AI and exit value · Answered by Golden Section from more than 400 B2B software companies observed
Quality of revenue is how much of your reported revenue a buyer can trust to recur, so improve it by making the recurring part larger, longer, more spread out and easier to prove. Separate subscription from services, implementation and AI inference revenue. Move unproven accounts into a provisional line outside core retention. Raise prices on the poorest-fit customers or let them go, and prove a price increase in run-rate. Lengthen terms and clean the assignment clauses in your largest contracts, and diversify away from any customer above about 10% of revenue. Then make all of it provable with an ARR schedule tied to a contract register and audited financials. Start at least six to eight quarters before a process, because buyers price what has held in trailing results. Next step: restate gross and net retention for your core book alone and look at it as a buyer would.
Start the structural work first, such as concentration and contract terms, because it takes years. Prove the operating work, such as retention and pricing, in at least two quarters of run-rate. Close the measurement gaps last, but never late.
| Metric | Value | What it means | Source |
|---|---|---|---|
| Concentration, exit-ready | largest customer ≤ ~10%, top five ≤ ~25% | share of revenue from the largest and the top five customersrule of thumb: a 10–20% discount once one customer clears 20–35% of revenue | Golden Section playbookPre-Sale Value Levers |
| Net revenue retention effect | above ~110% earns a premium; below 90% is discounted | commonly cited effect on the revenue multiplea rule of thumb, not a quote; buyers rebuild it from invoices cohort by cohort | Golden Section playbookPre-Sale Value Levers |
| Conditions for top-tier margins | gross retention ≥95% and gross margin ≥80% | conditions under which 45–50% steady-state cash-flow margins are sustainableeach point of gross retention below 95% costs 1.3–2.0 points of margin | Golden Section, publishedInvesting in Software addendum |
| Exit multiple base case | 4.0x revenue; 5.5–5.8x with genuine vertical focus | enterprise value ÷ revenue for a $15–20M revenue vertical software companyquality of revenue is much of what decides where in that range a company lands | Golden Section, publishedValuation model |
| Poor-fit customers | about 20% of any cohort | share of a customer cohort that fits poorlyraise their price or let them go | Golden Section operating viewGolden Section operating view |
| Underpricing signal | customer ROI above 10x | customer's monetized gain ÷ price paidan ROI that high says the price, not the product, is the weak point | Golden Section operating viewGolden Section operating view |
A buyer is purchasing contracts and the machine that produces more of them, and prices both by how far they can be trusted. Revenue that is contracted, recurring, spread across many customers and traceable to signed paper gets the multiple. One-time fees, unproven accounts and loosely contracted revenue get discounted, and a blended retention number gets the whole book priced at its weakest cohort. That is why the core and provisional split matters, and why it has to start well before a process: a revenue line created six months before a sale is evidence of nothing.
Buyers also read what kind of revenue it is. We now treat seat-priced contracts in headcount-exposed functions as carrying a discount to assumed retention, and AI inference revenue needs its own revenue and cost lines because buyers will test how scalable and sticky it is. Strategic acquirers add questions of fit: customer overlap, the category you hold, technical debt, key-person risk and whether contracts survive a change of control, which the contract register answers. Many $10–20M vertical companies are now bought as add-ons, priced off the platform's own math, so the quality of each dollar matters more than the headline. The meaningful exit plan names the buyer; the pre-sale value levers play moves the numbers.
The figures are invented. A company at $12M in annual revenue reports 104% net retention and plans to sell in two years. Split properly, the core book retains 112% while a cohort of accounts signed during a push into an adjacent segment retains 70%, and the largest customer is 18% of revenue. The company opens a provisional revenue line for that cohort, raises prices 8% on renewals for the poorest-fit fifth of the core book and loses a few of them, and signs its largest customer to a five-year term with a clean assignment clause. Eight quarters later the core book shows two years of retention above 110% and a price increase in run-rate, and the buyer's model starts from those numbers.
If one customer will still be above 20% at the target date, stop pretending otherwise in the plan; lengthen the contract, clean the assignment language and choose a buyer for whom that customer is an asset, often a strategic already selling to it. And a business that is deliberately services-heavy should present services margin plainly rather than dress it as recurring revenue.
From the Golden Section mistakes list, each paired with the play that prevents it.
Blending unproven customers into core revenue lets one weak cohort set the price of the whole book.
A buyer who has to argue about what counts as revenue lowers the price to cover the argument.
Keeping poor-fit customers at any cost drags down retention and margin in the numbers a buyer prices.
Pricing set without reference to customer value leaves the cheapest value lever unpulled before a sale.
In the order we would run them. Each is on its own page, most with a free Excel template.
Names the buyer archetype and builds the benchmark sheet of what it underwrites.
Separates unproven revenue from the core book so retention reflects the customers a buyer is really acquiring.
Shows every contract's terms and deviations, including assignment and change of control.
Keeps a consistent, contract-level record of recurring revenue over several years.
Moves pricing, retention, concentration and margin into run-rate where a buyer can see them hold.
Meaningful Exit Defines the exit first, so the work on revenue quality is aimed at the buyer you actually want.
Fit more than financial polish. Strategic acquirers price your customers' overlap with theirs, the category you occupy, integration risk, technical debt, key-person dependency and whether your contracts survive a change of control. Strong gross retention matters to them too, but they care less about this year's EBITDA than a private equity buyer does.
Decide early which buyer you are building for, then build what that buyer underwrites: durable retention, honest gross margin, low concentration, a team that runs without you, and records that reconcile. Write the exit down with the Meaningful Exit form and turn the gaps into quarterly priorities for years, not months.
Yes, early enough that the increase shows up in trailing revenue with retention holding, which takes at least two quarters. The play's range is 5% to 15%, tested on new business and then renewals; for most founders we would test a price 25% higher on new business first and aim for about 5% a year after that. A buyer will raise prices after close if you do not, and every dollar you prove first is paid for at the multiple.
It can, but buyers will inspect it separately. Report inference revenue and inference cost on their own lines so a buyer can see the margin and how the revenue behaves at renewal. Usage-driven AI revenue with no history of retention will be valued more cautiously than contracted subscription revenue.
Our equity work aims at exactly these metrics, and our Exit Platform connects portfolio companies with more than 100 private equity firms that buy vertical SaaS. A founder can do all of this without outside capital using the same plays.
Growth equity →Reviewed by Dougal Cameron, CEO & Co-Founder on 2026-10-22. Golden Section observations are labeled separately from external benchmarks and illustrative arithmetic.