The writer is a managing director at Frontline Analysts and the author of ‘The Unaccountability Machine’
The 2024 company results season has been well under way this month, leading into its strange postscript — the time of year when annual reports drop into inboxes and on to doormats. The earnings numbers are long since in the public domain, the conference calls finished and the price action has moved on, but still they arrive. Why do we bother?
Partly because it’s still a statutory requirement. But we can’t blame the regulators entirely for the extent to which annual reports have grown over the years. There’s an old joke among equity analysts that if you want to keep something really secret, publish it in a company’s annual report, somewhere between the section on management pensions and the statement on net zero emission goals. As the real action has moved to the headline announcements and investor relations calls, the annual report has turned into a repository of all those disclosures that everybody feels like companies ought to make, but which nobody wants to read.
But surely the numbers themselves are useful? Less and less so, unfortunately.
The big advantage an annual report has over a press release or investor presentation is that all the audited numbers are there. A skilled investor relations professional can spin gold out of the most unpromising straw; costs can be “adjusted”, revenues “normalised” and any bad event treated as a one-off. The annual report is where the confessions have to be made — where did the cash come from, and where did it go?
It is still possible to uncover insights if you take the trouble to read a set of annual accounts in detail, and have the skill to undo the work of the investor relations department in trying to paint as favourable a picture as possible. In fact, there are arguably more opportunities to do so than ever before — the permutations of “adjusted earnings” get more egregious every year. But, would you really advise an intelligent new graduate entering into the financial industry to spend years developing this skill?
For one thing, being able to decipher annual reports is less useful than it used to be. Over the past five years — as many active fund managers will ruefully tell you — outperformance has not been a matter of finding hidden treasure but of picking the megacap momentum plays and hanging on. The ability to ignore red flags — from Tesla’s inventories to Nvidia’s accounts receivable — has been a more reliable source of alpha than the ability to detect them.
And for another, this is a game of processing large amounts of information and combing through it to spot patterns and inconsistencies. It is surely bound to be taken over by artificial intelligence, and probably sooner rather than later. Not only will large language models be able to unspin the numbers faster than the investor relations teams can spin them, they may even be able to dig through the disclosures and find the occasional nugget.
So maybe we need to reinvent the whole concept of the annual report, taking advantage of new technology to do so. And we ought to think big. If we used artificial intelligence to free us from the constraint that a set of accounts had to be comprehensible to a human being, what might we be able to do?
One place to start might be the gap between management accounting and financial reporting. The most misleading numbers in any annual report are often the dates at the top of each column — they imply, often comically wrongly, that 12 months is the relevant period over which performance should be assessed.
Every other number in the accounts is also subject to a compromise between the reality of the business, and the need to provide investors with a set of comparable, consistent numbers that they can use to make choices. We call these compromises “Generally Accepted Accounting Principles”, and it feels like blasphemy for a financial analyst to say anything bad about them, but they are a trade-off of convenience rather than a law of nature.
If every company were to publish accounts today on the basis that it felt best represented its management system, it would be a disaster — they would use the freedom to swamp investors with even more self-serving garbage than they actually do. But in the future, there might be no point in doing so; GAAP might evolve into a set of principles to ensure that there is always enough data and metadata for an AI to re-present the numbers on any basis an investor might like.
That would let everyone find their own way to understand the numbers based on understanding the business, rather than the current quixotic quest of trying to do it the other way round. The future might be brighter and more fun for investors and analysts than we feared.