AI and exit value · Answered by Golden Section from more than 400 B2B software companies observed
Two numbers, added together. The first is the floor: what the customers you already have are worth if you stopped selling to new ones tomorrow, with revenue compounding at your net revenue retention, cash flow at the margin you earn without new-customer spend, and both discounted at today's price of money. The second is the growth premium: what your new-customer engine adds after paying for itself. At a 5.2% 10-year Treasury and a 14.2% discount rate, a company whose 110% NRR settles toward 102% over four years, harvested at a 30% margin, has a floor of about 3.4× revenue. Growing 35% and fading, the same company is worth about 4.9×; accelerating from 30%, about 7.1×. For comparison, our base case for $15–20M vertical software exits is 4.0× revenue, and 5.5–5.8× with genuine vertical focus. Next step: put your own retention, margin and sales efficiency into the valuation model and see which of the two numbers you are really building.
Your company is worth its floor plus whatever growth premium survives the cost of buying the growth. Persistence of retention sets the floor. Sales efficiency decides whether growth adds to it or subtracts from it.
| Metric | Value | What it means | Source |
|---|---|---|---|
| Floor, installed base only | ≈3.4× revenue | 110% NRR settling to 102% on a four-year half-life, 30% harvest margin, 14.2% discount rateHeld flat at 110% forever, the same book would be worth about 7.9× | IllustrativeGolden Section valuation model |
| Same company, growing 35% and fading | ≈4.9× | Growth half-life of three years, $1.20 of sales and marketing per $1 of new ARRA one-year half-life drops it to about 3.7× | IllustrativeGolden Section valuation model |
| Same company, accelerating | ≈7.1× | Growing 30% and adding 5 points a year for two years before fadingThe shape of the curve is worth as much as its level | IllustrativeGolden Section valuation model |
| When growth destroys value | ≈2.7× S&M ÷ new ARR | Past this sales efficiency, with the inputs above, each dollar of growth lowers what the company is worth | IllustrativeGolden Section valuation model |
| Top-tier harvest margins | 45–50% | Sustainable where gross retention is at or near 95% and gross margin near 80%At 45%, the floor above rises to about 5.1× | Golden Section, publishedInvesting in Software addendum |
| Exit multiple base case | 4.0×; 5.5–5.8× | Enterprise value ÷ revenue for a $15–20M revenue vertical software company, the higher figure with genuine vertical focus | Golden Section, publishedInvesting in Software addendum |
| Price of money | 5.17% | 10-year Treasury on September 25, 2026Each additional 100 bp takes about 0.3× off the floor and 0.5× off the total above | External benchmarkFederal Reserve H.15 via FRED |
A software company is a book of contracts and a machine that sells more of them. The book is worth the cash it throws off. When your customers expand by more than they leave, revenue grows with no acquisition cost at all, which is why net revenue retention sets the floor. But no book expands forever: seats fill, modules get bought, price increases run out. How long your retention holds matters more than today's figure, and a buyer will read it from your cohorts, not your headline. That is what the ARR schedule is for, and why unproven customers belong in a provisional line outside the core book.
Growth is not free. Every point above what retention supplies is bought with sales and marketing. At $1.20 of spend per dollar of new ARR, 25 points of new-customer growth consumes 30 points of margin. Growth adds value only when the revenue it buys outlasts what it cost, which is the whole job of the sales efficiency ratio and unit economics plays. Past a certain efficiency the premium turns negative, and the fastest way to raise the value of the company is to sell less, better.
Then the shape. A company growing 35% whose growth halves every year is worth about half of one growing 30% and accelerating. Buyers pay for the trajectory they can believe, and unpredictable growth costs twice: it pulls the median outcome down and it raises the return a buyer asks for. The price of money sits over all of it. Because a growing company's cash arrives late, higher rates cut its value faster than they cut the floor's.
Two vertical software companies each reach $8M in revenue, both growing 30%. The first has 112% NRR settling toward 103% and spends $0.90 in sales and marketing for each dollar of new ARR. The second has 97% NRR settling toward 95% and spends $1.80. The first has a floor of about 3.9× and is worth about 5.0× in total. The second has a floor of about 1.6× and is worth about 1.2×, less than its floor, because each dollar of growth it buys costs more than it returns. Same growth rate, four times the value. All figures are invented for illustration.
Intrinsic value is not a price. A strategic acquirer filling a gap in its product or customer base may pay well above what the cash flows justify, and small transactions clear at lower multiples than the arithmetic suggests. The model also leaves out taxes, working capital, debt, dilution and deal terms, each of which changes what you personally take home. Use it to know what the business is worth to you before someone tells you what it is worth to them.
From the Golden Section mistakes list, each paired with the play that prevents it.
A plan that assumes today's growth rate holds is valuing a curve that fades.
The floor rests on expansion someone has to earn every year.
One weak cohort inside a blended NRR lowers the floor for the whole book.
Past a certain sales efficiency, more growth lowers what the company is worth.
Value comes from cash flow that stays, not from the forecast in the deck.
In the order we would run them. Each is on its own page, most with a free Excel template.
Produces the cohort retention figures that set your floor.
Keeps unproven revenue out of the retention number your value rests on.
Earns the expansion that keeps NRR from settling early.
Tells you whether the next dollar of growth adds value or subtracts it.
Shows by cohort that the revenue you buy outlasts what it cost.
Names the buyer and benchmarks you against what that buyer underwrites.
Run your own numbers in the valuation model Your retention, harvest margin, growth curve and sales efficiency through the same arithmetic, at today's 10-year yield, with the floor and the growth premium side by side.
As the present value of its future cash flow, which splits cleanly into two parts: the installed base harvested with no new-customer spend, and the new-customer engine net of what it costs. Revenue multiples are a shorthand for that arithmetic, not a substitute for it.
With 110% NRR that settles over a few years, a 30% harvest margin and today's rates, the floor alone is roughly 3–4× revenue, and moderate efficient growth takes it toward 5×. Our published base case for $15–20M vertical software exits is 4.0×, and 5.5–5.8× with genuine vertical focus.
Less than founders expect on its own, and more than they expect through its shape and its cost. Five more points of growth add about 0.4× in the base case; the same growth accelerating rather than fading adds more than 2×; and growth bought at poor sales efficiency can subtract value.
Higher rates lower every software valuation and lower growth companies most, because their cash arrives later. In the base case, 100 bp on the 10-year takes about 0.3× off the floor and 0.5× off the total.
Only if it lasts. The same 110% is worth about 7.9× if it holds indefinitely and about 3.4× if it settles toward 102% over four years. Show a buyer cohort data that proves persistence, and the floor holds.
We price our own investments against the same two numbers, and hold entry to reasonable multiples of revenue so a founder is not asked to give away the growth. A founder who wants to build value without outside capital can use the same plays.
Growth equity →Reviewed by Dougal Cameron, CEO & Co-Founder. Golden Section observations are labeled separately from external benchmarks and illustrative arithmetic.