It has never been easier to make a young company look successful. Artificial intelligence allows a small team to build a prototype, launch a website, automate marketing and attract thousands of curious users within weeks. Rising traffic, early revenue and a polished investor presentation can create an impressive picture long before the company has proved that customers genuinely need its product..
As an investor, I distinguish between a startup moving because it is being pushed by capital, publicity and novelty, and one beginning to generate its own momentum. In a viable company, customers return, some buy more, acquisition becomes repeatable, economics improve and the team learns faster than the market changes..
This distinction is particularly important in AI. Bessemer Venture Partners divides high-growth AI companies into “Supernovas” and “Shooting Stars”. The Supernovas it studied reached an average of approximately $40 million in annual recurring revenue during their first year of commercialisation, but generated gross margins of only around 25%.
Shooting Stars grew more gradually, with stronger retention and average margins of approximately 60%.. The fastest-growing company is not necessarily the healthiest one. Revenue can be subsidised, users can be attracted by novelty and contracts can be cancelled.
No single metric proves viability. When I evaluate a startup, I apply eight interconnected tests.. The problem test.
My first question is not whether the product is impressive. It is what happens to the customer if the product does not exist.. Is the customer already losing money or time?
Is the company employing people, hiring consultants or combining several inadequate tools to manage the problem? Who experiences the pain, and who controls the budget?. Hypothetical enthusiasm is weak evidence.
When founders ask potential customers whether they would use a product, the answer is often yes because saying yes costs nothing. I am more interested in existing behaviour: how is the problem being solved today, how much does that solution cost and what happens when it fails?. Not every successful company addresses an obvious emergency.
Some challenge an inconvenience customers have accepted, while others create possibilities customers have never imagined. But founders must understand which journey they are taking. If the market does not yet recognise the problem, the company will need more time and capital to educate it..
A difficult market does not necessarily concern me. A founder who does not understand why the market is difficult does.. The customer-pull test.
The second test begins after the customer arrives.. Registrations, downloads and trials tell me that a product attracted attention. Retention tells me whether it created value.
Do customers continue using it? Do they renew, expand their contracts or introduce it to other teams? Would they notice immediately if it disappeared?.
ChartMogul recently analysed approximately 3,500 software companies. Among businesses with at least $250,000 in ARR, median net revenue retention was 82% for B2B SaaS but only 48% for AI-native companies. Some early-stage startups could still grow by more than 200% while retaining less than 40% of their revenue base.
They appeared to be expanding while constantly replacing customers who had left.. This is why I prefer cohort data to cumulative totals. Total registrations almost always rise.
A cohort shows how many customers who arrived in a particular month remain active six or twelve months later.. Not all revenue is equally valuable either. A customer experimenting with an AI application under a cancellable contract should not be treated like one that has embedded the product in an essential workflow..
The earliest evidence of product–market fit is not a founder declaring it. It is a pattern of customer behaviour that becomes difficult to explain as curiosity or chance.. The economic test.
A product can be useful and popular while remaining economically unhealthy.. Founders should understand customer acquisition cost, lifetime value, payback period, gross margin, burn rate and runway. But repeating the terminology is not enough.
I want to know whether the assumptions behind those numbers are credible.. A startup can understate CAC by excluding sales expenses, exaggerate lifetime value using retention history of only a few months or report attractive margins while ignoring implementation work and AI inference costs..
The relevant question is simple: does each additional customer strengthen the business or create another expense that must be subsidised?. For an AI company, viability may depend on model costs falling faster than prices. For an enterprise platform, it may depend on turning bespoke implementation into a repeatable process.
For a consumer subscription, it may depend on reducing churn before the cost of replacing departing users overwhelms revenue.. Runway is also more than the number of months before cash runs out. A founder should know which uncertainty must be resolved during that period: retention, pricing, distribution, implementation cost or regulatory approval..
Capital without a defined learning objective is not runway. It is a longer countdown.. The distribution test.
Many founders can explain how they will build a product but not how they will repeatedly find customers.. At an early stage, I prefer founders to participate directly in sales. The first sales conversations are part of product development.
They reveal which benefit matters, who makes the decision, what creates hesitation and why a trial does or does not become a real deployment.. A credible distribution strategy identifies the ideal customer, the buyer, the acquisition channel, the sales cycle and the event that makes the purchase urgent.
It must also show that the process can eventually be repeated without heroic involvement from the founder.. I become cautious when someone says, “The product is excellent; we only need marketing.” Marketing can accelerate existing demand, but it cannot permanently replace it..
The same applies to virality. A product does not become viral because a presentation says it will. Genuine virality exists when normal use exposes the product to new users and gives them a reason to join..
Technology creates a product. Distribution turns it into a business.. The learning-speed test.
Speed matters, but the correct measure is not the number of features released. It is how quickly a team converts an assumption into an experiment, collects real evidence and changes its decisions.. An MVP should therefore be the smallest credible product capable of answering an important question.
It is not simply a cheaper version of the final product.. Paul Graham has long argued that founders should release early and improve through user reactions. He also distinguishes determination from stubbornness.
A strong founder remains committed to the problem while remaining flexible about the solution.. When I meet a team, I ask what it believed six months ago that it no longer believes today. Which customer segment disappointed them?
Which feature was removed? Which evidence caused a change of direction?. Founders who claim never to have been wrong may simply have avoided testing their most important assumptions..
Constantly pivoting can indicate a lack of conviction. Refusing to pivot despite contrary evidence is equally dangerous. The quality investors seek is disciplined adaptability..
The fastest startup is the one reducing uncertainty most efficiently.. The defensibility test. When software becomes easier to build, an undifferentiated product becomes easier to replace..
This is now a central issue for AI startups. A company may build an attractive interface around a third-party model and gain users quickly. But what happens when the model provider introduces the same feature?.
I ask founders what becomes stronger with every new customer. The answer may involve proprietary data, network effects, deep workflow integration, regulatory approvals, distribution, trusted relationships or accumulated knowledge of a specialised industry.. Defensibility does not require an early-stage company to possess an impregnable moat.
It requires a credible mechanism through which success makes the product progressively harder to copy.. Terms such as “data advantage” and “switching costs” should also be demonstrated rather than asserted. A future data moat is meaningless if the company has no realistic way to collect unique information.
Switching costs are real only when customers remain despite credible alternatives.. The question is not simply whether somebody can copy the product today. It is whether copying it becomes more difficult as the company grows..
The founder-and-team test. At the earliest stage, the founders are the company. The product, customer segment and business model may all change, so investors must decide whether the team can continue making intelligent decisions as the original plan meets reality..
I look for founder–market fit: why these particular people understand this particular problem unusually well. That advantage may come from years inside an industry, personal experience, technical insight or exceptional proximity to an emerging market.. Age tells me very little.
A young founder may understand a new market better than an experienced executive whose assumptions belong to a previous technological era. What matters is the accuracy and speed of learning.. I also examine how founders work together.
Are their skills complementary? Are roles, equity and decision rights clear? How do they resolve disagreement?
Can they attract talented people? Will they remain committed when the company stops being exciting?. Most importantly, are they intellectually honest?.
I would rather hear a founder explain a weak metric precisely than watch one hide it behind confidence. Strong founders know what is not working, why it matters, what they are testing and which result would force them to change course.. Investors do not expect founders to know every answer.
They expect them to recognise the questions they cannot afford to ignore.. The capital-discipline test. Fundraising is not the objective.
Capital should purchase a specific reduction in risk.. A pre-seed round may demonstrate that the technology works. A seed round may establish customer retention.
A Series A may transform founder-led sales into a repeatable commercial system.. In each case, the founder should explain what the investment will prove. “We need 18 months of runway” is incomplete.
Runway to which milestone?. Capital discipline also includes dilution. Carta’s analysis of more than 45,000 startups found that the median founding team retains approximately 56.2% of its company after a seed round, 36.1% after Series A and 23% after Series B..
Avoiding dilution at any cost can deprive a company of necessary capital. Raising too much can encourage premature hiring, conceal poor economics and impose expectations inconsistent with the real market.. I pay close attention to what happens after a company raises money.
Does the team preserve its discipline, or begin treating the investment as proof that the difficult questions have been answered?. Financing confirms that an investor believes future success is possible. It does not confirm that the company has already achieved it..
The standard changes with the stage. These tests carry different weight at different stages.. At pre-seed, I focus primarily on the founders, their understanding of the problem, the originality of their insight, the potential market and their speed of learning..
At seed, customer behaviour should begin confirming the story. Are people using the product, paying for it and returning? Is an identifiable acquisition channel emerging?.
By Series A, promise is no longer sufficient. The company should demonstrate retention, repeatable distribution, improving economics and a credible path towards scale.. Metrics must also be interpreted according to the business model.
A long sales cycle may be a warning sign for a small-business application but entirely normal in defence or healthcare. Low margins may reveal a structurally weak model or a temporary investment in implementation.. Context matters.
But context should not become an excuse for the absence of evidence.. What investors are really looking for. A viable startup is a connected system..
The problem is important enough to create demand. The product solves enough of it to bring customers back. Retention makes acquisition economically worthwhile.
Distribution turns isolated success into a repeatable process. The team learns fast enough to keep improving. The product becomes more defensible as adoption grows.
Capital accelerates the system without being required to conceal its weaknesses.. When these elements reinforce one another, the company begins generating its own momentum. When they contradict one another, growth may be masking fragility..
A startup with rising revenue but collapsing retention is not necessarily healthy. A startup with millions of users but no economical way to serve them is not necessarily viable. A company with brilliant technology but no distribution is not yet a business.
And a profitable company operating in a limited market may be an excellent enterprise without being suitable for venture capital.. The most useful question for a founder is therefore not, “How can I make this company look impressive?” It is: “Which part of the system would fail first if capital became scarce, customer curiosity faded or a powerful competitor entered tomorrow?”.
A company is not alive simply because it is moving. It is alive when its customers, economics, team and learning processes have begun to sustain one another — and when new investment accelerates a real business rather than financing the appearance of one.. Featured image credit.
Tags: startupstrends
