How to Evaluate a Business Before the Spreadsheet Finds Out

Knowing how to evaluate a business means studying what does not fit in any ratio: what the company sells, who pays for it, who can take that customer away, and how management talks when something breaks. It is not an add-on to the valuation. It is the instruction manual for its numbers, because a number in the accounts is the consequence of a business decision taken quarters earlier: a price rise, a change of sales model, a lost fight for shelf space.

The spreadsheet is the last to know.

Qualitative versus quantitative analysis#

Quantitative analysis asks what happened. Revenue grew 9%, operating margin fell two points, net debt equals two years of operating cash flow. Those are facts, and there is nothing to argue about. Qualitative analysis asks something else: why it happened, and what would have to occur for that picture to stop being true.

Because the same 9% describes two different businesses. One won customers it did not have before. The other raised the price on the same customers as last year and sold them fewer units.

This layer answers three questions no table asks. Does the customer have anywhere else to go? Does the company set its own price, or take it from the market? And when something goes wrong, does management say what exactly went wrong? If you are still assembling your own process, the order of the steps for a first company lives elsewhere; this layer adds to that order rather than replacing it.

They sold less and earned more#

In its fourth-quarter and full-year 2023 release PepsiCo reported organic revenue growth of 9.5%. The split is more interesting than the headline: effective net pricing contributed 13 points, and organic volume took away 3. Which means the company sold fewer cans and bags than the year before, and revenue rose only because the customer paid more for each one.

A screener sees healthy growth.

Qualitative analysis sees a question with a shelf life: how long does a customer keep paying extra for the logo when the store's own brand sits on the same shelf. In the fourth quarter alone volume fell 4% and pricing contributed only 9 points, so the price lever was weakening quarter by quarter. That sits in the same document as the cheerful 9.5%.

The difference between the two routes to growth is a difference in durability. Volume growth can be repeated next year, because the customer came back of their own accord. Price growth spends down a reserve of patience that appears on no balance sheet, and it runs out at the moment the gap to the private label gets too wide to ignore at the shelf. I do not know whether habits built over decades survive another round of increases. I do know the answer shows up in the volume line, not the revenue line.

Splitting revenue into price and units is the same move as following one dollar through a company: you watch where it comes in and where it stays.

The year falling revenue was good news#

Adobe shows the same mechanism from the other side. In fiscal 2013 the company stopped selling boxed licences and moved to a subscription. Its annual report for fiscal 2013 shows revenue of $4.06bn against $4.40bn a year earlier, a decline of 8%. The same document shows paid subscriptions rising from 0.3 million to 1.4 million.

The accounts showed a shrinking company in exactly the year the business became more repeatable. A licence is one large payment today. A subscription is twelve small ones spread across the year and twelve more in the next, so the bookkeeping recorded a change of schedule and called it a decline.

A model would have read that year as a company losing customers. It gained more than a million.

This example has a second edge, and I would rather not hide it. "It is only the move to subscriptions" is also the favourite explanation of companies whose sales are simply falling. One thing separates them: in a real transition, the falling revenue line comes with a counter that rises and can be checked on its own, whether that is paid contracts or recurring revenue. When revenue falls and nothing rises, the explanation is an explanation, not a fact.

"Strong brand" is five claims#

Sentences that sound like conclusions are usually bags of unexamined claims. "The brand is strong" says five things at once: customers recognise it, prefer it to others, pay more for it, come back, and would not switch quickly to a substitute. Each of the five has a different source and can be true on its own.

Unpack them and you usually find the first two hold and the third is starting to crack.

Recognition shows up in research the company does not control. The premium shows up in the price per unit, and in whether it rises faster than the cost of the input. Repeat business shows up in sales to the same customers, and resistance to substitutes only reveals itself when the substitute becomes clearly cheaper. Four different pieces of evidence, gathered on four different clocks.

The same goes for "the company has a moat", which hides questions about switching costs, patents, scale, regulation and what competitors are actually doing. Or "management is excellent", which mixes capital allocation, the way failures get described, the bonus scheme and plain execution. The exercise that sorts this out fastest is fitting one company on one page, the way I did with Costco, because a single page leaves nowhere to hide five claims under one adjective.

Kodak had everything except one answer#

Kodak is the case I come back to most, because it is the cleanest example of popularity without durability. In 1975 a Kodak engineer, Steve Sasson, built the world's first digital camera. Management patented the device and put it on a shelf, because digital photography threatened film margins, and film paid for the whole company. For three decades a "Kodak moment" meant a memory captured, until on 19 January 2012 the company filed for Chapter 11 protection.

Kodak did not run out of technology or recognition. It ran out of an answer to the question of whether the model survives a change in the category, and no ratio asks that question for you.

The mechanism is ordinary, which is why it keeps repeating. A company with a fat margin on an old product measures every new thing against that margin, so the new thing always looks worse and always loses the internal budget. A competitor with nothing to defend measures the same thing from zero. That is why "what would this company have to cannibalise to win the next decade" is one of the few questions that genuinely splits an industry in two.

It also runs the other way: a company can work perfectly well while the ground shifts underneath it, and then no spreadsheet will save it. Which is why the question about the category comes before the question about the margin.

Four lenses instead of one impression#

Treat the qualitative layer as four lenses you change on purpose. First comes the customer: who pays, why they pay, and what they have instead. Next the industry, meaning whether the category is growing, crowded, regulated, commoditised, or changing shape the way photography did in Kodak's day. Third, management: does it speak in specifics, does it repeat last year's sentences, and can it own a weak quarter.

The fourth is the dullest and it polices the other three. It asks about the source: whether a claim comes from a filing, a transcript, someone else's article, or thin air. The first three lenses produce sentences, and the fourth decides what those sentences weigh.

Say you are looking at a sports equipment maker growing sales 20% a year. Customer: does the same person come back every two years, or does each sale need a new one bought with advertising? Industry: is the category growing, or just moving between brands? Management: does the word "volume" appear in the results commentary, or only "a difficult environment"? Source: is all of this in the annual report, or in an investor deck where the charts start at the best year?

Four answers say more about that company than the 20% does. Each also has its own shelf life: a category shifts over years, a price list over quarters, and the tone of management can change inside a single earnings call.

Is qualitative analysis just subjective#

The objection is fair and deserves to be stated in full. Two reasonable people will read the same transcript and part ways on the CEO's credibility: one hears composure, the other hears evasion. A ratio has the advantage of coming out the same regardless of who is looking at it, and you cannot argue with it about tone.

That sounds right. The subject, though, is qualitative; the method does not have to be.

Whether a sentence in the risk factors changed from last year's filing is a matter of record. So is whether management repeated a promise from the previous earnings call or dropped it without comment. Count how much of the revenue growth came from price, compare that share with the three years before, and you have a third answer nobody can argue with. Credibility stays a judgement call, which is one item on the list rather than the whole layer.

The real risk sits elsewhere and is worse. The story starts to replace the business: the better the narrative, the easier it is to forgive a third weak quarter because "it is early days", and customer concentration quietly becomes a "strategic partnership". So every load-bearing claim, meaning one whose collapse changes the whole picture of the company, gets a source or an explicit hypothesis label. Without that, qualitative analysis is not a method, it is sympathy with footnotes.

Not every claim earns that rigour. Two or three carry the structure: customer concentration, regulatory risk, the direction of demand, pricing power, one specific promise from management. The rest do not need the same interrogation, and it is better not to try, because research that covers everything ends at the first company.

How to evaluate a business in five steps#

1. Write one sentence: "this company makes money when...". No ticker, no slogan, no share price.
2. Open the business description in the annual report and list the segments, the geographies and the customer types.
3. Read the risk factors and mark the sentences that changed since last year.
4. Check in management's commentary how much of the revenue growth was price and how much was volume.
5. Pick three load-bearing claims and label each one: sourced, inferred or assumed.

The order is not accidental. Other people's write-ups come last, because the first source you read sets the narrative for everything after it, and the author of any market column has already decided what that narrative is. The whole route, rather than this one stretch of it, is one sentence from the company checked in three independent places.

On one company the whole thing takes two evenings, and it does not end in a verdict. It ends with a list of sentences you know the origin of, and a shorter list of the ones you have not checked. The second list matters more, because it tells you where the structure is thinnest, and it is the one you keep on top before the next set of results.

A named gap is far less dangerous than a gap hidden under an adjective.

At Taufolio this layer lives in the Full report: the business, the quality of management and the valuation laid out across 25 points. Every claim carries a link to the document it came from, and The gist of the report sits at the top for anyone who wants the whole shape first. The report does the reading for you. It does not pick which three sentences are load-bearing, because that depends on why you opened this file at all. Have a look at how the report is built and set it against your own list of questions about the firm you know best. Questions on your list that the report does not raise are precisely the part worth checking yourself.