is why coaching the conversation beats buying more leads.
The same exponent runs in reverse, which produces a counterintuitive result the book calls “less is more.” Because each extra meeting is another multiplication by a number below one, a shorter, sharper cycle can win more deals than a longer one: 20 × 0.9^11 ≈ 6.3 beats 20 × 0.9^12 ≈ 5.6. Adding a step to “be thorough” can cost you wins.
The closed loop compounds: GRR and NRR
The right side of the Bowtie is a closed loop. It feeds back on itself and either self-corrects or self-destructs, and its two dials are the retention metrics the metrics spine labeled CR7 and CR7 × CR8. Gross Revenue Retention measures how much of the existing base you keep, before any expansion. It cannot exceed 100%.
GRR = ( MRR_start − contraction − churn ) ÷ MRR_start
Take 100 customers at $1,000/mo ($100,000 MRR). Three cancel (−$3,000) and two downgrade by $500 (−$1,000). GRR = ($100,000 − $1,000 − $3,000) ÷ $100,000 = 96%. In the Data Model, GRR is CR7.
Net Revenue Retention adds expansion back in, so it can cross 100% and the book can grow with zero new logos.
NRR = ( MRR_start + expansion − contraction − churn ) ÷ MRR_start
Same base, and now 12 customers expand by $500 (+$6,000). NRR = ($100,000 + $6,000 − $1,000 − $3,000) ÷ $100,000 = 102%. In the Data Model, NRR is CR7 × CR8 (retention times expansion).
Move the three inputs yourself and watch where the line crosses 100%. Below it, new sales have to refill the leak before they add a dollar. Above it, the base grows on its own.
net retention
Above 100% the existing book grows on its own before a single new logo signs. Below 100% you are running up a down escalator: new sales have to refill the leak before they add anything.
net retention: 110.0%
Where the money actually is: LTV
Lifetime value is the sum from the additive side, and NRR is the rate it compounds at. That is the whole reason retention outweighs acquisition past scale: a rate above 100% turns a flat contract into a growing one, year over year, with no new selling.
LTV = Σ ARR(t) for t = 1 … Lifetime
LTV is the ARR summed over the years the customer stays. NRR sets the slope: the faster it compounds, the more it accelerates.
Run the same $100,000 first-year deal at two retention rates over five years and the divergence is the entire argument for funding the right side of the Bowtie (figures from Revenue Architecture’s worked example, illustrative):
| Year | NRR 110% | NRR 90% |
|---|---|---|
| 1 (land) | $100,000 | $100,000 |
| 2 | $110,000 | $90,000 |
| 3 | $121,000 | $81,000 |
| 4 | $133,100 | $72,900 |
| 5 | $146,410 | $65,610 |
| 5-year LTV | $610,510 | $409,510 |
Identical first year, $201,000 apart by year five, and the gap widens every year after because the curves are exponential, not straight. Note where the value sits: about 83% of the 110% deal’s lifetime arrives after year one, and only if impact keeps landing. The gong-quarter lesson from the factory chapter (/course/recurring-revenue-factory/) is this table in disguise. The scoreboard that counts only the first row cannot see the engine.
The Growth Formula
Put the multiplicative and additive halves together and you get the production capacity of a single motion, cause to effect. Revenue Architecture calls it the Growth Formula.
ARR = Visitors × Π CR[n] × List Price
Normalize price to list × (1 − discount) first, then multiply the funnel through. This is one motion's monthly ARR-generating capacity.
Normalized to a single win, the book’s low-touch inbound example reads cleanly: 253 visitors convert down through the funnel to 30 leads, then 5 opportunities, then 4 qualified, then one commit at a 26% win rate; a $24,000 list price less a 23% discount lands $18,408 in committed ARR. Unit economics is just the shorthand for that chain: 253 visitors buys one $18,408 customer, which at roughly 107% NRR is worth about $106,909 over five years. That one sentence is what you fund a motion against, and what you compare motions on.
Scenario analysis: stress the model before you spend
A Growth Formula is a model, which means you can stress-test it on a laptop before you commit a budget to it. Two tests earn their keep.
The second test is the downturn. Shave 5% off every conversion rate at once, a soft-market quarter where every step gets a little harder, and committed ARR per unit falls from $18,408 to $10,759. A 5% input change produced a 42% output drop: the open-loop volatility from earlier, now pointed at your number. The value of modeling it is that you see the cliff in a planning meeting rather than in the actuals.
Handing off
You now have the arithmetic of the factory. Acquisition multiplies, so coach the conversion rate and watch the whole product cascade in either direction. Retention compounds, so NRR above 100% is a second engine that runs while you sleep. And a Growth Formula lets you price a motion, project its LTV, and stress it against a discount change or a downturn before you fund a single hire.
The math tells you what the engine can do. It does not tell you how to run the tools, the methodologies, and the language so the whole organization operates one engine instead of a dozen non-interoperable ones. That is the operating model, and it opens Module 3. Continue to The Operating Model (/course/operating-model/).