Valuing Companies: The Most Valuable Part Does Not Fit the Model

I have been watching company valuation for ten years from both sides of the negotiating table. Value has always been calculated from the past and from a projection of the future. That model feels outdated to me, but we have no replacement for it. So in this article I try to describe what I look for instead.

As an entrepreneur and investor, I have been through several negotiations about selling part or all of a company. I have been on both sides: as the seller and as the buyer. And every time, value stood on two legs: past results and a projection of future earnings.

It worked according to a simple key. Very young companies were valued mainly on the promise of the future, because they had nothing else. The older the company, the firmer the foundation the past provided. From the history of revenues, margins and customer behaviour, much could be predicted.

But today’s era is changing entire industries and demolishing this proven approach.

What am I talking about? I am not talking about companies in sectors that require large investments in physical technology: machines, production halls, vehicle fleets. There, capital still creates a barrier and the past says something about the future. I am talking about companies that provide services, develop software and hardware. About companies in dynamic sectors where the rules change faster than a single financial model can run its course.

And there, existing company valuation models feel outdated to me.

Company Valuation Rests on Projections That Lie to Themselves

A company that projects its product will be the same or similar over the next three to five years, that the pricing model will stay unchanged and everything will simply grow, is lying to itself. And the buyer cannot take it seriously.

I am not alone in this and it is not an armchair opinion. Look at what happened to the price of software companies that were issuing exactly these projections just two years ago.

According to Multiples, the median for publicly traded horizontal SaaS in August 2026 sits around 2.2 times revenue. Not the ten times the market got used to in 2020 and 2021. Just over two. And the same analysis says something every software company owner should hear: the market today does not value by market size, but by whether a company uses AI or is dying because of it. The spread between categories is enormous. While DevOps sells at 8.7 times revenue, AdTech goes for 0.9.

The reports started calling this collapse the SaaSpocalypse, and according to Forrester it wrote off value from software companies in the trillions of dollars. It did not fall because those companies suddenly stopped growing. The market simply stopped believing the projection. Exactly the projection I am talking about.

What the Head of Microsoft Says About His Own Category

Satya Nadella put it most bluntly back at the end of 2024, long before the market priced it in. In his view, business applications as we know them will collapse in the era of agents, because at their core they are just databases with a layer of business logic. And the agent will take that logic. The head of Microsoft said this about a category Microsoft itself operates in.

What companies actually charge for is changing too. The shift from paying per user to paying per outcome is no longer theory: Intercom charges 0.99 dollars per resolved case, Salesforce launched Agentforce at two dollars per conversation. When the unit a company invoices changes within three years, what does a projection built on licence counts mean?

This is my first objection to today’s company valuation practice. We do not ask hard enough whether the model the projection is built on will even exist in three years.

Large Companies Will Fall Faster, Small Ones Will Grow Up Faster

I recently discussed this area with Senta Čermáková and we agreed on one thing: we will live through a period where large companies fail much faster and young ones become mid-sized very quickly.

AI will help them do it. Scaling companies will happen through systems, not people, and it will be significantly less capital intensive.

The numbers back this up from both sides. According to a Huron report, the average time a company stays in the S&P 500 index shortened from thirty-three years in 1965 to twenty in 1990 and is heading toward fifteen. Aswath Damodaran, professor of valuation at NYU, puts it even more harshly about technology companies: they age like dogs. They grow from zero to something incredibly fast, do not stay mature for long, and then decline. And he adds a sentence I would put on the wall of every owner: the most value destroyed in the world comes from companies that refuse to act their age.

A Hundred Million ARR With Fifty People

On the small side, exactly what Senta and I discussed is happening. Cursor went from its first million to a hundred million dollars in annual revenue in roughly a year, with just a few dozen people. Lovable did the same with forty-five people. Ten years ago, reaching a hundred million ARR would have required hundreds of employees and several funding rounds.

And this does not apply only to startups. Klarna shrank its team from around five thousand people to just under three thousand, while revenue per employee grew from three hundred thousand dollars to 1.3 million. Shopify went further and made it a rule: before a team asks for new people, it must document why AI cannot handle it. OpenAI CEO Sam Altman had been betting friends for years about a one-person billion-dollar company and recently announced he won the bet.

All of this places enormous demands on the adaptability and flexibility of people, management and owners. If large companies fail to manage the transformation in time and at speed, their positions will be shaken at the foundations and some will not survive.

Conversely, small companies where adaptability, innovation and flexibility are part of the cultural DNA will be highly valued. Because in an era changing this fast, they are the ones who will survive.

A Strategy That Talks About Neither Products Nor EBITDA

I recently had a conversation with a friend, a very capable CEO of a company with more than two hundred IT specialists. He does not build a two to three year strategy about what his company will do, at what volume and with what profitability. He does not know what will be true that far ahead. They work more on staying relevant in the market and going through the changes together with their clients.

Do you hear that? There is no talk of products, services, revenues and EBITDA. The talk is about relevance, which comes from adaptability, innovation and flexibility.

It is not a new idea. Years ago, Jeff Bezos spoke about irrelevance as the phase that comes right after stagnation and long before death. The difference is the pace. Bezos warned about a slow slide. My friend builds his company on the assumption that the slide can come within a single season.

I will add one objection of my own, because otherwise this would sound too neat. Adaptability without a firm intent is not adaptability, it is floundering. A company that changes direction every quarter because it is carried by the latest trend is not flexible, it is rudderless. The goal must hold. The path changes.

Buyers Already Pay for Adaptability. They Just Do Not Call It That

Here I come to what I find most interesting about the whole thing.

I have seen several times that a large strategic partner bought a smaller company not for its product or profitability, but for the capabilities of its team. The buyer had a plan in which those capabilities would generate far more value than the original product or service ever could.

For a long time I thought it was an exception. It is not. Over the past two years it has become one of the most expensive transaction categories in the world.

Microsoft paid roughly 650 million dollars for Inflection and it was not about a chatbot, it was about Mustafa Suleyman and his research team. Google paid 2.7 billion for Character.AI for a non-exclusive licence and two founders. And a year later, 2.4 billion dollars for Windsurf CEO Varun Mohan and the core of his research team joining DeepMind. In total, big tech spent over forty billion dollars on this type of deal in 2024 and 2025. That is more than the entire previous history of acqui-hires combined.

Notice what was not valued in those deals. Not one of them can be justified by an EBITDA multiple. Not one rests on a product revenue projection. What was being bought was the ability to build the next thing.

The Earn-Out Is the Most Honest Sentence in the Contract

And then there is a subtler trace I notice in ordinary mid-market transactions. Buyers already pay for adaptability, they just do not call it that and they do not pay for it in the multiple, but in the structure of the deal.

Earn-outs, deferred payments, key people contractually tied in for years ahead. According to SRS Acquiom data, the share of transactions with an earn-out rose from 19 % in 2014 to 24 % in 2025, and among the smallest deals under 25 million dollars it reaches a full 35 %. When a buyer says you will get a third of the price if these specific people stay and the company meets these targets, what they are really saying is: I am not confident this company can adapt without me, without you, or without both.

It is the most honest sentence in the entire purchase agreement. And it is hidden in the clause about payment terms, not in the valuation.

Accounting Does Not See It and Company Valuation Cannot Capture It

There is a reason adaptability does not appear in the models: it has nowhere to be recorded. A company that spends a year teaching people to work with AI, rebuilds its processes, kills its own established product and builds a new one, will report worse results for that year. All that money went through as cost. Nothing stayed on the balance sheet. A company that did nothing for a year and simply milked its old contract base will report a better number. And in a standard company valuation model, it comes out looking better.

That is a systemic flaw. According to Ocean Tomo, intangible assets make up roughly 90 % of the market value of companies in the S&P 500 index. In 1975 it was 17 %. The market sees that value. Accounting largely fails to capture it, and a financial model derived from accounting cannot capture it either.

So we ask about what is measurable instead of what matters.

What the Hejnic Number Might Look Like

How do you capture this capability in company valuation? We investors will have to work it out. Perhaps a set of parameters and questions will emerge from which a coefficient can be derived. Maybe my colleagues and I will come up with the Hejnic number.

But whatever it ends up being called, I have a fairly clear idea of the property such an indicator must have to be useful: it must be readable before the outcome is known. In hindsight, everyone who survived looks adaptable. That is worthless.

Seven Questions I Would Ask

Here is what I would ask about. It is not a finished methodology, just a first draft I want someone smarter to tear apart.

1. The share of revenue from things that did not exist two years ago. 3M has used this indicator since 1988 under the name New Product Vitality Index and today most large companies with their own development track it in some form. For dynamic sectors I would shorten the window to twenty-four months.

2. How long it takes the company to decide. Not the quality of the decision, that only shows after the outcome. The time from we know there is a problem to it is decided and we are moving. This can be established from history and cannot be staged for a single meeting.

3. How many times the company killed something of its own that was still earning. The ability to let go of yesterday is rarer than the ability to launch something.

4. How much of the know-how sits in heads and how much in systems. A company where the process is documented, automated and transferable adapts. A company where everything rests on three people adapts only as fast as those three can retrain.

5. The trend in revenue per employee, not its absolute level. The direction of this curve over the past two years says more about how a company scales than an entire chapter on strategy.

6. Retention of key people, and whether the buyer can even name them. When nobody in due diligence can say which five people the company rests on, that in itself is a finding.

7. How many decisions are reversible. A company built only on long irrevocable commitments has less room to manoeuvre regardless of how good its culture is.

One last, uncomfortable question: how do you tell adaptability from luck? Evaluate the process, not the outcome. Translated into due diligence language: do not ask whether the company got the last big change right, but how it arrived at that decision, who knew about it first and what the company did with that information. Getting it right can be chance. How a company learned about the change and how quickly it responded is not.

Valuation models can do the work with data, but they cannot replace judgement. A model will not calculate this question for us. I cannot do it yet either. But I know that whoever answers this question before the others will buy significantly better than the rest of the market.