The standard · version 1.1.0 · 15 November 2026
Measures
Reported beside the six pillars in every review, and never scored. None has a pass mark.
Gross margin
The share of revenue left after the cost of delivering the product
- Why investors ask
- Gross margin is what is left of each pound of revenue to pay for everything else. For an AI company it is the first question, because every query costs money to serve. A low margin also makes a customer slower to pay back and growth more expensive.
- How it is read from a deck
- The unit economics slide, or the slide on the AI. Look for gross margin as a percentage, the period it covers, and what is counted in the cost: hosting, model inference, data and software bought in, and the people who support or deliver the product. Costs charged on each use, such as card fees or text messages, belong in it too. SVV's material does not define gross margin; Craft Ventures' published definition is in the definitions. A slide that says what its cost counts has defined its figure. Where it does not, the report asks how the figure was worked out. For an AI product, look also for the cost of one unit of work beside its price, and for whether free credits from a model provider are hiding a cost. If the deck gives no gross margin, the report says so and it becomes one of the questions.
- For AI companies
- The classic figures and the AI figures are not comparable. Read an AI company against the AI lines. Y Combinator warns that free credits from a model provider hide a real cost: a company that claims a high margin on credits has a shock coming when they run out (Y Combinator).
What is published
- Classic software, at $1M to $10M of ARR
- 70% on average. 85% or more for the top performers
- SourceBessemer Venture Partners, 21 September 2021Scaling to $100 Million
- Classic software, at any size
- An average of 65% to 70%, with the middle half of companies between about 60% and 80%
- SourceBessemer Venture Partners, 21 September 2021Scaling to $100 Million
- Classic software, the long-term aim
- At least 75%
- SourceCraft Ventures, 18 October 2021The SaaS Metrics That Matter
- Sales-led B2B software, by First Round's levels of product-market fit
- Level 1 (pre-seed or seed): not yet known. Level 2 (seed or Series A): not less than 50%. Level 3 (Series A to C): 60% to 70% or more. Level 4: 80% or more
- SourceFirst Round, 8 April 2024Levels of PMF
- AI companies
- 50% to 60%. Bessemer sets that against 80% to 90% for classic software (2026). a16z set it against 60% to 80% or more (2020)
- SourcesBessemer Venture Partners, 10 February 2026The AI pricing and monetization playbookAndreessen Horowitz, 16 February 2020The New Business of AI (and How It's Different From Traditional Software)
- AI products, from a survey of about 300 software companies
- 45% in 2025, with 53% projected for 2026 and 59% for 2027
- SourceICONIQ, 2026State of AI: The Builder’s Economy (July 2026)
- Fast-growing AI companies, by kind
- About 25%, and often negative, for the fastest group, which Bessemer calls Supernovas. About 60% for the group that grows like strong software companies, which it calls Shooting Stars
- SourceBessemer Venture Partners, 13 August 2025The State of AI 2025
- AI companies, the low end
- Many at 20% to 40%
- SourceNotion Capital, 2 October 2025Not All ARR is Created Equal
- An exception worth knowing
- Twilio ran at about 50% for much of its time in Bessemer's portfolio, because it paid telecom carriers, and is one of the strongest performers in Bessemer's index of public cloud companies
- SourceBessemer Venture Partners, 21 September 2021Scaling to $100 Million
Capital efficiency
Cash burned for each unit of net new ARR: the burn multiple
- Why investors ask
- It shows what growth costs. David Sacks, who named it, calls it a catch-all: a margin problem, a sales problem or a churn problem each shows up in it. He adds that a company burning three times its new ARR or more may not have the product-market fit it appears to have.
- How it is read from a deck
- The financial slide. The burn multiple is net burn divided by net new ARR over the same period: the cash that went out for each pound of recurring revenue added. Net burn is the fall in the bank balance. Net new ARR is the rise in ARR after churn. Look for the multiple itself, or for the two figures it is made from, over the same quarter or year. Where the deck gives only the two figures, the report sets them side by side and shows the division. Before there is revenue the multiple cannot be worked out, and the report says so.
- For AI companies
- These figures come from classic software. An AI company's model costs lower its gross margin, and Sacks lists a gross margin problem first among the things that push the multiple up (David Sacks 2020).
What is published
- David Sacks's scale, for venture-stage companies
- Under 1×: amazing. 1 to 1.5×: great. 1.5 to 2×: good. 2 to 3×: suspect. Over 3×: bad
- SourceDavid Sacks, 23 April 2020The Burn Multiple
- By stage, David Sacks
- A seed company "might have a Burn Multiple of 3 because it just started selling". After the Series A, "it might drop to 2"
- SourceDavid Sacks, 23 April 2020The Burn Multiple
- Before there is revenue, David Sacks
- The multiple cannot be worked out. Sacks says that is fine, and that the job is to keep burn low and get to first revenue quickly
- SourceDavid Sacks, 23 April 2020The Burn Multiple
- Fast-growing SaaS companies, Craft Ventures
- Under 1× is amazing. Anything under 2× is still quite good
- SourceCraft Ventures, 18 October 2021The SaaS Metrics That Matter
- Point Nine's napkin, by round
- Pre-seed: nothing asked. Seed: no figure, "Resourcefulness, ability to get a lot done with a small team". Series A: a burn multiple under 3 to 4×
- SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
- How much it matters to investors
- 91% of the 35 investors Point Nine asked in 2023 said capital efficiency mattered more to them than the year before
- SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
- Sales-led B2B software, by First Round's levels of product-market fit
- Level 1 (pre-seed or seed): not yet known. Level 2 (seed or Series A): not worse than 5×. Level 3 (Series A to C): 1 to 3×. Level 4: 0 to 1×
- SourceFirst Round, 8 April 2024Levels of PMF
Where demand comes from
How customers found the company, and how much of that rests on the founders' own contacts
- Why investors ask
- A customer won cold proves more than a customer won through a friend. First Round builds its levels on this: early demand comes mostly from the founders' network, and the test of the next level is a channel that does not depend on warm introductions. Bessemer tells founders not to count sales to former colleagues as proof of demand.
- How it is read from a deck
- The traction slide and the slide on how the company sells. Look for where each customer or lead came from: the founders' contacts, introductions from investors, cold outreach, inbound enquiries, referrals from customers, partners, or self-serve sign-ups. For a self-serve product, look for where sign-ups and new paying accounts come from, by channel, and what share of sign-ups go on to pay: the report states these with no level beside them. A wall of logos says nothing about this. If the deck does not say, the report says so and it becomes one of the questions.
- For AI companies
- First Round's levels were written for sales-led companies. Much of the demand for AI products is self-serve: see Menlo's figure.
What is published
- First Round, level 1 (pre-seed or seed)
- Mostly friends and network, with some cold outreach. One in 10 to 20 warm introductions becomes a customer
- SourceFirst Round, 8 April 2024Levels of PMF
- First Round, level 2 (seed or Series A)
- Early signs of a scalable channel that does not depend on warm introductions from investors or friends. Without a warm introduction, close to 10% of first calls end in a closed deal
- SourceFirst Round, 8 April 2024Levels of PMF
- First Round, level 3 (Series A to C)
- One or more scalable channels. More than 10% of inbound enquiries come from referrals and word of mouth. 10% to 15% of first calls end in a closed deal
- SourceFirst Round, 8 April 2024Levels of PMF
- First Round, level 4 (later)
- Two or more scalable channels. More than 15% of first calls end in a closed deal
- SourceFirst Round, 8 April 2024Levels of PMF
- Point Nine's napkin, by round
- Seed: "Organic traction" and "Success with founder-led sales". Series A: "Success with at least one acquisition channel"
- SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
- Bessemer, on the first million of ARR
- The founders should win the first dozen customers themselves, putting cold outreach first to test real demand. Sales to former colleagues or close contacts should not be counted as a sign of success
- SourceBessemer Venture Partners, 11 February 2025The founder’s playbook for scaling to $1 million ARR
- Unusual Ventures, on early customers
- If the early customers were all "friendlies", the company will stumble when it starts to sell in earnest
- SourceUnusual Ventures, 2 September 2026Startup Fundraising Guide: From Pre-Seed to Series A Capital
- AI applications bought self-serve
- 27% of all spending on AI applications comes through product-led growth, against 7% in traditional software
- SourceMenlo Ventures, 9 December 20252025: The State of Generative AI in the Enterprise
Regretted churn
Customers lost that the company wanted to keep
- Why investors ask
- It tests need, not liking. First Round, on how it judges satisfaction: "we’ve intentionally used “need” and not “love” here". A customer who needed the product and still left is the signal an investor looks for. Losing early customers who were a poor fit is normal while the product settles.
- How it is read from a deck
- The retention slide. Look for how many customers have left, why each left, and which of them the company would have wanted to keep. First Round sets those apart from customers who were never the right fit, which it calls non-regretted. A deck that gives one churn figure has not made the split, and the report says so.
- For AI companies
- Early churn in AI products runs higher. Notion Capital reports many AI companies losing 10% to 30% of their customers every month (Notion Capital 2025). a16z counts retention from month three, after the people who were only trying the product have gone (Andreessen Horowitz 2025). A deck for an AI product should say which customers it counts.
What is published
- First Round, level 1 (pre-seed or seed)
- No renewals yet, and no regretted churn. Some non-regretted churn is expected
- SourceFirst Round, 8 April 2024Levels of PMF
- First Round, level 2 (seed or Series A)
- 10% to 20% regretted churn. Above 20% is one of its warning signs at this level
- SourceFirst Round, 8 April 2024Levels of PMF
- First Round, level 3 (Series A to C)
- Under 10% regretted churn. Above 10% is one of its warning signs at this level
- SourceFirst Round, 8 April 2024Levels of PMF
Speed
How much has been done in the time the company has had
- Why investors ask
- Investors read progress against the clock. Y Combinator: "the longer they’ve been around, the more they have to have done." a16z says start-ups raising now "need a strong velocity story": in revenue if there is any, and otherwise in how fast they ship.
- How it is read from a deck
- The title, team, traction and plan slides. Look for dates: when the company started, when the product first shipped, when the first customer paid, and what has happened since the last round. Set what has been done beside the time it took. Where the deck gives no dates, the report says so and it becomes one of the questions.
- For AI companies
- The figures from before 2025 are from classic software and are now the low end. a16z: what was best in class "is now on the lower end of growth we’re seeing" (Andreessen Horowitz 2025).
What is published
- From launch to $1M of ARR, B2B, investors asked in 2022
- Within a year is good. Within nine months is great
- SourceLenny Rachitsky, 26 April 2022What is a good growth rate
- Before AI, best in class
- $1 million of ARR in the first twelve months
- SourceAndreessen Horowitz, 6 June 2025What "Working" Means in the Era of AI Apps
- Enterprise AI companies, revenue twelve months after the first revenue
- $1.2M for the bottom quarter. $2.1M at the median. $5.3M for the top quarter
- SourceAndreessen Horowitz, 6 June 2025What "Working" Means in the Era of AI Apps
- Enterprise AI companies, time from first revenue to Series A
- 13 months for the bottom quarter. 9 months at the median. 7 months for the top quarter
- SourceAndreessen Horowitz, 6 June 2025What "Working" Means in the Era of AI Apps
- Fast-growing AI companies that grow like strong software companies
- About $3M of ARR within the first year of revenue
- SourceBessemer Venture Partners, 13 August 2025The State of AI 2025
- From seed to Series A, Y Combinator
- A typical 18 to 24 months. A company that has been going longer needs more to show for it
- SourceY Combinator, undatedHow to build a great Series A pitch and deck
- At seed, before there is much revenue, Point Nine
- "Team has shown that it can build fast with a small team and can iterate quickly based on new insights"
- SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
- A dissenting view
- A couple of investors told Lenny Rachitsky they do not mind how long a company takes to get going, as long as it takes off once it does
- SourceLenny Rachitsky, 26 April 2022What is a good growth rate
Customer references
Customers who will speak for the product
- Why investors ask
- A Series A investor will call customers. Point Nine lists strong customer references beside high usage and low churn as the evidence of product-market fit it looks for at Series A.
- How it is read from a deck
- The traction slide or the customer slide. Look for named customers quoted in their own words, case studies with results, or a statement of how many customers will take a call from an investor. A logo is not a reference. If the deck names no one who will speak, the report says so.
What is published
- Point Nine's napkin, Series A
- "Strong customer references", listed with clear evidence of product-market fit: high usage, low churn
- SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
- Unusual Ventures, by the end of the seed phase
- A solution that delights early customers, and "a tight set of references to build on"
- SourceUnusual Ventures, 2 September 2026Startup Fundraising Guide: From Pre-Seed to Series A Capital