What the next investor will test at Series A, pillar by pillar. Each indicator is read by its parts, and each part gets its own status. A part has a floor and, where sources publish a range, a typical and a strong level. Each level says where it comes from.
Type a word or two. Only what holds them stays.
Nothing on this page holds those words.
Team
Can this team sell and ship without a founder doing every part of it?
Coverage
The people who build the product are in the company: a technical founder at the start, named owners of product and engineering by Series A.
The bar at Series A
Floor
Product and engineering each have a named owner
SVV's judgementNo published threshold. Point Nine asks at Series A for a team that has "Built a high-performing tech team", which a deck cannot show.
How it is read, and what counts
How it is read from a deck
The team slide. Look for the name of the person who builds, and what that person has shipped before: a product, a system or a company, with a place and a year.
Shown
At pre-seed and seed: the deck names the person who builds the product and gives three facts about software that person has released before: what it was, the company it was built at, and the year. At Series A: the deck names the person who owns product and the person who owns engineering. One person may hold both.
Claimed
At pre-seed and seed: the deck names a technical founder, gives someone a title such as CTO, or says the team is technical, and any of the three facts is missing. At Series A: the deck says product and engineering are covered, or shows a technical team, and names no owner for one or both.
Not met
The deck says the product is built by an agency, by contractors or by anyone outside the company, or says the company has no technical founder or is looking for one. At Series A: the deck says the post of product owner or of engineering owner is vacant. Where this applies it is Not met, whatever else the deck says.
Unknown
The deck is silent.
What would evidence it
The code repository and its history; a reference call with someone who has worked with the technical founder; at Series A the organisation chart.
Applies to
Every company.
For AI companies
a16z asks whether AI has ended the old rule against non-technical founders, and gives the case against: "there’s still a big gap between a demo and a reliable product that can scale". A founder who built the first version with AI tools still has to show who will make it reliable.
People other than the founders are selling.Sold by a sales team
The bar at Series A
Floor
One person who is not a founder is selling and has closed deals
SVV treats a VP of Sales or CRO in post six months and at quota as the strong level, not the floor. Published guidance puts that hire later. Point Nine asks at Series A for "Early evidence of a repeatable, efficient sales motion beyond founder sales" and for "2-3 excellent ICs in product/marketing/sales" (individual contributors: people who do the work and do not manage a team). It puts "at least 1-2 proven VPs" at Series B. Bessemer: "Don’t hire a VP of Sales too early", and consider one only with revenue "ideally approaching $1 million ARR" and strong retention. Unusual Ventures asks before a Series A for an "Early GTM team with a motion that’s working now" (GTM is go-to-market: the people who sell and market). How much of the new revenue the team closes is read under Revenue.
How it is read, and what counts
How it is read from a deck
The team slide and the slide on how you sell. Look for a named person in a selling role who is not a founder, and for deals that person closed.
Shown
The deck names a person in a selling role who is not a founder and gives at least one deal that person closed: a named customer, a count of deals, or new revenue split by who closed it.
Claimed
The deck says there is a sales team, a sales hire or a sales leader, and gives no deal closed by a named person who is not a founder.
Not met
The deck says every deal so far was closed by a founder, or that the first sales hire has not started.
Unknown
The deck is silent.
What would evidence it
New ARR by who closed it, founder or team, from the CRM; the start date of each seller.
Applies to
Only companies that sell through a sales team.
Where the figures are from
Point Nine’s figures are in US dollars, from 86 rounds and Carta’s US data. Bessemer and Unusual Ventures are American. Notion Capital, a UK investor, places the move "beyond founder-led sales" between $1m and $3m of ARR.
A named person owns growth: how users find the product, start using it and start paying.Bought self-serve
The bar at Series A
Floor
A named owner of growth
SVV's judgementIt is the self-serve counterpart of the part asked of sales-led companies, that people other than the founders are selling. No published threshold. Bessemer says a company with a product-led motion "may not need a traditional sales team at all". Point Nine asks at Series A for "Success with at least one acquisition channel".
How it is read, and what counts
How it is read from a deck
The team slide and the slide on how customers arrive. Look for the person who owns sign-ups, first use and conversion to paid, and for the channel that brings users in.
Shown
The deck names the person who owns growth and gives at least one dated number that person answers for: sign-ups, first use or conversion to paid.
Claimed
The deck says growth is "product-led" or "organic", or describes how users arrive, and either names nobody who owns growth or gives no dated number for it.
Not met
The deck says nobody owns growth, or that it waits for a hire not yet made.
Unknown
The deck is silent.
What would evidence it
Product analytics by channel: sign-ups, first use and conversion to paid, by month.
Y Combinator’s Series A checklist asks the company to list any officer "not currently devoting 100 percent of their business time to the Company".
How it is read, and what counts
How it is read from a deck
The team slide. Look for "full-time" beside each founder, with a date. At pre-seed, a date by which each founder leaves their other job counts.
Shown
The deck says each founder is full-time and gives a date for each: the date they started, or at pre-seed the date they will start, no later than the close.
Claimed
The deck says the founders are full-time, "committed" or "all in", and for at least one founder gives no date. A team slide with names and titles and nothing about time is Unknown, not Claimed.
Not met
The deck shows a founder who holds another job or a role at another company and gives no date for leaving it, or at pre-seed gives a leaving date after the close. Where this applies it is Not met, whatever the deck says of the other founders. In a university spinout, an academic founder who keeps a university post is read under the university relationship and does not count against this part.
Unknown
The deck is silent.
What would evidence it
Employment contracts or resignation letters with dates; the university’s written approval where a founder keeps an academic post.
Applies to
Every company.
The founders’ shares vest: a founder who leaves early does not keep them all.
Rarely on a slide. It may be on the team slide, the ask, or a cap table in the appendix. Look for the word "vesting" with its length.
Shown
The deck says founder vesting is signed and gives the vesting period. At pre-seed, a statement that it will be signed at the close, with the period, also counts.
Claimed
The deck mentions founder vesting and leaves out one of the two things Shown asks for: that it is signed (or at pre-seed will be signed at the close), or the period.
Not met
The deck says founder shares are held outright with no vesting and gives no plan to add it.
Unknown
The deck is silent.
What would evidence it
The founder agreement or shareholders’ agreement, and the cap table.
Applies to
Every company.
The company owns the intellectual property its product depends on, or holds a signed licence to it.
For a university spinout a signed licence counts. Its terms are read under the university relationship.
How it is read, and what counts
How it is read from a deck
The team slide, the product slide or the appendix. Look for who owns the code, the models and the rights to data, and whether founders, staff and contractors have assigned their work to the company.
Shown
The deck says the IP is assigned to the company and names who assigned it: founders, staff, contractors or a previous employer. For a university spinout: it names the signed licence and its date. At pre-seed, a statement that the assignment or licence will be signed before the close also counts.
Claimed
The deck says the company owns its IP, its code or its technology, and does not say who assigned it or name a licence.
Not met
The deck says work the product depends on belongs to someone else (a founder personally, a previous employer, a contractor or a university) and names no assignment or licence, signed or to be signed before the close.
Unknown
The deck is silent.
What would evidence it
IP assignment agreements for founders, staff and contractors; for a spinout, the signed licence.
Applies to
Every company.
An option pool is set up, and the options promised to staff have been granted.
The bar at Series A
Floor
A pool set up, and the grants promised to staff made
Y Combinator: "If you have any pending or promised equity grants you haven't already taken care of before you start your Series A process, do that ASAP." NFX lists "Not enough ESOP for current employees (10% ESOP standard for post Series A)" among the common cap table pitfalls. The floor asks one thing of the grants: those the company promised have been made.
How it is read, and what counts
How it is read from a deck
The cap table, usually in the appendix or on the ask slide. An option pool (ESOP) is the share of the company set aside for employees. Look for its size as a percentage and how much of it is granted.
Shown
The deck gives the option pool as a share of the company and says how much of it has been granted.
Claimed
The deck mentions an option pool or an ESOP and leaves out its size, how much is granted, or both.
Not met
The deck says there is no option pool, or that grants promised to staff have not been made.
Unknown
The deck is silent.
What would evidence it
The cap table with the pool and each grant; board approval of the grants.
Applies to
Every company.
Where the figures are from
Both sources are American. SVV found no published figure for the UK or Ireland.
Y Combinator’s Series A checklist asks for "Any correspondence or documents relating to any pending or threatened action, suit or proceeding or investigation".
How it is read, and what counts
How it is read from a deck
A deck does not usually say. A cap table on a slide, with nothing flagged, is a claim.
Shown
Not available from a deck. Only a cap table checked against the company’s records shows it. This part is Claimed, Not met or Unknown.
Claimed
The deck shows a cap table or says the cap table is clean, and mentions no dispute.
Not met
The deck mentions a dispute over shares, a founder who has left with shares still unresolved, or a contested promise of equity.
Unknown
The deck is silent.
What would evidence it
The cap table reconciled to the share register; the lawyers’ disclosure of any claim.
Applies to
Every company.
Hiring
The hires the last round was raised to make have been made.
The bar at Series A
Floor
Two or three hires made in product, marketing or sales, and an engineering team in place
SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
Point Nine’s Series A team row: "Built a high-performing tech team. Recruited 2-3 excellent ICs in product/marketing/sales." The share of the hiring plan filled on time is reported separately, with no pass mark. First Round’s level 2, seed or Series A, is a team of up to 20 people.
How it is read, and what counts
How it is read from a deck
The team slide. Look for people hired since the last round into product, marketing or sales, and for who is in engineering.
Shown
The deck names at least two people who are not founders in product, marketing or sales roles, and either names an engineer who is not a founder or gives the size of the engineering team.
Claimed
The deck gives a headcount, a team photograph or a list of roles, and leaves out one of the two things Shown asks for: two named people in product, marketing or sales, or the engineering team.
Not met
The deck says the team is still the founders alone, or that the roles the last round was to pay for are still open.
Unknown
The deck is silent.
What would evidence it
The hiring plan against the hires made, and the organisation chart.
Applies to
Every company.
For AI companies
ICONIQ’s 2026 survey of about 300 executives at software companies building AI products: 78% of companies are rethinking workforce planning and 33% expect smaller teams. A small team at Series A is not a gap in itself.
How much of the hiring plan was filled, and when.
The bar at Series A
No pass mark The report states the hires planned and the hires made, where the deck gives them. It gives no pass mark for the share of a hiring plan filled on time.
A figure sometimes quoted is 80% of a twelve-month plan filled on time. Nothing published supports it as a pass mark, and the trend runs the other way: 91% of the investors Point Nine surveyed in 2023 said capital efficiency mattered more than before, and a third of the companies ICONIQ surveyed in 2026 expect smaller teams. A company that hired fewer people than it planned and still hit its numbers has done well.
How it is read, and what counts
How it is read from a deck
The team slide or the plan. Look for hires planned against hires made.
Shown
The deck gives two numbers for a stated period: the hires planned and the hires made.
Claimed
The deck says hiring is on plan, ahead or behind, and leaves out one of the two numbers or the period.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
The hiring plan against the hires made, with dates.
Applies to
Every company.
Who has left in the last twelve months.
The bar at Series A
No pass mark The report states any departures the deck gives. It gives no pass mark.
A figure sometimes quoted is no regretted leavers in twelve months. No investor framework publishes one, and SVV found nothing published on staff leaving. A founder leaving is a red flag of its own.
How it is read, and what counts
How it is read from a deck
Decks rarely say. Look on the team slide for a note of a departure, or for a founder or senior hire named in the story and missing from the team.
Shown
The deck gives the number of people who left in the last twelve months, which may be none, and the role of each.
Claimed
The deck says the team is stable, or that it keeps its people, and gives no number.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
The list of leavers with dates and reasons.
Applies to
Every company.
Governance
The board meets at least every quarter and gets a written pack beforehand.
The bar at Series A
Floor
Four meetings a year, each with a written pack
SVV's judgementNone of the sixteen investors whose published guidance SVV read names governance as a reason to invest. Y Combinator’s Series A checklist asks for "All minutes of directors’ and stockholders’ meetings" once a term sheet is signed, so the record has to exist.
How it is read, and what counts
How it is read from a deck
The team slide or the appendix. Look for who sits on the board and how often it meets.
Shown
The deck names the directors, says the board meets at least four times a year and says a written pack is issued.
Claimed
The deck mentions a board or shows board members and leaves out any of the three: the directors’ names, how often it meets, or the pack.
Not met
The deck says there is no board beyond the founders, or gives a number of meetings a year below four.
Unknown
The deck is silent.
What would evidence it
Board minutes and packs for the last twelve months.
Applies to
Every company.
At least one director is not a founder: an investor or an independent.
The bar at Series A
Floor
One investor or independent director
SVV's judgementNo published threshold.
How it is read, and what counts
How it is read from a deck
The team slide. Look for board members who are not founders, and whether each is a director or an observer.
Shown
The deck names a person who is not a founder and says that person is a director, as an investor or as an independent.
Claimed
The deck names investors or advisers as being on the board, or as attending it, without listing who the directors are.
Not met
The deck lists the directors and all of them are founders. An observer is not a director.
Unknown
The deck is silent.
What would evidence it
The list of directors at the company register; the shareholders’ agreement.
Applies to
Every company.
Where the figures are from
Seed investors differ. Seedcamp, a European seed investor: "we do not take board seats but we do typically take up board observer roles", usually "until at least the next round of funding or until the Series A". So a seed-backed company can reach its Series A with observers and no outside director.
Red flags
Any one of these caps the pillar at level 2.
A founder who is not full-time. At pre-seed: a founder who will still be part-time after the round closes.
Intellectual property the product depends on has not been assigned to the company: work a founder did before it was formed, work done at a previous employer, or work done by a contractor. Raised only where the deck says or shows this. A silent deck is not checked.
A founder has left in the last twelve months, or a dispute between founders is unresolved.
Product
Is there a product customers rely on, and can it be defended?
In production
The product is live in production and customers use it on their own data. Not a demo.
The bar at Series A
Floor
10 customers in production
SVV's judgementIt is SVV’s own figure and no source sets it as a floor. It sits inside First Round’s level 2, seed or Series A, which runs from 5 to 25 customers.
Unusual Ventures reports hearing this bar and disagrees with it. It lists "Having 10-plus customers" among the things founders are told investors must see, and then says: "At Unusual, we disagree." Its objection is that a count says nothing about whether the customers are alike: "if these early customers are a diverse assortment of ICPs, this will not yield efficient growth going forward". So the count is read with the ICP match under Market. First Round’s level 3, Series A to C, is 25 to 100 customers.
How it is read, and what counts
How it is read from a deck
The product and traction slides. In production means live and used for real work, not a demo and not a test account. Look for how many customers or design partners use it, named or counted, and for a screen or a number from real use. A customer the deck lists only as a letter of intent, or as a pilot not yet started, is not using it. For a self-serve product, a dated count of paying or active accounts is the number.
Shown
The deck gives the number of customers or design partners using the product in production, the number is at or above the floor for the round, and it backs the number with at least one of: named customers, a dated usage figure, or a screen of the product in use. A list of named customers counts as the number.
Claimed
The deck says the product is live or in production, or gives a number of users or customers at or above the floor, and gives none of the three: named customers, a dated usage figure, or a screen of it in use.
Not met
The deck gives a number in production that is below the floor for the round, or says the product is a demo, a pilot not yet started or still being built.
Unknown
The deck is silent.
What would evidence it
Production metrics and usage by customer; a live demonstration on real data.
Applies to
Every company.
Where the figures are from
Both sources are American and describe sales-led B2B companies. A self-serve product with many small accounts passes the count easily; for it the usage parts say more.
How much of the time the product is available to customers.
The bar at Series A
No pass mark The report states the uptime the deck gives, the period and the source. It gives no pass mark.
A figure sometimes quoted is 99.5% over six months. No investor framework publishes an uptime bar, and SVV found nothing published that supports one. What a customer needs differs by product and is set in its contract.
How it is read, and what counts
How it is read from a deck
The product or trust slide, or the appendix. Uptime is the share of time the product was available. Look for the percentage, the months it covers and where the figure comes from.
Shown
The deck gives an uptime percentage, the period it covers and where it comes from: a monitoring tool or a public status page.
Claimed
The deck says the product is "reliable" or "enterprise grade", or gives an uptime percentage and leaves out the period, the source or both.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
Uptime from the monitoring tool for the last six months, and the incident log.
Applies to
Every company.
Usage
Accounts use the product in every period it is meant to be used.
The bar at Series A
Floor
80% of paying accounts active in each period
SVV's judgementNo published figure for accounts. The direction is supported: Point Nine asks at Series A for "Clear evidence of PMF (high usage, low churn, high NPS)".
The same rule on the period applies as at seed. Craft Ventures publishes the only engagement figures SVV found, and they measure something else: users, not accounts. "A good metric for most SaaS startups is 40% DAU/MAU" (daily active users over monthly active users) and "60% DAU/WAU" (daily over weekly), on working days.
How it is read, and what counts
How it is read from a deck
The traction slide. Look for active accounts as a share of all accounts, what counts as active, and the period: a week for most products, a month or a quarter where the job the product does comes round that often.
Shown
The deck gives the share of accounts active in each period, says what counts as active, names the period and dates the figures, and the share is at or above the floor for the round. At pre-seed: it shows each design partner’s use for the last six weeks, and each used the prototype in at least four.
Claimed
The deck says something about how much the product is used (for example "daily" or "heavily", or a share of active accounts at or above the floor) and leaves out any of what Shown asks for: the share, what counts as active, the period or the dates. Revenue billed on what accounts used in a period says something about use: with no share of active accounts it is Claimed, not Unknown.
Not met
The deck gives a share of active accounts that is below the floor for the round, with or without the rest. At pre-seed: it shows a design partner who used the prototype in fewer than four of the last six weeks.
Unknown
The deck is silent.
What would evidence it
Product analytics: active accounts in each period for the last six months, with the meaning of active.
Applies to
Every company.
For AI companies
a16z finds that early use of AI products overstates what lasts: people sign up to try them and leave within a couple of months, so it measures from month three. Read usage from the first weeks after sign-up with that in mind.
Customers stay: of the customers you had twelve months ago, the share that are still customers.
The bar at Series A
Floor
90% where customers are enterprises, 85% for mid-market, 70% for small businesses
Craft Ventures: "90-95% is common for enterprises, 85% for mid-market, and 70-80% for small businesses. Logo Retention below these benchmarks could be evidence of a problem." The floor here is the lower end of each of Craft’s ranges and the strong level is the upper end. That split is this standard’s reading of Craft’s words; Craft does not publish a floor. First Round counts only regretted churn, customers the company wanted to keep: 10 to 20% at its level 2 and under 10% at level 3.
How it is read, and what counts
How it is read from a deck
The traction or retention slide. This is called logo retention. Look for it, or for customers lost, over twelve months, with the number of customers it is worked out on.
Shown
The deck gives logo retention over twelve months, or the customers at the start and the customers lost, says which kind of customer it sells to (enterprises, mid-market or small businesses), and the figure is at or above the floor for that kind.
Claimed
The deck says churn is low or that no customer has left, or gives a retention figure, and leaves out the twelve-month period, the number of customers it is worked out on, or which kind of customer it sells to. A figure below the floor is Not met, not Claimed.
Not met
The deck gives a logo retention figure that is below the floor for the kind of customer it says it sells to, or below 70% where it does not say which kind.
Unknown
The deck is silent.
What would evidence it
The customer list at the start and the end of the twelve months, from the billing system.
Applies to
Every company.
For AI companies
Early retention in AI products is less reliable than in classic software. a16z measures retention from month three, after the people who were only trying the product have left.
Where the figures are from
Craft’s figures are from US software companies in 2021 and predate AI companies. SVV found no published European figure.
Technical position
The company holds something a competitor cannot copy quickly: data, a workflow customers depend on, or integrations.
The bar at Series A
Floor
The data or workflow advantage shown at work, with evidence
Point Nine at Series A: "Increasing conviction that you can create defensibility e.g. by leveraging proprietary data or network effects". a16z: "Defensibility still comes from owning the workflow end-to-end, embedding deeply into customer systems, becoming the system of record, or earning trust over time."
How it is read, and what counts
How it is read from a deck
The slide on the product, the AI or the moat. Look for what is named as the company’s own and for anything that shows it at work.
Shown
At pre-seed: the deck says which of three things will be the company’s own (data only it holds, a workflow customers run their work through, or integrations into customers’ systems) and how it will be built up. At seed: the deck names one such asset that exists and gives a fact about it: its size, how it grows with use, or the customers whose systems it is built into. At Series A: as at seed, and the deck gives a result or a deal won that it puts down to the asset.
Claimed
The deck says its technology is proprietary, unique or defensible and falls short of Shown for its round: it names no asset of the three kinds, or names one and gives nothing more.
Not met
The deck says the product has no data, workflow or integrations of its own and gives no plan for any, or the only thing it names as its own is a set of prompts on a third-party model. An integration counts as the company’s own. Prompts alone do not.
Unknown
The deck is silent.
What would evidence it
The IP summary; the list of data sets with the rights to each; the list of customer integrations.
Applies to
Every company.
For AI companies
a16z: "AI is an extraordinary engine for product differentiation. But it’s not a moat." Point Nine: a moat from AI needs "proprietary data that is impossible to obtain for others", and that data must lead "to superior model performance".
The product has a measured advantage on the things customers care about.
The bar at Series A
Floor
An advantage measured in customers’ live use, against a named alternative
SVV’s judgement on the wording. No published threshold. For AI companies the direction is published: Bessemer says "AI vendors will need to surface evidence of effectiveness before purchase, not just after deployment."
How it is read, and what counts
How it is read from a deck
The product, traction or competition slide. Look for a result from customers’ live use (time saved, cost, accuracy, revenue), what it is compared with, and the date.
Shown
The deck gives a result as a number, says it was measured in customers’ live use, names what it is compared with, and dates it.
Claimed
The deck says the product is better, faster or more accurate and any of the four is missing: the number, that it comes from customers’ live use, what it is compared with, or the date. Results from the company’s own tests only are Claimed.
Not met
The deck gives numbers in which the product does no better than the alternative it names.
Unknown
The deck is silent.
What would evidence it
The results on live work, and customer references who will confirm the numbers.
Applies to
Every company.
For AI companies
For an AI product the measure is the quality of the output on live work, with its date. A score on a public benchmark does not count: Bessemer says "Public benchmarks aren’t enough."
There is a written note on what is protected and how.
The bar at Series A
Floor
A written IP strategy
SVV's judgementNo investor framework sets a stage for it.
How it is read, and what counts
How it is read from a deck
Rarely on a slide; sometimes in the appendix. Look for what is covered by patents, copyright or trade secrets, and who owns each.
Shown
The deck says what is protected and by which means: a patent, copyright or a trade secret.
Claimed
The deck says "patent pending", "protected IP" or similar and leaves out what is protected, the means, or both.
Not met
The deck says nothing is protected, or that protection has not been considered.
Unknown
The deck is silent.
What would evidence it
The IP strategy and any filings.
Applies to
Every company.
Engineering maturity
The product can be built and released the same way every time.
The bar at Series A
Floor
Automated testing and deployment on every change (CI/CD)
SVV's judgementNone of the sixteen investors whose published guidance SVV read names internal systems as a reason to invest. It is checked in diligence, so have it ready. A deck is not expected to show it.
How it is read, and what counts
How it is read from a deck
A deck does not usually say. It may be on a product or technology slide.
Shown
The deck says how the product is built and released, and what it describes meets the level for the round: at pre-seed a build anyone on the team can repeat, at seed automated deployment, at Series A automated testing and deployment on every change.
Claimed
The deck says something about how often or how well it releases and does not say how releases are made, or what it describes is short of the level for the round.
Not met
The deck says releases are made by hand, or that only one person can build or release the product.
Unknown
The deck is silent.
What would evidence it
The repository and the deployment pipeline, shown live.
Applies to
Every company.
The quality of what the model produces is monitored on live use.AI only
Bessemer: "Private, continuous evaluation is mission-critical".
How it is read, and what counts
How it is read from a deck
The slide on the AI. Look for how the quality of output is tracked in production: accuracy on live work, drift (quality slipping over time), failures, and how often it is checked.
Shown
The deck names what is measured on live use (for example accuracy on live work, drift or failures) and gives a figure for it with a date.
Claimed
The deck says the model is "continuously monitored" or "continuously improving" and leaves out what is measured, the figure or its date.
Not met
The deck says quality is checked only before release, or only when a customer complains.
Unknown
The deck is silent.
What would evidence it
The monitoring dashboard and the log of incidents.
Applies to
Only AI companies.
The live product is monitored: errors, speed and availability, with alerts.Not AI
The bar at Series A
Floor
Production monitoring with alerts
SVV's judgementIt is the counterpart, for software that is not AI, of the production model monitoring asked of AI companies. No published threshold.
How it is read, and what counts
How it is read from a deck
A deck does not usually say. It may be on a product or trust slide.
Shown
The deck names what is monitored in production (errors, speed or availability) and says alerts are in place.
Claimed
The deck says the product is "monitored" or "robust" and leaves out what is monitored, the alerts or both.
Not met
The deck says faults are found only when customers report them.
Unknown
The deck is silent.
What would evidence it
The monitoring dashboard and the log of incidents.
Applies to
Only software companies that are not AI companies.
Known shortcuts in the code are written down in a register.
The bar at Series A
Floor
A technical-debt register kept
SVV's judgementNo published threshold. Unusual Ventures lists "removing early technical debt" among the product work that follows a Series A, so an investor will ask what there is.
How it is read, and what counts
How it is read from a deck
A deck does not usually say. Technical debt is the shortcuts taken to ship sooner that will need rework. Look for it on a product or plan slide.
Shown
The deck says a technical-debt register is kept and names at least one item on it.
Claimed
The deck says the code is "clean" or "scalable", or mentions technical debt, and either does not say a register is kept or names no item on it.
Not met
The deck says technical debt is not recorded.
Unknown
The deck is silent.
What would evidence it
The technical-debt register.
Applies to
Every company.
What shipped in the last two quarters, against what was planned.
The bar at Series A
No pass mark The report lists what the deck says shipped in the last two quarters, with dates. It gives no pass mark for the share of a roadmap shipped on time.
A figure sometimes quoted is 70% of the last two quarters’ roadmap shipped on time. No investor publishes it. What investors publish is about speed, measured differently: Point Nine asks at seed for a team that can "iterate quickly based on new insights", and ICONIQ reports that "cycle time and deploys per week still hold up as velocity proxies". A team that changed its roadmap because of what it learned would fail a share-on-time test for doing the right thing. How much has been done in the time is read in the case.
How it is read, and what counts
How it is read from a deck
The product or plan slide. Look for what was released in the last two quarters, with dates.
Shown
The deck lists what shipped in the last two quarters with a date for each item, or gives planned against shipped as two numbers.
Claimed
The deck says the team "ships fast" or "delivers on its roadmap" and gives neither a dated list nor the two numbers.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
The roadmap against what shipped, by quarter.
Applies to
Every company.
Red flags
Any one of these caps the pillar at level 2.
The company neither owns the code and models its product is built on nor holds a signed licence to them. Raised only where the deck says or shows this. A silent deck is not checked.
No product in production. Raised only where the deck says or shows this. A silent deck is not checked.
A critical dependency on one third-party model, with no tested way to run on another. Raised only where the deck says or shows this. A silent deck is not checked.AI only
The product depends on one third-party platform, interface or data source that could be withdrawn, with no fallback. Raised only where the deck says or shows this. A silent deck is not checked.Not AI
Market
Does the revenue come from the customer you defined, in a market big enough for a venture investor?
ICP
The customers you are winning match the ideal customer you wrote down.
The bar at Series A
Floor
80% of new ARR in the last twelve months from customers who match the ICP
SVV's judgementNo published figure. The direction is supported by Unusual Ventures and by Bessemer’s advice to keep the ICP "uncomfortably narrow".
Unusual Ventures, on the customers won before a Series A: "if these early customers are a diverse assortment of ICPs, this will not yield efficient growth going forward".
How it is read, and what counts
How it is read from a deck
The slide on how you sell, or traction. At seed, look for how many design partners and open deals match the ICP. At Series A, look for the share of new ARR (annual recurring revenue added) in the last twelve months that came from customers who match it. For a self-serve product, read pipeline as new sign-ups.
Shown
The deck gives the share that matches the ICP, with the counts or the revenue behind it, and the share is at or above the floor for the round: half of design partners and pipeline taken together at seed, 80% of new ARR in the last twelve months at Series A. For a self-serve product: the share of new paying accounts.
Claimed
The deck says its customers are in its target segment, or shows their logos, and gives no share, or gives a share at or above the floor with no counts or revenue behind it.
Not met
The deck gives a share, or counts from which the share can be worked out, and it is below the floor for the round.
Unknown
The deck is silent.
What would evidence it
The CRM export with each customer and open deal marked as matching the ICP or not; new ARR by customer. For a self-serve product: sign-ups and paying accounts by kind of customer, from product analytics.
Applies to
Every company.
Market size
The market you can reach with this product and these channels is large enough.
The bar at Series A
Floor
£250m a year
SVV’s judgement, in sterling. No investor publishes a threshold for SAM. Published thresholds are for the total market.
How it is read, and what counts
How it is read from a deck
The market slide. SAM is the part of the total market you can reach with this product and the ways you sell, in every country you mean to sell in. Look for it as a yearly figure. A figure the deck calls its serviceable or reachable market counts, whatever the label. Markets the deck sizes one by one, each with a count and a price, are read together, and the report shows the addition as the reader’s working. A market that depends on something the deck says is not yet shown does not count. A deck in dollars or euros is read against the same number in its own currency.
Shown
The deck gives a SAM as a yearly figure at or above the floor for the round, and shows the sum behind it: a count of customers and a price.
Claimed
The deck gives a SAM at or above the floor and no sum behind it. A figure for the reachable market taken from a report is Claimed.
Not met
The deck gives a SAM below the floor for the round, with or without working. A figure for the first market alone, where the deck names further markets and leaves them out of the count, is not the SAM: the part is Unknown, and the report states the figure and what was left out.
Unknown
The deck is silent.
What would evidence it
The bottom-up market model.
Applies to
Every company.
Where the figures are from
These bars are SVV’s own, in sterling, written for UK and Irish enterprise AI software companies. They are for everything the product can reach, not for the UK and Ireland alone.
The market size is built bottom-up: a count of target customers multiplied by what each would pay.
The market slide. Bottom-up means the number of target customers multiplied by what each would pay in a year. Look for the count, where the count comes from, and the price.
Shown
The deck shows the count of target customers, the yearly price used and the multiplication. At pre-seed the count also carries a named source.
Claimed
The deck gives a market size and any of what Shown asks for is missing: the count, the price, the multiplication, or at pre-seed the source of the count. A figure quoted from a report with no sum of the deck’s own is Claimed. So is a market worked out top-down only: a total from a report with a percentage applied, such as "1% of a £30bn market".
Not met
The deck says its target customers have not been counted, or that it has no price to multiply by.
Unknown
The deck is silent.
What would evidence it
The market model, with the source and date of each count.
Applies to
Every company.
The total market is large enough for a venture investor.
The bar at Series A
Floor
£1bn a year
SVV’s judgement, in sterling. It is SVV’s own figure. No source publishes a threshold in pounds. The published figures are in dollars and are not converted: Sequoia says "a $1B potential", Y Combinator says "Total Available Market (TAM) >$1B if possible", and Balderton says "a billion-dollar revenue opportunity".
Strong
Several billion dollars
SourcePoint Nine, 30 September 2023What does it take to raise capital, in SaaS, in 2023? (the 2023 SaaS Funding Napkin, Christoph Janz)
The sources are Sequoia Capital, Y Combinator and Balderton Capital 2022. Each is quoted in its own currency and nothing is converted. A deck that gives its market in dollars or euros is read against a billion in that currency. Point Nine asks at Series B for "a multi-billion dollar market", in dollars.
How it is read, and what counts
How it is read from a deck
The market slide. TAM is what every target customer, everywhere, would spend in a year at your price. Look for the figure and the sum behind it. A figure for one country, one region or a first market is not a TAM, and a reader does not add regions together to make one: where the deck gives no figure for the whole market, the part is Unknown. A deck in dollars or euros is read against the same number in its own currency.
Shown
The deck gives a TAM and shows the sum behind it: a count of target customers and a yearly price, multiplied. At Series A the figure is also a billion a year or more.
Claimed
The deck gives a TAM with no sum of its own behind it. A figure quoted for a whole industry or taken from a report is Claimed, whatever source is named.
Not met
At Series A: the TAM the deck gives is below a billion a year, with or without working. Not given at seed, where this part has no pass mark.
Unknown
The deck is silent.
What would evidence it
The market model, bottom-up and top-down, with sources.
Applies to
Every company.
Where the figures are from
The published figures are in US dollars. Balderton is a European investor; Sequoia and Y Combinator are American. Where the market does not exist yet no figure can be shown: Balderton says "some markets may not yet exist", and that founders then have to work harder to convince an investor it is there.
There is a worked path from today’s numbers to £10m of ARR.
The bar at Series A
Floor
£10m of ARR at today’s contract value and win rate
SVV's judgementNo published figure at this level. Published paths are longer: NFX asks how "you get to $100M+ in revenue with a bottoms-up calculation", and Sequoia tells pre-seed and seed founders "You should be able to chart a path to $500M in revenue".
NFX: "The VC isn’t questioning whether you can be a $5-6M business, they’re questioning whether you can be a $100M revenue business."
How it is read, and what counts
How it is read from a deck
The market or plan slide. ARR is annual recurring revenue; ACV is the average contract value, what one customer pays in a year. Look for how many customers at today’s ACV make £10m of ARR, and how many deals at today’s win rate that takes. For a self-serve product, read today’s price and conversion rate. A deck in dollars or euros is read against the same number in its own currency.
Shown
The deck shows three numbers and the sum that links them: today’s ACV, the number of customers that makes £10m of ARR at that ACV, and the deals or sign-ups needed at today’s win or conversion rate.
Claimed
The deck shows revenue reaching £10m or more and any of the three numbers is missing.
Not met
The number of customers the deck needs for £10m of ARR, at the contract value it gives, is larger than the number of target customers it counts. A forecast that stops short of £10m because the plan’s period ends is not Not met: it is silent on the rest of the path, which is Unknown.
Unknown
The deck is silent.
What would evidence it
The model: customers, average contract value, win rate and sales capacity by quarter.
Applies to
Every company.
Where the figures are from
NFX and Sequoia are American and write in US dollars. SVV’s £10m is in sterling.
Why now
There is a named, dated reason customers buy now.
The bar at Series A
Floor
One documented trigger: a regulation, a platform shift or a budget line
Point Nine asks at Series A for a compelling "why now".
How it is read, and what counts
How it is read from a deck
The why-now slide. Look for one outside change with a date: a regulation, a shift in a platform or a technology, or a new budget line.
Shown
The deck names one change outside the company (a regulation, a shift in a platform or a technology, or a new budget line), gives its date, and gives one fact about it that can be looked up: the name of the regulation, a published price or a published target.
Claimed
The deck names a change or a trend outside the company (for example "AI is here") and leaves out its date, the fact that can be looked up, or both.
Not met
The only reasons the deck gives for the timing are inside the company: the product is ready, the team is in place, the round is open.
Unknown
The deck is silent.
What would evidence it
The regulation, price curve or published target itself, with its date.
Applies to
Every company.
Customers name the trigger themselves.
The bar at Series A
Floor
5 customer references cite it
SVV's judgementNo published count. The direction is supported: Point Nine asks at Series A for "Strong customer references".
How it is read, and what counts
How it is read from a deck
The why-now, problem or traction slide. Look for quotes from interviewees or customers, each with a name or with a role and organisation, giving the trigger as their reason.
Shown
The deck quotes or cites people who give the trigger as their reason, each identified by name or by role and organisation, and there are enough of them: 2 interviewees at pre-seed, 2 customers at seed, 5 customers at Series A. Whether anyone was prompted is not read from a deck.
Claimed
The deck says customers give this reason and identifies fewer of them than the round asks for, or none, without saying those are all there are.
Not met
The deck quotes customers on why they buy and none of them gives the trigger, or says how many customers cite the trigger and the number is below what the round asks for.
Unknown
The deck is silent.
What would evidence it
Interview notes at pre-seed; reference calls at seed and Series A.
Applies to
Every company.
A budget owner is identified in the deals you close.Sold by a sales team
The bar at Series A
Floor
In 80% of closed deals
SVV's judgementNo published figure. The direction follows Sequoia’s advice to "Target customers who will move fast and pay a premium", and this standard’s pre-seed red flag for a buyer with no budget.
Does not apply to a self-serve product, where the user pays.
How it is read, and what counts
How it is read from a deck
The slide on how you sell. Look for the role that signs and pays, and in how many closed deals that person was identified.
Shown
The deck names the role that holds the budget and gives the share of closed deals in which that person was identified, and the share is 80% or more.
Claimed
The deck names a buyer or a role that holds the budget and gives no share of closed deals.
Not met
The deck gives a share below 80%, or says deals are signed by users with no budget of their own.
Unknown
The deck is silent.
What would evidence it
The CRM export: closed deals, with the budget owner named on each.
Applies to
Only companies that sell through a sales team.
Competitive position
Wins and losses are recorded for at least ten deals.Sold by a sales team
The bar at Series A
Floor
10 deals with the outcome recorded
SVV's judgementNo published count.
How it is read, and what counts
How it is read from a deck
The competition slide or the slide on how you sell. Look for deals won and lost, by competitor.
Shown
The deck gives the number of deals won and the number lost, and together they are 10 or more.
Claimed
The deck says it records wins and losses and gives no numbers.
Not met
The deck gives numbers of deals won and lost that together are below 10, or says outcomes are not recorded.
Unknown
The deck is silent.
What would evidence it
The CRM export with stage definitions and the outcome of each deal.
Applies to
Only companies that sell through a sales team.
You win at least 30% of the deals in which a named competitor was also considered.Sold by a sales team
The bar at Series A
Floor
30% against named competitors
SVV's judgementNo published figure for a win rate against competitors. First Round publishes a different measure, the share of first calls that end in a closed deal: "approaching 10%" at its level 2 and 10 to 15% at level 3.
How it is read, and what counts
How it is read from a deck
The competition slide or the slide on how you sell. Look for a win rate, the competitors it is measured against and the number of deals behind it.
Shown
The deck gives its win rate in deals where a named competitor was also considered, with the number of deals behind it, and the rate is 30% or more.
Claimed
The deck gives a win rate of 30% or more and leaves out the competitors, the number of deals or both, or says it wins most of its deals and gives no figure.
Not met
The deck gives a win rate below 30%.
Unknown
The deck is silent.
What would evidence it
The CRM export: deals by competitor and outcome.
Applies to
Only companies that sell through a sales team.
Where the figures are from
First Round’s figures are for US sales-led B2B companies.
Lost deals are grouped by reason.Sold by a sales team
The bar at Series A
Floor
Loss reasons categorised
SVV's judgementNo published threshold.
How it is read, and what counts
How it is read from a deck
The competition slide or the slide on how you sell. Look for why deals were lost, with a count for each reason.
Shown
The deck groups its lost deals by reason and gives a count for each reason.
Claimed
The deck gives one or more reasons for losing deals and no counts.
Not met
The deck says the reasons for losing are not recorded.
Unknown
The deck is silent.
What would evidence it
The win and loss record, with the reason for each loss.
Applies to
Only companies that sell through a sales team.
You know what paying customers used before, and why they changed.Bought self-serve
The bar at Series A
Floor
Recorded for 10 paying customers
SVV's judgementIt is the self-serve counterpart of the win/loss record asked of sales-led companies, with the same count of ten. No published figure.
How it is read, and what counts
How it is read from a deck
The competition or traction slide. Look for what customers used before, why they switched, and how many customers that is based on.
Shown
The deck says what paying customers used before and why they changed, and says it comes from 10 or more paying customers.
Claimed
The deck says what customers switch from, or why, and gives no number of customers it is based on.
Not met
The deck bases it on fewer than 10 paying customers, or says it is not known.
Unknown
The deck is silent.
What would evidence it
The survey or interview notes, and product data on what customers imported or replaced.
Applies to
Only companies whose customers buy self-serve.
Red flags
Any one of these caps the pillar at level 2.
No written ideal customer profile. Raised only where the deck says or shows this. A silent deck is not checked.
The ideal customer profile changed more than twice in the last twelve months.
One customer is more than half of revenue.
An incumbent, a supplier the customer already uses, has shipped the same feature and there is no differentiated win to show.
Revenue
Is there a repeatable way to win?
Level
Annual recurring revenue (ARR)
The bar at Series A
Floor
£1m of ARR, or £600k of ARR growing 15% a month
SVV's own figure, set as the floor. It is at the bottom of the published range, which starts at about $1M (Point Nine 2023). A founder who also plans to pitch larger funds should read it as the least an investor will consider, not as the target.
Typical
About $1M to $5M of ARR. For enterprise AI companies, a median of $2.1M of revenue twelve months after the first revenue
Point Nine's napkin gives the range, about $1M to $5M, and no strong level. Its author added a view of his own, offered as a possibility and not a finding: that most Series A investors may now expect $2.5M to $3M or more, against $1M to $2M in earlier years. Most Series A companies in his survey had several million dollars of ARR. The only top-quarter figure published is a16z's, and it covers enterprise AI companies alone. Craft Ventures wrote in 2021, when rounds were happening earlier, that the threshold had fallen from $1 million to around $500k (Craft Ventures 2021). SVV's second route, £600k growing fast, belongs to that earlier view.
How it is read, and what counts
How it is read from a deck
The traction slide. ARR here is headline ARR as SVV defines it: recurring subscription or usage revenue, invoiced and paid under a signed contract, stated as an annual run-rate. Pilot fees, set-up fees and services are not part of it. SVV's definitions ask for services to be flagged, and so shown on their own, when they are above 20% of revenue. Services above 50% of revenue is a red flag, which caps the pillar. Contracts signed and not yet invoiced are shown separately. Revenue the deck shows is not billed (a pilot with no fee, a letter of intent, a forecast) is not ARR. For a self-serve product paid by card, ARR is twelve times the recurring revenue paid in the month, with one-off revenue left out: see MRR and ARR in the definitions.
Shown
The deck gives ARR with its date and says what is counted in it, and the figure is at or above the floor. On the floor's second route, the lower figure with growth of 15% a month, it also shows the monthly figures behind the growth.
Claimed
The deck gives a figure at or above the floor with no date and nothing to say what is counted. "ARR" with nothing to say what is in it is a claim.
Not met
The deck's own figure is below the floor: under the lower of the floor's two figures, or under the higher with growth below 15% a month.
Unknown
The deck is silent.
What would evidence it
The contract register: each signed contract with its start date, its annual value and whether it has been invoiced and paid. SVV's ARR bridge template is that register. For a self-serve product: recurring revenue by month, from the billing system.
Applies to
Every company.
For AI companies
AI companies reach these levels faster, so more is expected of them. In a16z's data on enterprise AI companies (June 2025), revenue twelve months after the first revenue was $1.2M for the bottom quarter, $2.1M at the median and $5.3M for the top quarter. The median company raised its Series A nine months after its first revenue. Before AI, $1 million of ARR in the first twelve months was best in class (Andreessen Horowitz 2025).
Where the figures are from
Every published figure here is American and in dollars. SVV found no European investor that publishes a revenue threshold for seed or Series A. SVV's floor was set for UK and Irish companies.
Growth
How fast recurring revenue is growing
The bar at Series A
Floor
3× year on year, or 15% a month sustained for six months
The floor is the benchmark Craft Ventures publishes for leading a Series A: at least 15% a month below $1M of ARR and 10% above it, where 10% a month is about three times a year. Y Combinator says a trend needs at least four to six months to be believed. Bessemer's cloud companies at $1M to $10M of ARR grew by nearly 200% a year on average, which is the same as the floor, and its top performers by 230% or more. Point Nine's 2023 survey of SaaS rounds found a lower range: about two to three times a year, and now and then one and a half to two (Point Nine 2023). Where a deck shows two to three times, this part is Not met against the floor and the report says the figure is inside Point Nine's range.
How it is read, and what counts
How it is read from a deck
The traction slide. Look for recurring revenue month by month, and for what the deck says the figures count. Recurring revenue is billed and paid for, with pilot fees, set-up fees and services left out, and a label on a chart does not make it so. The rate has to hold over the whole period the bar names, so one strong month does not count. Monthly figures that can be read off a running total count: a chart that only adds up is a fault of the slide and does not lower this status.
Shown
The deck shows recurring revenue month by month, or gives the start and end figures with their dates, over the period the bar names, and says what the figures count. What they count is recurring revenue as the definitions have it. The growth is at or above the floor.
Claimed
The deck states a growth rate at or above the floor and does not show the monthly figures or the start and end figures behind it. Or it shows figures for a shorter period than the bar names. Or it shows figures it calls ARR, MRR or revenue and does not say what they count, or says they include pilot fees, one-off fees or services without the recurring part at the start and the end of the period.
Not met
Over the period the bar names, the deck's own figures give growth below the floor. Or the deck's own account of its customers shows that none of them pays, so there is no recurring revenue to grow, whatever a chart is labelled.
Unknown
The deck is silent.
What would evidence it
Twelve months of monthly recurring revenue, from the billing system or the accounts.
Applies to
Every company.
For AI companies
These figures come from classic software, and AI companies are growing faster. The fast-growing AI companies Bessemer tracks quadruple their ARR in each of their first two years (Bessemer Venture Partners 2025). In a16z's data the median enterprise AI company went from $0.7M of revenue six months after its first revenue to $2.1M at twelve months (Andreessen Horowitz 2025).
Where the figures are from
All of these figures are American. Nothing comparable is published for the UK or Ireland.
Repeatability
New ARR closed by someone other than a founderSold by a sales team
The bar at Series A
Floor
40% of new ARR in the last two quarters
SVV's own figure. No source publishes a percentage. The direction is supported. Point Nine looks at Series A for "Early evidence of a repeatable, efficient sales motion beyond founder sales" (Point Nine 2023). Bessemer has the founder selling alone to $100k of ARR, then with a first hire, then a sales team from $1M (Bessemer Venture Partners 2025).
Does not apply to a self-serve product, where nobody closes the sale.
How it is read, and what counts
How it is read from a deck
The slide on how the company sells. Look for new ARR in the last two quarters, split by who closed it: a founder, or a member of the team.
Shown
The deck gives the split of new ARR for the last two quarters by who closed it, and 40% or more was closed by people who are not founders.
Claimed
The deck says the team closes deals without the founders, and gives no split of new ARR.
Not met
The deck's own split shows less than 40% of new ARR closed by people who are not founders, or the deck says the founders close every deal.
Unknown
The deck is silent.
What would evidence it
The CRM export of deals closed in the last two quarters, with the name of the person who closed each.
Applies to
Only companies that sell through a sales team.
The time from first conversation to signed contract is not getting longerSold by a sales team
The bar at Series A
Floor
No longer than in the period before
SVV's judgementA stable sales cycle is asked for, with no figure. It is written here as a test of direction so that two readers agree. Nothing is published on a threshold.
How it is read, and what counts
How it is read from a deck
The slide on how the company sells. The sales cycle is the number of days from a first qualified conversation to a signed contract. Look for it for two periods or more, such as the last two quarters.
Shown
The deck gives the sales cycle for two or more periods in a row, and the latest is no longer than the one before.
Claimed
The deck gives one figure for the sales cycle, or says it is stable or getting shorter, with no figure for an earlier period.
Not met
The deck's own figures show the sales cycle longer in the latest period than in the one before.
Unknown
The deck is silent.
What would evidence it
The CRM export of closed deals with the date each was opened and the date each was signed.
Applies to
Only companies that sell through a sales team.
The share of deals won is not fallingSold by a sales team
The bar at Series A
Floor
No lower than in the period before
SVV's judgementA stable win rate is asked for, with no figure. It is written here as a test of direction so that two readers agree. Nothing is published on a threshold for stability.
First Round measures something close: the share of first calls, made without a warm introduction, that end in a closed deal. It gives about 10% at its second level and 10% to 15% at its third (First Round 2024). A win rate counted from qualified deals will be higher, so the report says which the deck uses.
How it is read, and what counts
How it is read from a deck
The slide on how the company sells. The win rate is the share of qualified deals that end in a signed contract. Look for it for two periods or more, and for where the count starts: from the first call, or from a qualified deal.
Shown
The deck gives the win rate for two or more periods in a row, and the latest is no lower than the one before.
Claimed
The deck gives one win rate, or says it is stable or improving, with no figure for an earlier period.
Not met
The deck's own figures show the win rate lower in the latest period than in the one before.
Unknown
The deck is silent.
What would evidence it
The CRM export of deals won and lost, by quarter.
Applies to
Only companies that sell through a sales team.
How many months a new customer takes to pay back what it cost to win them (CAC payback)
The floor of 18 months is First Round's figure for its third level and Bessemer's target for companies selling to mid-sized customers. 15 months is the average for Bessemer's cloud companies at $1M to $10M of ARR. Bessemer sets the target by the size of customer: under 12 months selling to small businesses, under 18 to mid-sized companies, under 24 to large enterprises. Lenny Rachitsky's panel has the same shape: for enterprise customers 18 months is good and 24 is acceptable (Lenny Rachitsky 2021). Where a deck shows 18 to 24 months and sells to large enterprises, this part is Not met against the floor and the report says the figure is inside that allowance. The typical and strong levels are for a figure on gross profit that counts everyone who sold. Lenny Rachitsky: "The biggest mistake founders make when calculating their payback period is looking at revenue". a16z says CAC "should be the full cost of acquiring users" (Andreessen Horowitz 2015).
How it is read, and what counts
How it is read from a deck
The unit economics slide. CAC is customer acquisition cost: what is spent on sales and marketing to win one customer. SVV's material does not define CAC payback. The published definitions are in the definitions, and they measure it on gross profit with the full cost of selling in it. So the deck has to say three things about its figure: the period the costs and the new customers are taken from, what is counted in the cost, and whether payback is measured on gross profit or on revenue.
Shown
The deck states CAC payback in months, it is 18 months or less, and the deck says all three: the period, what is counted in the cost, and whether it is measured on gross profit or on revenue. The level is the floor and no higher where the figure is measured on revenue, or where the cost leaves out the pay of the people who did the selling, founders included. Say which in the reason.
Claimed
The deck states a payback of 18 months or less and leaves out one or more of the three.
Not met
The deck's own figure is more than 18 months.
Unknown
The deck is silent.
What would evidence it
Sales and marketing cost by month with the cost of founders' selling time in it, new customers and their revenue by month, and the gross margin used.
Applies to
Every company.
For AI companies
An AI company's gross margin is lower, so the same sales cost takes longer to earn back. Craft notes that lower-margin products with a high cost of acquisition do poorly on payback (Craft Ventures 2021). Check that the figure uses gross profit after model and hosting costs.
Retention and quality
Gross revenue retention over the last twelve months
The bar at Series A
Floor
90%
SVV's own figure. It is also the line in SVV's own definitions: "Investors look for 90% or more by Series A." (Sure Valley Ventures 2026). No outside source sets 90% as a minimum.
What is published is Bessemer's band: gross retention of 85% to 90% at every size of cloud company, and 70% to 80% in successful companies that sell to small businesses. SVV's floor is the top of that band, so the typical level here sits below the floor. Where a deck shows 85% to 90%, this part is Not met against SVV's floor and the report says the figure is inside Bessemer's band. No source gives a level above 90%.
How it is read, and what counts
How it is read from a deck
The retention slide. SVV's definition: opening ARR less churn and contraction over twelve months, divided by opening ARR. It leaves out expansion, so it cannot be above 100%. The deck states it with its period, or shows the opening ARR, contraction and churn it comes from. A figure called only "retention" or "revenue retention", with nothing to say whether it is gross or net, counts as silence on this part: the part is Unknown and the report asks which it is. So does a figure measured another way, such as customer retention, or revenue at month twelve over month three. So do words with no figure, such as "low churn". A shorter period counts where the deck states it and shows the opening ARR and the changes: the report states the period.
Shown
The deck gives a figure it calls gross revenue retention, GRR or gross dollar retention, with its twelve-month period, or shows the opening ARR, contraction and churn. The figure is 90% or more. Where one customer is above 30% of ARR by the deck's own figures, Shown also needs the figure without that customer. Without it the part is Claimed, and the report says the figure may rest on one customer.
Claimed
The deck gives a figure it calls gross retention, 90% or more, with no period and nothing it comes from.
Not met
The deck's own gross revenue retention is below 90%.
Unknown
The deck is silent.
What would evidence it
The cohort table from the billing system: opening ARR, expansion, contraction and churn for the twelve months.
Applies to
Every company.
For AI companies
a16z measures AI companies another way: revenue at month twelve as a share of revenue at month three. On that basis the median is 85%, the top quarter 95% and the top tenth 100% or more (Andreessen Horowitz 2025). It is not the same measure as gross revenue retention, and a deck that uses it should say so.
Where the figures are from
Bessemer's figures are from its own portfolio, mostly American, to 2021.
Floor: First Round's second level. Typical: Lenny Rachitsky's "good" for enterprise customers, and First Round's third level. Strong: Craft Ventures' figure for the best companies, and First Round's fourth level. All are US sources on classic software. For companies selling to small and mid-sized businesses, Lenny Rachitsky's panel put good at about 90% and great at about 110%.
How it is read, and what counts
How it is read from a deck
The deck states NRR with its period, or shows opening ARR, expansion, contraction and churn. SVV's definition: opening ARR plus expansion less contraction and churn, divided by opening ARR. A figure called only "retention" or "revenue retention", with nothing to say whether it is gross or net, counts as silence on this part: the part is Unknown and the report asks which it is. So does a figure measured another way, such as revenue at month twelve over month three. A shorter period counts where the deck states it and shows the opening ARR and the changes: the report states the period.
Shown
The deck gives a figure it calls net revenue retention, NRR or net dollar retention, with its twelve-month period, or shows the opening ARR, expansion, contraction and churn. The figure is 100% or more. Where one customer is above 30% of ARR by the deck's own figures, Shown also needs the figure without that customer. Without it the part is Claimed, and the report says the figure may rest on one customer.
Claimed
The deck gives a figure it calls net retention, 100% or more, with no period and nothing it comes from.
Not met
The deck's own net revenue retention is below 100%.
Unknown
The deck is silent.
What would evidence it
The cohort table from the billing system.
Applies to
Every company.
For AI companies
Early AI retention is less reliable. a16z measures it from month three to month twelve, to leave out the people who were only trying the product (Andreessen Horowitz 2025).
No single customer is too large a share of ARR
The bar at Series A
Floor
No customer over 30% of ARR
SVV's own figure. No source publishes 30%. The direction is supported: Craft calls it a significant risk when one or two customers make up most of the revenue (Craft Ventures 2021).
One customer over half of ARR is a red flag, which caps the pillar.
How it is read, and what counts
How it is read from a deck
The retention slide or the revenue quality slide. Look for the largest customer's share of ARR, or a customer list or chart with contract values from which it can be read.
Shown
The deck gives the largest customer's share of ARR, or lists customers with their contract values, and no customer is over 30%.
Claimed
The deck says revenue is spread or not concentrated, and gives no share and no contract values.
Not met
By the deck's own figures, one customer is more than 30% of ARR.
Unknown
The deck is silent.
What would evidence it
The customer list with contract values.
Applies to
Every company.
How customers are contracted and billed: by the year, for shorter terms, or by usage
The bar at Series A
No pass mark The report states the split as the deck gives it, and applies no pass mark.
A figure sometimes quoted is 60% or more of ARR on annual contracts. It is not used as a pass mark, because it runs against the trend. ICONIQ reports contracts getting shorter, with terms under a year on the rise, and usage-based pricing up from 35% to 42% of companies in six months (ICONIQ 2026; ICONIQ 2026). a16z's survey of 100 chief information officers found that most still prefer to pay for AI applications by usage (Andreessen Horowitz 2025). Such a bar would mark down a usage-priced AI company for being typical. SVV's own definition of ARR counts usage revenue under a signed contract (Sure Valley Ventures 2026).
How it is read, and what counts
How it is read from a deck
The pricing slide or the revenue quality slide. Look for the share of ARR on contracts of a year or longer, on shorter terms, and billed by usage.
Shown
The deck gives the split of ARR by contract length or by how it is billed, with its date.
Claimed
The deck describes its contracts in words, such as "mostly annual" or "usage-based", with no figures.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
The contract register, with the term and the billing basis for each customer.
Applies to
Every company.
Red flags
Any one of these caps the pillar at level 2.
Services or consulting over 50% of revenue
Pilots have finished and none has become a paid contractSold by a sales team
One customer over 50% of ARR
The deck counts as revenue or ARR something no customer is yet bound to pay for: a verbal yes, a letter of intent, a proposal or a pilot with no fee. For a self-serve product: revenue from accounts that have not paid. Raised where the slides show it, such as an ARR figure beside customers who are all unpaid.
Systems
Can the company run itself?
Cash and finance
A cash plan that reaches the next round
The bar at Series A
Floor
A model for the next 18 months that joins profit and loss, cash and headcount
SVV's judgementNothing is published on the length or the form. Y Combinator's Series A diligence checklist asks for the current business plan and any financial projections (Y Combinator).
How it is read, and what counts
How it is read from a deck
The plan slide and the ask slide. Look for cash by month or quarter and the month it runs out. At pre-seed, look also for the target the deck itself names for its seed round: a number, such as customers or revenue, and the date it expects to reach it. That target is what the floor means by the number the seed round needs. The report does not judge here whether the target is the right one: the Revenue pillar's seed bar does that. A runway figure on its own is a claim. At Series A the plan is a full model: profit and loss, cash and headcount joined together.
Shown
The deck shows the plan itself, as a table or chart by month or quarter, covering the months the bar for the round names, with the cash left at the end. At pre-seed the deck also names its target for the seed round and the date it expects to reach it, and the plan shows cash left at that date.
Claimed
The deck gives a runway figure, or says a plan or a model exists, and shows none of it. At pre-seed, a plan that is shown but names no target for the seed round, or no date for it, is also Claimed.
Not met
The plan the deck shows stops short of the months the bar names, or its own figures show the cash running out before them. At pre-seed: its own figures show the cash running out before the date the deck gives for reaching its seed target.
Unknown
The deck is silent.
What would evidence it
The plan or model itself, as a spreadsheet, with its assumptions.
Applies to
Every company.
Management accounts every month
The bar at Series A
Floor
Produced every month
SVV's judgementThat monthly accounts exist is SVV's judgement. Y Combinator's diligence checklist asks for the most recent financial statements at closing (Y Combinator).
A figure sometimes quoted is within ten working days of the month end. No source gives a figure for how fast, so there is no pass mark for speed. The report states how soon the accounts are ready, where the deck says.
How it is read, and what counts
How it is read from a deck
The financial slide or the appendix. Management accounts are the monthly profit and loss, balance sheet and cash figures a company produces for itself and its board. Look for actual monthly figures, and for how soon after the month end they are ready.
Shown
The deck shows figures month by month that it labels as actuals or as from the management accounts, and names the months. Figures labelled as forecast, plan or projection do not count.
Claimed
The deck says monthly management accounts are produced, and shows no monthly figures labelled as actuals.
Not met
The deck says accounts are produced less often than monthly.
Unknown
The deck is silent.
What would evidence it
The last twelve months of management accounts, with the date each was issued.
Applies to
Every company.
Actual cash has stayed close to the plan
The bar at Series A
Floor
Actual cash within 10% of the plan in each of the last two quarters
SVV's own figure. No source gives a figure for it, and no investor names it as a test. Nothing published supports the 10%. It is SVV's own measure of what it asks of this pillar: "Monthly numbers a board trusts" (Sure Valley Ventures 2026).
How it is read, and what counts
How it is read from a deck
The financial slide. Look for planned cash set beside actual cash, or planned burn beside actual burn, for each of the last two quarters.
Shown
The deck shows planned and actual cash, or planned and actual burn, for each of the last two quarters, and the gap is under 10% in both.
Claimed
The deck says the company is on plan or ahead of it, with no figures for plan and actual.
Not met
The deck's own figures show a gap of 10% or more in either quarter.
Unknown
The deck is silent.
What would evidence it
The plan as it stood at the start of each quarter, and the bank balance at the end.
Applies to
Every company.
KPI reporting
A board pack every month, on time
The bar at Series A
Floor
Issued on time in at least ten of the last twelve months
SVV's own figure. No source gives a figure, and no investor names it as a test. Y Combinator's talk on metrics supports a regular written update to investors with revenue, burn rate and runway at the top. It says nothing on how often or how promptly a board pack should go out (Y Combinator).
How it is read, and what counts
How it is read from a deck
A board pack is the set of figures and commentary sent to the directors before a board meeting. Look for how often the pack goes out, or an extract from one. A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck or its appendix shows an extract from a monthly board pack, or lists the dates the last twelve were issued, with at least ten on time.
Claimed
The deck says a monthly board pack is produced, and shows nothing from it.
Not met
The deck says packs go out less often than monthly, or that more than two of the last twelve were late or missed.
Unknown
The deck is silent.
What would evidence it
Twelve months of board packs, with the date each was sent.
Applies to
Every company.
The KPIs come straight from the CRM and the finance system
The bar at Series A
Floor
Drawn from the CRM and the finance system, not kept by hand in a spreadsheet
SVV's judgementNothing is published as a test. Y Combinator advises one agreed definition of each metric and one central way of measuring it (Y Combinator).
How it is read, and what counts
How it is read from a deck
Look for where the numbers come from: a dashboard fed by the CRM and the accounts, or a spreadsheet kept by hand. A CRM is the record of every prospect and deal. For a self-serve product the systems are the billing system and the product's own analytics. A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck names the systems its KPIs are drawn from, or shows a dashboard with its source.
Claimed
The deck says the numbers come from the company's systems, and names none.
Not met
The deck says the KPIs are kept by hand in a spreadsheet.
Unknown
The deck is silent.
What would evidence it
A walk through the dashboard, from one number back to the record it comes from.
Applies to
Every company.
Sales system
The revenue forecast has been close to the result
The bar at Series A
Floor
Within 20% of the result in each of the last two quarters
SVV's own figure. No source gives a figure for forecast accuracy, and no investor names it as a test at Series A. Nothing published supports the 20%. It is SVV's own measure of what it asks of this pillar: "Monthly numbers a board trusts" (Sure Valley Ventures 2026).
How it is read, and what counts
How it is read from a deck
The financial slide or the appendix. Look for what was forecast at the start of each of the last two quarters, set beside what was achieved.
Shown
The deck shows forecast and actual new revenue for each of the last two quarters, and the gap is 20% or less in both.
Claimed
The deck says the company hits its forecasts or is on plan, with no forecast and actual figures.
Not met
The deck's own figures show a gap of more than 20% in either of the last two quarters.
Unknown
The deck is silent.
What would evidence it
The forecast as it was written at the start of each quarter, and the new revenue won by the end.
Applies to
Every company.
The pipeline is reviewed on a fixed rhythmSold by a sales team
The bar at Series A
No pass mark The report states how often the pipeline is reviewed, where the deck says, and applies no pass mark.
A weekly pipeline review is sometimes asked for. No source gives a frequency, so there is no pass mark for one.
How it is read, and what counts
How it is read from a deck
Look for how often the open deals are reviewed and by whom. A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck says how often the pipeline is reviewed and shows something from a review, such as a dated report.
Claimed
The deck says the pipeline is reviewed regularly.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
The recurring diary entry and recent review notes.
Applies to
Only companies that sell through a sales team.
The way the company sells is written down so that a new hire can follow itSold by a sales team
The bar at Series A
Floor
A written sales playbook
SVV's judgementNo investor publishes it as a test. The direction is supported: Bessemer puts playbook building between $100k and $1M of ARR (Bessemer Venture Partners 2025), and the invented company First Round uses to show its top level has "repeatable playbooks" (First Round 2024).
How it is read, and what counts
How it is read from a deck
The slide on how the company sells. A playbook is the written guide a new salesperson is trained from: who to approach, what to say, the stages, the prices and the common objections.
Shown
The deck shows an extract from the playbook, or lists what it covers.
Claimed
The deck says there is a documented or repeatable playbook, and shows nothing of it.
Not met
The deck says the sales approach is not written down.
Unknown
The deck is silent.
What would evidence it
The playbook.
Applies to
Only companies that sell through a sales team.
People systems
The staff option pool is set up, and granted as planned
The bar at Series A
Floor
Options granted as the plan said, with none promised and not yet granted
SVV's judgementY Combinator's diligence checklist tells founders to make any pending or promised equity grants before the Series A process starts (Y Combinator).
How it is read, and what counts
How it is read from a deck
The team slide, the ask slide or the cap table. Look for the size of the pool and, at Series A, how much of it has been granted against the plan.
Shown
The deck gives the size of the pool as a share of the company. At Series A it also gives how much has been granted, and to how many people.
Claimed
The deck says there is an option pool, and gives no size.
Not met
The deck says there is no pool, or at Series A that options promised to staff have not been granted.
Unknown
The deck is silent.
What would evidence it
The option scheme documents and the register of grants.
Applies to
Every company.
How the company hires and brings people in is written down
The bar at Series A
Floor
A written hiring process and a written onboarding plan
SVV's judgementNo investor publishes it as a test. Y Combinator's diligence checklist asks for the employee handbook and the standard offer letter at closing (Y Combinator).
How it is read, and what counts
How it is read from a deck
A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck or its appendix shows the written hiring process or the onboarding plan, or an extract from either.
Claimed
The deck says hiring and onboarding are documented, and shows neither.
Not met
The deck says neither is written down.
Unknown
The deck is silent.
What would evidence it
The hiring process, the onboarding plan, the handbook and the standard offer letter.
Applies to
Every company.
Staff performance reviews are running
The bar at Series A
Floor
One round of reviews completed, and the next one dated
SVV's judgementA review cycle that is running is written here as one round done and the next dated, so that two readers agree. Nothing is published, and no investor names it as a test.
How it is read, and what counts
How it is read from a deck
A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck or its appendix gives the date of the last round of reviews and the date of the next.
Claimed
The deck says reviews happen, with no dates.
Not met
The deck says no round of reviews has been run.
Unknown
The deck is silent.
What would evidence it
The review calendar and the completed reviews.
Applies to
Every company.
How long it takes to fill a key role
The bar at Series A
No pass mark The report states the time to hire as the deck gives it, and applies no pass mark.
A figure sometimes quoted is 60 days or less for key roles. No source gives a figure for time to hire, so there is no pass mark for it.
How it is read, and what counts
How it is read from a deck
The team slide. Look for the time from opening a role to the person accepting, for the senior roles and the ones that are hard to fill.
Shown
The deck gives the time to hire for named roles, with the dates.
Claimed
The deck says the company hires quickly, with no dates.
Not met
Not given for this part. It has no pass mark, so a deck cannot fall short of it.
Unknown
The deck is silent.
What would evidence it
For each key role, the date it opened and the date the offer was accepted.
Applies to
Every company.
Red flags
Any one of these caps the pillar at level 2.
The deck says there is no cash model, or its own figures give less than six months of runway and it does not say so. Raised only where the deck says or shows this. A silent deck is not checked.
The numbers in the deck do not agree with the company's own records: its CRM, billing system or accounts. Two slides that disagree are a fault of the deck, read under the check on figures agreeing, and do not raise this flag. From a deck alone it is not checked.
Staff without employment or IP agreements. Raised only where the deck says or shows this. A silent deck is not checked.
Trust
Will an enterprise buyer sign?
Security posture
SOC 2 or ISO 27001: held, or an audit under way with a date
The bar at Series A
Floor
SOC 2 Type I or ISO 27001 held, or an audit under way with a dated plan
SVV's judgement on timing. No investor framework sets a stage for certification, and Y Combinator's Series A diligence checklist has no section on security. The direction is supported, because buyers ask early. Unusual Ventures: "your customers will require it of you if you’re processing or storing any amount of their data, even if you’re only in the design partner phase" (Unusual Ventures 2023).
How it is read, and what counts
How it is read from a deck
The trust slide. SOC 2 and ISO 27001 are independent audits of how a company protects its customers' data. Type I is the first stage of SOC 2. Look for the certificate or the report and its date, or for the name of the auditor and the date the audit is due to finish.
Shown
The deck gives the date of the certificate or report, or names the auditor and the date the audit will finish.
Claimed
The deck says "SOC 2 compliant", "SOC 2 ready" or "in progress", with no auditor and no date.
Not met
The deck says neither is held and no audit has started.
Unknown
The deck is silent.
What would evidence it
The certificate or audit report, or the auditor's engagement letter with its dates.
Applies to
Every company.
Where the figures are from
In the UK the government's minimum standard is Cyber Essentials. The National Cyber Security Centre says a growing number of organisations require suppliers to hold it before they can bid for work (National Cyber Security Centre). It covers five technical controls and does not stand in for SOC 2 or ISO 27001 in this part. Where a deck shows it, the report lists it as a strength.
An outside penetration test in the last twelve months, with the serious findings fixed
The bar at Series A
Floor
Tested in the last twelve months, with the critical findings fixed
SVV's judgementNo investor publishes a security or data test for any stage, and none names one as a reason to invest. The direction is supported by what buyers ask. Unusual Ventures says customers will require compliance certification of a supplier that processes or stores their data, even at the design partner stage (Unusual Ventures 2023).
How it is read, and what counts
How it is read from a deck
The trust slide. A penetration test is an attack on the product by a hired security firm, to find its weaknesses. Look for who did it, when, and what happened to the critical findings.
Shown
The deck names the firm, gives a date within the last twelve months, and says the critical findings are fixed.
Claimed
The deck says the product is penetration tested, with no firm and no date.
Not met
The deck says no test has been done in the last twelve months, or that critical findings are still open.
Unknown
The deck is silent.
What would evidence it
The test report and the record of the fixes.
Applies to
Every company.
Data and privacy
A data processing agreement for customers
The bar at Series A
Floor
A signed DPA with every enterprise customer
SVV's judgementNo investor publishes a security or data test for any stage, and none names one as a reason to invest. The direction is supported by what buyers ask. Unusual Ventures says customers will require compliance certification of a supplier that processes or stores their data, even at the design partner stage (Unusual Ventures 2023).
How it is read, and what counts
How it is read from a deck
The trust slide. A DPA, or data processing agreement, is the contract that says how a supplier may handle a customer's personal data. At seed, look for a standard template ready to sign. At Series A, look for one signed with every enterprise customer. For a self-serve product the DPA is part of the terms every customer accepts.
Shown
At seed: the deck says a DPA template exists, and shows it or gives the link. At Series A: the deck gives the number of enterprise customers and the number with a signed DPA, and they match. For a self-serve product, at either round: the deck gives the link to the DPA in its published terms.
Claimed
The deck says DPAs are "in place" or "available", with no template shown and no counts.
Not met
The deck says there is no template, or at Series A that an enterprise customer has no signed DPA.
Unknown
The deck is silent.
What would evidence it
The template. At Series A, the signed DPAs set against the customer list.
Applies to
Every company.
A data protection impact assessment, where the law requires one
The bar at Series A
Floor
Done where required, or the reason none is required written down
SVV's judgementNo investor publishes a security or data test for any stage, and none names one as a reason to invest. The direction is supported by what buyers ask. Unusual Ventures says customers will require compliance certification of a supplier that processes or stores their data, even at the design partner stage (Unusual Ventures 2023).
Whether one is required is a legal question. The report does not decide it. It states what the deck says.
How it is read, and what counts
How it is read from a deck
The trust slide. A DPIA, or data protection impact assessment, is a written assessment of the risks a use of personal data creates for the people it is about. Look for a statement that one has been done, or the reason none is needed.
Shown
The deck says a DPIA has been done and what it covers, or says why none is required.
Claimed
The deck says the company is "GDPR compliant", and does not mention a DPIA.
Not met
The deck says one is required and has not been done.
Unknown
The deck is silent.
What would evidence it
The DPIA, or the written note of why none is needed.
Applies to
Every company.
A published list of the other companies that handle customer data
The bar at Series A
Floor
A published list of sub-processors
SVV's judgementNo investor publishes a security or data test for any stage, and none names one as a reason to invest. The direction is supported by what buyers ask. Unusual Ventures says customers will require compliance certification of a supplier that processes or stores their data, even at the design partner stage (Unusual Ventures 2023).
How it is read, and what counts
How it is read from a deck
The trust slide. A sub-processor is another company the product passes customer data to, such as a cloud host or a model provider. Look for the list, or a link to it.
Shown
The deck shows the list or gives the link.
Claimed
The deck says a list is available on request.
Not met
The deck says no list is published.
Unknown
The deck is silent.
What would evidence it
The page or document that lists them.
Applies to
Every company.
For AI companies
For an AI product the list should name every model provider that customer data is sent to.
A rule for how long customer data is kept, and proof that it is followed
The bar at Series A
Floor
A retention policy that is enforced
SVV's judgementNo investor publishes a security or data test for any stage, and none names one as a reason to invest. The direction is supported by what buyers ask. Unusual Ventures says customers will require compliance certification of a supplier that processes or stores their data, even at the design partner stage (Unusual Ventures 2023).
How it is read, and what counts
How it is read from a deck
Look on the trust slide for how long customer data is kept and how it is deleted. A deck rarely says this. If the deck is silent the part is Unknown, and it becomes one of the questions to the founder.
Shown
The deck states how long customer data is kept and how deletion is carried out, such as by an automatic job.
Claimed
The deck says there is a retention policy, and gives no period.
Not met
The deck says data is kept without limit, or that the policy is not applied.
Unknown
The deck is silent.
What would evidence it
The policy, and the logs of the deletion job.
Applies to
Every company.
AI governance
A written AI governance policyAI only
The bar at Series A
Floor
An AI governance policy
SVV's judgementNothing is published on AI governance by stage. The direction is supported by what buyers ask. Bessemer expects AI vendors to "surface evidence of effectiveness before purchase", and says evaluations become "a core requirement for procurement and governance" (Bessemer Venture Partners 2025).
How it is read, and what counts
How it is read from a deck
The trust slide. An AI governance policy says who is responsible for the AI in the product, how changes to models are approved, how it is tested, and what happens when it goes wrong.
Shown
The deck shows the policy or an extract, or names it with its date and the person who owns it.
Claimed
The deck says the company practises "responsible AI" or has governance in place, and names no policy.
Not met
The deck says there is no written policy.
Unknown
The deck is silent.
What would evidence it
The policy.
Applies to
Only AI companies.
The AI is watched in production for slipping quality and for incidentsAI only
The bar at Series A
Floor
Drift monitoring and incident monitoring in place
SVV's judgementNothing is published on AI governance by stage. The direction is supported by what buyers ask. Bessemer expects AI vendors to "surface evidence of effectiveness before purchase", and says evaluations become "a core requirement for procurement and governance" (Bessemer Venture Partners 2025).
For high-risk systems the Act will expect providers to monitor a system once it is on the market, and to report serious incidents (European Commission 2026). Those duties apply from 2 December 2027.
How it is read, and what counts
How it is read from a deck
The trust slide or the slide on the AI. Drift is the slow fall in a model's accuracy as the world or the model changes. Look for what is measured on live work, how often, and the record of incidents.
Shown
The deck shows a measure of quality on live work over time, or describes the monitoring and gives the number of incidents in a stated period.
Claimed
The deck says the models are monitored, with no measure and no period.
Not met
The deck says quality in production is not measured, or that there is no record of incidents.
Unknown
The deck is silent.
What would evidence it
The monitoring dashboard and the incident log.
Applies to
Only AI companies.
A record is kept of what the AI producedAI only
The bar at Series A
Floor
An audit trail of outputs
SVV's judgementNothing is published on AI governance by stage. The direction is supported by what buyers ask. Bessemer expects AI vendors to "surface evidence of effectiveness before purchase", and says evaluations become "a core requirement for procurement and governance" (Bessemer Venture Partners 2025).
For high-risk systems the Act will require "logging of activity to ensure traceability of results" from 2 December 2027 (European Commission 2026).
How it is read, and what counts
How it is read from a deck
The trust slide or the product slide. Look for whether each output is logged with its input, the model version and the time, so that a result can be traced later.
Shown
The deck says outputs are logged, and says what is recorded with each.
Claimed
The deck says the product is "auditable" or "traceable", and does not say what is logged.
Not met
The deck says outputs are not logged.
Unknown
The deck is silent.
What would evidence it
A sample of the log.
Applies to
Only AI companies.
Customers are told, in writing, what the AI does and where its limits areAI only
The bar at Series A
Floor
Transparency documentation written for customers
SVV's judgementNothing is published on AI governance by stage. The direction is supported by what buyers ask. Bessemer expects AI vendors to "surface evidence of effectiveness before purchase", and says evaluations become "a core requirement for procurement and governance" (Bessemer Venture Partners 2025).
Part of this is now law for some products. The Act's transparency rules have applied since 2 August 2026: people must be told when they are dealing with an AI system such as a chatbot, and AI-generated content must be identifiable. Systems already on the market before that date have until 2 December 2026 to mark generated content (European Commission 2026; Future of Life Institute 2026).
How it is read, and what counts
How it is read from a deck
The trust slide. Look for a document or page written for customers: what the AI is used for, what it should not be relied on for, and how people are told they are dealing with an AI.
Shown
The deck shows the document or page, or gives the link.
Claimed
The deck says the company is "transparent" about its AI, and shows nothing written for customers.
Not met
The deck says nothing of the kind exists.
Unknown
The deck is silent.
What would evidence it
The document or page written for customers.
Applies to
Only AI companies.
Procurement readiness
Enterprise buyers' procurement reviews passedSold by a sales team
The bar at Series A
Floor
Three reviews passed
SVV's own figure. Nothing is published on a number. The direction is supported: a16z's survey of 100 chief information officers found that buying AI now follows the checks of ordinary enterprise software buying (Andreessen Horowitz 2025).
How it is read, and what counts
How it is read from a deck
The trust slide or the customer slide. A procurement review is a buyer's check of a supplier's security, legal terms and finances before it will sign. Look for how many have been passed, and with whom.
Shown
The deck gives the number passed, three or more, and names the buyers or says what kind they are.
Claimed
The deck says the company sells to enterprises, or shows enterprise logos, and gives no count of reviews passed.
Not met
The deck's own count is under three.
Unknown
The deck is silent.
What would evidence it
The procurement history: the buyer, the date the review started and the date it was passed.
Applies to
Only companies that sell through a sales team.
How long those reviews tookSold by a sales team
The bar at Series A
Floor
A median of 45 days or less
SVV's own figure. Nothing is published on how long a procurement review should take. SVV's aim for the pillar is that a company passes procurement in weeks, not months.
How it is read, and what counts
How it is read from a deck
The trust slide. Look for the median time from the start of a procurement review to passing it. The median is the middle one when they are put in order.
Shown
The deck gives the median time, 45 days or less, and how many reviews it is based on.
Claimed
The deck says procurement is fast, or gives the sales cycle only.
Not met
The deck's own median is more than 45 days.
Unknown
The deck is silent.
What would evidence it
The same procurement history.
Applies to
Only companies that sell through a sales team.
Cyber insurance in placeSold by a sales team
The bar at Series A
Floor
A cyber insurance policy in force
SVV's judgementNo investor publishes it as a test. Y Combinator's diligence checklist asks for any insurance policies the company holds (Y Combinator).
How it is read, and what counts
How it is read from a deck
The trust slide. Cyber insurance covers the cost of a data breach or an attack. The trust pack lists insurance among what a buyer asks to see.
Shown
The deck names the insurer or gives the level of cover.
Claimed
The deck says the company is insured, with no insurer and no level of cover.
Not met
The deck says there is no cyber cover.
Unknown
The deck is silent.
What would evidence it
The policy schedule.
Applies to
Only companies that sell through a sales team.
Red flags
Any one of these caps the pillar at level 2.
A breach or unresolved security incident
Models are trained on customer data with no right in the contract to do so. Raised only where the deck says or shows this. A silent deck is not checked.AI only
Unable to complete a standard security questionnaireSold by a sales team