Your gross margin payback period for startups can be double what you report AI startups that report customer acquisition cost payback on revenue can need exactly twice as long to recover CAC when gross margin is 50%, because the revenue-based figure ignores inference, hosting and human review costs, according to an analysis of the margin-adjusted formula. The corrected calculation — CAC ÷ (new monthly recurring revenue per customer × gross margin) — turns a $15,000 CAC and $1,500 of monthly revenue into a 10-month revenue payback but 18.2 months at a 55% margin, and routing easy requests to a cheaper model lifted that example's margin to about 66% and cut payback to 15.2 months. Bessemer Venture Partners' State of AI report put typical gross margins at roughly 25% for fast-growing AI startups it calls Supernovas and 60% for Shooting Stars, against 75% to 80% for traditional SaaS, while David Skok's rule of thumb is recovering CAC in under 12 months. Divide acquisition cost by revenue and an AI startup looks healthy. Divide it by gross profit and the same company can need twice as long to earn its money back. - CAC payback measured on revenue ignores inference, hosting and review costs, so an AI company at 50% gross margin takes exactly twice as long to recover CAC as its revenue figure says. - The formula is CAC divided by new monthly recurring revenue per customer times gross margin, and it takes four inputs: loaded sales and marketing spend, new customers, starting MRR and cost of revenue. - In the worked example, a $15,000 CAC and $1,500 of monthly revenue gives a 10-month revenue payback but 18.2 months at a 55% margin. - Routing easy requests to a cheaper model lifted that example's margin to about 66% and cut payback to 15.2 months without signing a single customer. - Benchmarks run from about 6 to 12 months for small-business sales up to 18 to 24 for enterprise, and David Skok's rule of thumb is under 12 months. A founder tells an investor the company earns back its customer acquisition cost in ten months. The investor asks a single follow-up: ten months of what? If the answer is revenue, the number flatters you, and the gross margin payback period for startups is the figure that replaces it. You can work it out in five minutes with four inputs from your billing system and your cloud invoices. The gap matters more for AI companies than it ever did for classic software. Bessemer Venture Partners' State of AI report sorted fast-growing AI startups into two groups, the Supernovas and the Shooting Stars, and put typical gross margins at roughly 25% for the first and 60% for the second. Traditional SaaS businesses usually run at 75% to 80%. Inference bills, GPU rentals and human review all sit inside cost of revenue, and they grow with usage, so every dollar a customer pays leaves less behind to cover what it cost to win them. Lead with the margin-adjusted number yourself. An investor will compute it within a minute of seeing your deck, and it lands far better as something you volunteered than something they caught. The usual version is short. Take your fully loaded sales and marketing spend for a period, divide it by the new customers you won, and you have CAC. Divide CAC by the monthly recurring revenue a new customer brings and you have payback in months. David Skok, the investor behind the For Entrepreneurs blog and its SaaS metrics series, has long argued that you should recover CAC in under 12 months, and he frames the calculation on gross margin rather than revenue. How to calculate gross margin for AI startup books before investors do https://startupfortune.com/how-to-calculate-gross-margin-for-ai-startup-books-before-investors-do/ Ledgerly, an invented fintech-adjacent company, shows how four misfiled bills change the picture: a $720,000 LLM API bill filed under R&D, plus vector database, extraction checks and contract reviewers. Total COGS rises from $690,000 to $1,810,000, so the reported 82.75% falls by 28 points. - how to calculate gross margin for AI startup https://startupfortune.com/how-to-calculate-gross-margin-for-ai-startup-books-before-investors-do/ - AI startup gross margin compared to SaaS https://startupfortune.com/how-to-calculate-gross-margin-for-ai-startup-books-before-investors-do/ Most founders skip that last step. Revenue payback is the easy one, your billing tool hands you the inputs, and with an 80% gross margin the gap is small: a 10-month payback becomes 12.5. Nobody loses sleep over that. The adjusted version looks like this: CAC payback in months = CAC ÷ new MRR per customer × gross margin . If you can't count customers cleanly, use the aggregate form instead. Divide the quarter's sales and marketing spend by net new ARR times gross margin, then multiply by 12. Two details trip people up. Fully loaded means salaries, commissions, sales tools, agency fees and the share of founder time that goes into selling, not only ad spend. And if your sales cycle runs a quarter, lag the spend: money spent in January produced the customers who signed in April, so divide January's spend by April's wins. Decide, too, whether you're measuring new-logo CAC or a blended figure. Blended CAC divides all spend by all new ARR, including expansion from existing accounts, which makes payback look shorter. Investors accept it for companies with heavy expansion, but they'll ask for the new-logo version, so build that one first. A worked example at 55% gross margin Say you sell an AI support agent to mid-sized e-commerce teams. Last quarter you spent $600,000 on sales and marketing, covering two account executives, one marketer, tooling and a webinar series, and you signed 40 customers. CAC is $15,000. Each customer pays $18,000 a year, which is $1,500 a month. These figures are illustrative, picked to be easy to follow. Revenue payback: $15,000 divided by $1,500 is 10 months. That's the number on the slide. Now open the cost side. For each customer you pay about $480 a month in model inference, $75 for hosting and vector storage, and $120 for support plus a human reviewer who checks the agent's riskiest replies. That's $675 of cost of revenue against $1,500 of revenue, leaving $825 of gross profit and a gross margin of 55%. Payback becomes $15,000 divided by $825, or 18.2 months. How to Calculate CAC Payback Period for a B2B SaaS Startup https://startupfortune.com/how-to-calculate-cac-payback-period-for-a-b2b-saas-startup/ A worked example shows the math: $300,000 in fully-loaded sales and marketing spend, 20 new customers at $2,000 a month each, and 75% gross margin produce a 10-month payback. Bessemer Venture Partners treats anything under 12 months as capital-efficient and over 24 months as a flag. - how to calculate CAC payback period for SaaS https://startupfortune.com/how-to-calculate-cac-payback-period-for-a-b2b-saas-startup/ - CAC payback period formula investors expect to see https://startupfortune.com/how-to-calculate-cac-payback-period-for-a-b2b-saas-startup/ Eight extra months, and nothing about the business changed. Push the margin down to 50% and payback is 20 months, exactly double the revenue figure. At 40% it's 25 months, well past the 12 to 18 months investors expect at this deal size. Now the fix, which shows why the number is worth calculating at all. Suppose you route the easy 60% of requests to a smaller, cheaper model and inference falls by a third, to about $320. Cost of revenue drops to $515, gross profit rises to $985, margin lands near 66%, and payback shortens to 15.2 months. You didn't sign a customer or cut a salary. You moved three months by changing where the requests go. What belongs in cost of revenue for an AI company The margin is only as honest as what you put in it. Inference is the obvious line, whether you pay a model provider per token or rent GPUs by the hour. Add embeddings and vector storage, hosting, third-party data you pay for per query, payment processing, the support and onboarding staff who keep accounts running, and any labelers or reviewers in the loop. If a person checks the agent's output before the customer sees it, that wage is cost of revenue. Credits are the quiet distortion. Model providers and cloud platforms hand out startup credits, and a company living on them can show a 75% margin that will vanish the month the credits run out. Calculate at list price, or at the rate in your committed-spend contract, and show the credit separately if you want investors to see it. Free trials need a decision too. Inference spent on prospects who never pay is an acquisition expense, so it belongs in CAC, not cost of revenue. Putting it in neither is what makes some payback numbers look so clean. If your product burns real compute during a trial, divide that spend across the deals you closed. Check the spread as well. An average margin hides the customer who runs ten times the queries of everyone else. Pull gross margin by customer for the last three months, and if the bottom fifth is negative, your blended 55% is carrying accounts you'd be better off repricing. Setting up the sheet so you can rerun it monthly Build one row per monthly cohort of new customers. The columns are lagged sales and marketing spend, new customers, CAC, average starting MRR, cost of revenue per customer, gross margin and payback months. Pull cost of revenue from the model-provider and cloud invoices for that cohort's accounts rather than spreading a company-wide figure across everyone, since a cohort of heavy users and a cohort of light ones don't share a margin. Keep revenue payback in a column right beside it. The distance between the two columns is the cost of serving your product, measured in months, and it's the first thing an investor's eye will go to. Rerun it every month, because a quarter of data can hide one bad cohort that a monthly view would have flagged. Benchmarks investors use to judge the result Skok's line is 12 months or less. Bessemer's cloud benchmarks have long broken payback out by deal size: roughly 6 to 12 months for small-business sales, 12 to 18 for mid-market and 18 to 24 for enterprise, measured on gross margin. The example above sells at $18,000 a year, so it sits in the mid-market band, and 18.2 months is on the very edge of it. The 15.2 months after rerouting is comfortably inside. AI companies get some slack at the high end because margins tend to climb as model prices fall and routing improves, and investors know it. Slack isn't a free pass. They'll want to see margin by monthly cohort, trending up for at least two quarters, before they accept a forward-looking number. If you forecast 70% on a company that's at 45% today, expect most investors to run the math on 45%. Compute the figure per segment and per channel, not once for the whole company. Outbound enterprise deals and self-serve signups have different CACs and often different margins, and a healthy blend can hide a channel that takes 30 months to pay back. What the payback number still leaves out Payback says nothing about whether customers stay. With 18.2 months of payback and 30% annual logo churn, a simple constant-churn estimate says roughly four in ten customers leave before they've repaid what they cost to acquire. Pair the payback figure with gross-margin-adjusted net revenue retention. If retention is weak, no payback target will rescue the model. It also treats a dollar of revenue as a dollar of margin forever. Usage-based pricing narrows that problem, since revenue rises with compute. Flat seat pricing widens it: a customer paying $18,000 a year who triples their usage on an unlimited plan pays you the same while your cost to serve them triples. Annual prepay changes the cash picture. A customer who pays $18,000 upfront covers the $15,000 CAC in cash on day one, before any cost of serving them, and many investors look at that cash payback alongside the margin-based figure. Report the two separately. Margin payback tells you whether the unit economics work, and cash payback tells you how much funding the growth needs. To run it tonight, pull four numbers for last quarter: fully loaded sales and marketing spend, new customers, average starting MRR and cost of revenue at list price. Divide CAC by MRR times margin. Then compare it with the revenue version and put both in the deck. At a 55% margin, a 10-month revenue payback is 18.2 months. At 50% it's 20. The investor will do that division in their head, so do it first. Also read: How to set an AI agent tool call timeout before one slow API stalls a run https://startupfortune.com/how-to-set-an-ai-agent-tool-call-timeout-before-one-slow-api-stalls-a-run/ • How to Build a Golden Dataset for AI Agent Evaluation From Your Logs https://startupfortune.com/how-to-build-a-golden-dataset-for-ai-agent-evaluation-from-your-logs/ • How to write an AI agent memory retention policy before it bites https://startupfortune.com/how-to-write-an-ai-agent-memory-retention-policy-before-it-bites/ This article is posted in Startup News https://startupfortune.com/category/startup/ , check it out for more related stories. Join the discussion Open in the community → https://startupfortune.com/community/ Almost there. Sign in and your reply posts straight away.