The Illusion of Precision

The final week of January saw one of the sharpest relative declines in software stocks in recent years. A series of earnings announcements triggered a broad sell-off. By the end of the week, software and services stocks were down 7.1 per cent in the US and 11.8 per cent in Europe, making this one of the sector’s weakest relative performances since 2020. The move was widely described as a renewed reckoning with AI, as concerns about margin pressure, competitive disruption and even long-term viability returned with force. 

Yet while this explanation has gathered some deserved steam and will undoubtedly be a topic for future discussion (as it has morphed into a broader question on the viability of Software-as-a-Service as a business model), it sits uncomfortably as a trigger. Notably, the immediate catalysts were not Anthropic’s rollout of plugins or structural breaks in demand or profitability, but modest deviations from guidance or expectations of accelerated growth that did not materialise. The market has drawn a crosshair around particular metrics and placed huge pressure on companies to hit them, chalking any miss as evidence of potential AI disruption.

The issue is that these metrics are forecasts—and the challenge is not simply that forecasting is difficult, but that the architecture surrounding financial reporting and guidance creates a false sense of precision. Companies are held to targets that imply a degree of control and foresight that neither management nor accounting systems can realistically deliver. Software businesses are especially exposed. They trade on higher multiples, rely on intangible and contract-based products, bundle services, and exercise significant judgement in revenue recognition and cost capitalisation. As a result, both reported results and forward guidance are unusually sensitive to assumptions. 

A feedback loop then emerges where Management estimates are issued as guidance, this guidance is treated as a firm commitment and Management credibility becomes contingent on achieving this commitment.¹ The result is a system that rewards numerical precision over economic resilience. In stable environments, this distinction is easy to ignore but when conditions shift, the cost of that confusion becomes visible all at once.

Information Laundering

The gap between reported numbers and economic reality is both a feature and a bug. It is neither feasible nor desirable to evaluate a business by looking at every transaction and operational decision. Financial statements exist to compress reality—to transform high-frequency, noisy activity into a form that can be analysed. Accrual accounting in particular smooths cash volatility, in order to reflect underlying economic activity more accurately than movements of cash.

The cost of this compression is that reported outputs such as revenue, earnings, and margins are just estimates built on layers of assumptions rather than hard facts. Each layer improves readability and comparability while filtering out detail. The result is a cleaner signal, but one that embeds substantial hidden judgement. In treating the signal as all that matters, markets often overlook the information contained in the discarded noise, which frequently carries important information about a business’s underlying risk.

The hard truth of software

Software businesses sit at the extreme end of this abstraction. Revenue recognition standards for example impose a common framework for recognising revenue from customer contracts. The cost of this standardisation is discretion. Management must determine when a “performance obligation” has been satisfied and when “control” has transferred. These judgements are simple in straightforward transactions but become highly uncertain in complex, bundled software contracts. 

For example, bundling implementation services with subscription access can allow revenue to be recognised earlier, by asserting that value transfers at the point of access. These judgements are fully compliant with accounting standards, yet they can materially change reported results. The uncertainty does not disappear; it is absorbed into the accounts. 

The same dynamic applies to capitalisation. Software development costs may be capitalised once “technological feasibility” is reached, but feasibility itself is a managerial judgement. Lowering that threshold or reclassifying research as development defers expense recognition and smooths reported earnings, giving the veneer of stability. This is despite the fact that it has no effect on the economic reality whatsoever.

Precision ≠ accuracy

This brings us back to SAP’s selloff and what happens when precision is mistaken for accuracy. Alongside results that were broadly in line with expectations for the final quarter of 2025, SAP reported current cloud backlog (CCB) growth of 25 per cent in constant currency, versus a market expectation of 26 per cent. That one-percentage-point shortfall triggered the largest single-day selloff in the stock since the major guidance reset of October 2020, with the stock settling down -16 per cent.

The market reaction implied a slowdown in underlying economics. The reality was almost the opposite. Management attributed the outcome to three factors. First, a higher mix of large deals, which are structurally more back-end loaded. This reflected SAP’s growing success in closing deals above €5m—an area that had previously concerned analysts.  Second, an increase in sovereign cloud contracts, which are more complex and take longer to implement than standard public cloud deals. Third, a larger share of government contracts signed in the quarter included “termination for convenience” clauses, which allow customers to cancel unilaterally, even without performance issues.²

The shortfall reflected accounting norms rather than weak demand, rising competition, or poor execution. It also showed how the AI narrative has stripped category leaders of the benefit of the doubt. What looked like a miss against a precise target was in fact a function of deal mix and accounting norms. The irony is clear. Greater success in securing larger, more complex and more strategic contracts reduced a near-term accounting metric and was then read as evidence of weakening fundamentals. Precision: one. Accuracy: zero. 

Measure what matters

The purpose of this analysis is not to argue that predictability is an illusion, but to show that much of what is presented as predictability is a byproduct of accounting architecture, rather than a law of business physics. The “smooth line” that markets prize is often a managed output. It may act as a trailing indicator of past stability, but not necessarily a reliable guide to future outcomes. The distinction matters because it separates earned confidence from inferred certainty.

First, management guidance is not privileged foresight; precision in guidance should not be confused with control over outcomes.

Second, much of apparent predictability is internally manufactured to allow a messy reality to transform into stable reported results. Volatility is repackaged rather than reduced and when markets reward the output without interrogating the mechanism, they confuse accounting discipline with business resilience.

Finally, when conditions deteriorate, accuracy becomes less valuable than adaptability. In stable environments, precise forecasts are helpful. In unstable ones, they become liabilities. The relevant question is not whether earnings hit a target, but whether the business can absorb missing it. Stress testing margins, liquidity, customer behaviour and capital intensity reveals far more about durability than marginal improvements in earnings precision. Understanding the real drivers of a business will be far more helpful than holding management to outdated guidance.

In a world of non-linear shocks and rapid technological change, the most valuable attribute a business can have is not a predictable earnings path, but the capacity to endure when things go wrong. Predictable earnings are a consequence of resilience, not its source. For investors, the task is not to punish imprecision, but to identify companies whose economics hold up when accounting smoothness gives way to business reality.