To know what you know, and what you do not know: that is knowledge. Confucius, Analects, Book II
The Light and the Key
Nasruddin, the wise fool of Middle Eastern folklore, is on his hands and knees under a streetlamp, patting the ground. A neighbour stops to help. “What have you lost?” “My key.” They search together for an hour. Nothing. Finally, the neighbour asks where, exactly, he dropped it. “In the house.” “Then why are we looking out here?” “Because the light is better here.”
The story has survived seven centuries because every generation recognises itself in it. Ours should recognise financial reporting. Our accounts are a streetlamp: precise, audited, comparable and lit. Everything inside their circle is measured to the cent. And the things that decide whether your business survives the next decade sit in the dark, a few metres away, unrecorded, because there is no accounting entry for them.
Consider four things the light of the streetlamp cannot reach. The cost of equity, because no invoice ever arrives for it. The future value of the business because tomorrow has no journal entry. The impending approach of failure, because companies file immaculate accounts until the day they collapse. And the intangible assets that now carry roughly ninety-two per cent of the market value of the S&P 500 yet scarcely appear on the balance sheet at all (Ocean Tomo, 2026). Four of the most consequential facts about any enterprise. None of them in the light.
Artificial intelligence has made this problem more dangerous before it makes it better. It reads a ledger faster than any human who ever lived, flagging at a scale no audit team could approach. It is a magnificent instrument for searching under the streetlamp. The key is still in the house.
The danger is no longer that managers lack information. It is that artificial intelligence gives them the comforting illusion that more information is the same thing as better judgement. It has improved our ability to analyse what can be measured and done almost nothing to improve our judgement about what cannot.
The Man Who Kept the Plan
It’s the things that I knew enough to do that I didn’t do. Warren Buffett
A student once asked Warren Buffett to name the worst investment he had ever made. Buffett looked at him and said: “How long do you have?” Then he said something stranger. The mistakes that cost him most do not show up anywhere. Not in the losses, not in the accounts. They are mistakes of omission, not commission. He has never lost that much on any single investment. What haunts him is different. Ten billion dollars’ worth, he reckoned, of things he knew enough to do and did not.
The ten billion was Walmart’s. He set out to buy a hundred million shares at about twenty-three dollars. He bought a few. The price ticked up. He thought it might come back. He stopped. Walmart went on to become the largest retailer on earth, and the shares he did not buy are the ten billion. Charlie Munger, Buffett’s partner of sixty years, had a word for waiting instead of acting, and Buffett has used it ever since: thumb-sucking. What was missing was not information. He had every fact he needed and said so. He was waiting for certainty, and certainty was never coming. The twenty-three dollars was not a valuation. It was a price he had grown fond of, and he let the anchor make the decision the analysis should have made.
He had made the identical error thirty years earlier and been rescued by luck. In 1972 the family controlling the US company See’s Candy wanted thirty million dollars; Munger said it was worth it, but Buffett would not go past twenty-five, because three times net tangible assets made him gulp. His caution, he later wrote, could have scuttled a terrific purchase, yet the sellers happened to take his bid. See’s has since earned $1.9 billion pre-tax on forty million of added investment (Buffett, 2015). The same mistake. Once rescued by luck. Once punished by reality. Three decades apart.
Let me put the countercase, and it is my own. Twenty-five years ago, I bought a very small apartment in Bandra, then as now one of Mumbai’s most sought-after addresses. I had no model to speak of. Real estate valuation is famously three words, and they are not taught at Harvard: ‘location, location, location’. A few years later the apartment directly above mine came up for sale. I offered the owner more than he was asking for to close it quickly. Sometime after that the adjacent apartment came up, and I paid the asking price without an argument. In all three cases, I bought because of the location.
But three words is still a model, and a model gives you a threshold. My three-word valuation was crude, but enough to tell me that each asking price sat below what the apartment was worth to me. So, when the seller wanted more, I could say yes. A crude valuation beats a precise anchor.
Then something happened that no valuation of any one of the three apartments would have shown. The three together were worth more than the three apart, and that value existed nowhere until I created it. My premium was now seen. With hindsight, the number I could not see on the day of purchase was the only figure that mattered: what the decision would be worth twenty years later.
And it goes deeper still. In 1964 Buffett had agreed verbally to tender his Berkshire Hathaway shares to Seabury Stanton at $11.50. The written offer arrived at $11.375: an eighth of a point chiselled off. Buffett, irritated, refused to sell, bought control, fired Stanton, and found himself owning a dying New England textile mill. He calls it a monumentally stupid decision and says he became the dog who caught the car. He struggled with textiles for eighteen years before closing in 1985 and confesses he delayed far too long. He has since put the cost of that pique at roughly two hundred billion dollars: what Berkshire would be worth today had he taken Stanton’s price and bought a decent insurer instead.
The eighth of a point appears in the accounts. It is $0.125, precisely, and it can be audited. The two hundred billion does not appear anywhere at all. No algorithm would have caught it. There is nothing to catch. A transaction that never happened generates no data. The most expensive category of error in the career of the most successful investor in history is, by construction, invisible to every detection system ever built, including the ones now being sold as artificial intelligence.
The Invoice That Never Arrives
That which is seen, and that which is not seen. Frédéric Bastiat, 1850
Bastiat’s essay of 1850 is the founding document of opportunity cost. The bad economist, he argued, stops at the visible effect. The good one accounts for the effects that must be foreseen. The visible effect in your own accounts: you borrow at eight per cent, and the interest appears as a line in the profit and loss. It is seen. It is charged. Nobody argues.
Here is what is not seen. Your shareholders also want a return, and more than the lenders, because they rank behind them and carry the risk. But no invoice arrives. No cheque leaves the building. So, the cost of equity, usually the largest single cost your business bears, appears nowhere in your statements. This is why accounting profit is not profit: it is revenue less the costs that happen to send you a bill. Peter Drucker, the father of modern management, put it plainly in 1995. A business that fails to return more than its cost of capital is operating at a loss, however healthy its earnings look.
Economic value added is the profit and loss with the shareholder’s invoice put back in. NOPAT, less a capital charge, and the charge is what the money would have earned elsewhere at the same risk. It is Bastiat, with a number attached. And the number is where the discipline lives. Any executive will tell you, over a drink, that a division has been limping for three years. Very few will tell you its economic profit is negative $4.2 million, because they have never computed it, and once computed, it can no longer be a feeling. It becomes a figure with a sign in front of it, and figures with signs demand meetings.
I have paid for this lesson myself and only found the invoice because I went looking for it. Twenty years ago, I bought a house in Bentleigh, a suburb of Melbourne, at auction, which is how most property is bought there. Several bystanders were kind enough to tell me I was a fool and that I would regret it. The house is worth about four times what I paid.
By any reading of my own accounts, that was a good decision. Here is the entry that does not appear in them. To buy it, I sold blue chip shares I already owned. Years later I did the sums, and those shares would today be worth considerably more than the value of the house today. The sum is directional, not precise, but the direction is not in doubt. So, the four times is real, and seen, and also a loss. There is no line for it anywhere. It exists only because I bothered to check.
And yet I have never regretted the house, because the value of my home to me has always exceeded its price. That is not sentiment dressed up as analysis. It is the oldest distinction in this discipline: fair market value is what a hypothetical stranger would pay, investment value is what the asset is worth to the person holding it. The bystanders were pricing. I was valuing. They were not wrong. They were answering a different question, and it was not mine. Similarly, the real-estate investments I never made would today be worth millions. So are the stock investments, and so, more recently, are the crypto investments I never made.
Artificial intelligence will calculate your cost of equity in a millisecond, before you have finished asking, but it will not automatically subtract it from your profit. That is a decision everyone, including experts like Warren Buffett and Aswath Damodaran, must make for themselves. It is the one step no machine has ever taken for anyone: charging yourself for money that never sends a bill and letting the smaller number change what you do.
Less Wrong Than Everyone Else
Uncertainty is a feature, not a bug. Aswath Damodaran
So why does a man who knows, not move? Aswath Damodaran, who teaches valuation at New York University’s Stern School and has written five books on it, has spent a career on this question. Uncertainty, he says, is not a flaw in your valuation. It is the valuation. And our responses to it are unhealthy: denial at one end, paralysis at the other (Damodaran, 2013).
Paralysis – that is Buffett and the ten billion dollars.
What makes Damodaran worth listening to is what he admits about himself. He knows he will be wrong on every number he estimates. He publishes them anyway, with his workings. The most qualified valuation teacher alive leads with the certainty that he is mistaken.
His diagnosis of the profession is that it has split into two tribes who do not speak. The numbers people believe valuation is modelling and that stories introduce irrationality. The narrative people believe valuation is about great stories, and that models manufacture false confidence. Both are half right, which is the most dangerous thing to be. A valuation not backed by a story, he writes, is soulless and untrustworthy (Damodaran, 2017).
And his test is the most practical instrument in the field. Any story must survive three questions in sequence. Is it possible? Is it plausible? Is it probable?
Most clear the first question. Almost anything is possible.
Theranos was possible. The company claimed it could run hundreds of blood tests from a single finger-prick, and a device that does that does not violate physics. So, the story cleared question one, and the validation began: the US supermarket chain Safeway retrofitting eight hundred stores for three hundred and fifty million dollars, an FDA clearance for a single test, and a board of two former Secretaries of State, a Defence Secretary, a Senate Majority Leader, an admiral and a general. A board assembled to validate a story rather than evaluate a technology. It contained one epidemiologist. Nobody on it could evaluate the technology, and the technology was the entire company. Theranos ceased operations in 2018; its founder Elizabeth Holmes and her deputy Ramesh Balwani were convicted of fraud in 2022.
Tesla in December 2020 was plausible. The market valued it at five hundred and fifty-four billion dollars, more than Volkswagen, Hyundai, General Motors and Ford combined, while it built half a million cars, one per cent of the global market. Read the valuation backwards, asking what would have to be true for it to be right, and it required either ten million cars or a margin roughly twenty times Volkswagen’s. Plausible. A long way from probable. The third question is the expensive one, and we will come to it at the end, with an airline.
Damodaran’s answer is one sentence, and it is the most liberating line in this subject. You do not have to be right. You only have to be less wrong than everyone else.
The Scream Nobody Heard
Tell me where I’m going to die, so I’ll never go there. Charles Munger
In Melbourne, on 14 September 2001, Ansett Australia went into voluntary administration, and thousands lost their jobs in a city that had regarded the airline as part of its furniture. Apply the Z-score Edward Altman built in 1968, and which has been predicting corporate failure ever since, to Ansett’s final accounts and it returns 0.188. The bankruptcy threshold is 1.10. Anything above 2.60 is safe. Ansett was not near the line, nor in the grey area. It was scoring at one sixth of the level at which a company is deemed to be failing.
Now put the others beside it. Enron in 2000: 0.86, bankrupt. One Tel in 2000: 1.32, grey area. WorldCom in 2001: 1.71, grey area. Four companies, four warnings, four sets of audited statements, filed on time, signed off, sitting in the light. The number was screaming in every one, computable in four minutes from four ratios by anyone who cared. Nobody did. Or rather, plenty did, and it changed nothing, which is more disturbing still.
This is where the artificial intelligence conversation usually starts, and it starts in the wrong place. Machine learning models now outperform Altman, ingesting hundreds of variables where he used four. Detection has become nearly free. Detection was never the problem.
Ansett’s problem was not that the score was unavailable. It was that no one would act on a number that implied firing friends and admitting error. A chief executive with capital sunk in a declining operation, Buffett wrote, almost never redeploys it, because doing so requires that long-time associates be fired and mistakes be admitted (Buffett, 2015). He was describing himself and eighteen years of textile mills.
The movement that matters is not from ignorance to detection. That journey is complete, and the machines finished it. It is from detection to anticipation, and from anticipation to action, and no technology has ever helped with the last step. Altman’s model is not a verdict. Reverse it. Ask what would have to change in each ratio to move WorldCom from 1.71 to 2.60, which of them management could actually move, how fast, and at what cost. Munger’s line applies: find out where you are going to die, and don’t go there.
Ninety-Two Per Cent
What is essential is invisible to the eye. Antoine de Saint-Exupéry, 1943
At its 2025 fiscal year-end, Apple traded at more than fifty-five times its book value. Read that as what it is. Apple’s audited balance sheet was describing less than two per cent of Apple. The other ninety-eight per cent was real, generated among the highest profits in the Fortune 500, and had no entry. That gap is not sentiment. It is a forecast: the market has already priced the economic profits it expects Apple to earn and is waiting to receive them.
IAS 38 explains why, with perfect clarity and no embarrassment. Internally generated brands, mastheads, customer lists and goodwill are not recognised. Buy an intangible in an acquisition and it appears. Build the same intangible yourself, brilliantly, over twenty years, and it does not.
The standard is not stupid. It was written for a world where value sat in machinery, and in 1975 the position was exactly reversed: tangible assets were 83 per cent of S&P 500 market value and intangibles just 17. But the four largest companies on earth now require almost no net tangible assets. The largest news distributor owns no content, the largest accommodation provider no hotels, the largest taxi companies no vehicles. The instrument was calibrated for a world that has ended, and it is still returning readings.
Artificial intelligence is the sharpest illustration. It is among the most consequential assets on earth, and almost none of it appears on a balance sheet. A company can spend a decade building a data advantage, a trained model and the people who understand both, and the accounts will record expense, expense, expense, and then nothing. Accounting is extraordinarily good at recording yesterday’s investment and silent about tomorrow’s value. Saint-Exupéry was writing about love. He could have been writing about IAS 38.
The Second Set of Books
You don’t have to be right, just less wrong than everyone else. Aswath Damodaran
Almost every case so far has been a man who knew and did not move. Let me end with the opposite, on a scale that dwarfs three apartments in Bandra. In September 2020, with Virgin Australia in administration, Bain Capital paid A$3.5 billion. The headline is misleading, and instructively so: it included roughly A$2.3 billion owed to secured creditors and A$450 million in worker entitlements. The equity component was around A$1.2 billion. Unsecured bondholders, who had put some two billion into the airline, recovered between nine and thirteen cents on the dollar.
Take Virgin’s Pre-COVID accounts, run a shareholder value analysis on a two-year horizon, and the corporate value comes out at about A$1,244 million. Bain paid roughly A$1.2 billion for the equity. The model, applied to public accounts by an outsider, landed within a rounding error of what the most sophisticated private equity buyer in the market actually paid.
But the model does something else, and this is the part that matters. Stress the seven value drivers Alfred Rappaport identified sales growth, operating margin, cash tax rate, fixed capital investment, working capital investment, the planning period and the cost of capital, and it tells you Bain’s price could only be justified if the operating margin moved from two per cent to about six while sales growth fell to five (Rappaport, 1998). In the middle of a pandemic, with borders shut and aircraft parked, that looked close to fantasy.
Now look at what happened. Virgin returned to the Australian Securities Exchange on 24 June 2025 at A$2.90 a share, a market capitalisation of about A$2.3 billion. The fleet simplified, Tigerair closed, long-haul run capital-light through a wet-lease with Qatar Airways, which supplied both the aircraft and the crews, domestic seat share up from twenty-one per cent to 31.2. The EBITDA margin now runs at over fourteen per cent. The margin transformation the model demanded is exactly what Bain manufactured.
So, Bain did not buy a mispriced asset. Bain bought the right to do the thing the arithmetic said would have to be done.
Buffett waited for certainty about a company he already understood. Bain had no certainty available and bought the airline anyway, in a pandemic, with the borders shut and the aircraft parked. The party with less information did more with it.
And that is the whole lesson of this discipline, stated as plainly as it can be:
Valuation does not tell you what a business is worth. It tells you what would have to be true for the price to be right, and then it asks whether you are willing to go and make it true.
The market’s verdict is not a fairy tale. The shares reached A$3.80 in October 2025 and fell to A$2.15 in April 2026. Could the margins be transformed? Yes. Was the exit price right? Not yet. Two questions, two answers, one company. That is the method working on5 both, honestly.
The Second Ledger
Which returns us to Nasruddin under the streetlamp, and to the three problems every senior executive carries at once. There are things you do not know. That is a gap, and gaps are the smallest problem you have.
There are things you know and do not know that you know. Decades of judgement about your own business, sitting in a box you have never opened, because nobody has walked past and asked you what you are sitting on. And there are things you know, and know that you know, and still cannot use. That is Buffett’s problem; it is the expensive one, and no amount of information will touch it. He knew about Walmart. He said so himself. And he had made the same mistake thirty years before.
Valuation is the second set of books.
It is the ledger you keep for everything the first set refuses to record: the cost of the equity that never invoices you, the future that has no journal entry, the distress that files clean statements, the ninety-two per cent that IAS 38 declines to name. Its purpose is not to tell you what you do not know, but to take what you already know, privately, in the part of your judgement no auditor has inspected and convert it into a number you can no longer avoid.
Every one of those numbers will be wrong. Damodaran’s numbers are wrong. Buffett’s were wrong by two hundred billion dollars over an eighth of a point. Yours will be wrong. Being right was never the point: a wrong number you have to defend beats a right instinct you never acted on.
Buffett says he does not dwell on his mistakes, that he does not look back. He also wrote fourteen pages listing them by name, with the figures attached, and published it to a million shareholders. Those are not contradictory, and the distinction is the whole point. Not dwelling is not the same as not knowing. He does not carry the guilt. He absolutely carries the number.
Goethe understood this two hundred years ago, in a passage he appended to the end of Wilhelm Meister’s Journeyman Years: Knowing is not enough; we must apply. Willing is not enough; we must do. Johann Wolfgang von Goethe, Wilhelm Meisters Wanderjahre, 1829
You will find that line attributed everywhere to Leonardo da Vinci. It is not his. No published occurrence has ever been located before Goethe, and it appears in his own hand in his own book. Which is a fitting way to end. Even the sentence telling us that knowing is not enough turns out to be something the world is quite certain it knows, and does not.
Dr Chris D’Souza is Deputy CEO of CMA Australia and New Zealand
References
Altman, E.I. (1968) ’Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy’, Journal of Finance, 23(4), pp. 589 – 609.
Bastiat, F. (1850) Ce qu’on voit et ce qu’on ne voit pas. Paris.
Buffett, W.E. (2015) ’Berkshire: Past, Present and Future’, in Berkshire Hathaway Inc. 2014 Annual Report. Omaha: Berkshire Hathaway, pp. 24 – 38.
Damodaran, A. (2013) ’Living with Noise: Valuation in the Face of Uncertainty’, Journal of Applied Finance, 23(2).p 34.
Damodaran, A. (2017) Narrative and Numbers: The Value of Stories in Business. New York: Columbia University Press.
Drucker, P.F. (1995) ’The Information Executives Truly Need’, Harvard Business Review, 73(1), pp. 54-62.
Goethe, J.W. von (1829) Wilhelm Meisters Wanderjahre, oder Die Entsagenden. 2nd edn. Stuttgart: Cotta’sche Buchhandlung.
International Accounting Standards Board (2004) IAS 38 Intangible Assets. London: IFRS Foundation.
Ocean Tomo, a part of J.S. Held (2026) Intangible Asset Market Value Study, 2025 Release. New York.
Rappaport, A. (1998) Creating Shareholder Value: A Guide for Managers and Investors. 2nd edn. New York: Free Press.
Saint-Exupéry, A. de (1943) Le Petit Prince. New York: Reynal and Hitchcock.