The dismal science  Data-driven rationality is ruining decision-making

Bower immediately responded that any company that listed profitability as its top priority did not deserve to be in business in a few years’ time. So, Ogilvy reordered his list.

It was this later, rejigged version which I saw when I joined the company in 1988. Profitability was now at the bottom of the list, not at the top. Ogilvy had hence recognised what Bower had seen immediately: that profit is best attained as a consequence of other activities, rather than as an end in itself. It is best understood as an ultimate, not a proximate, objective.

This is an example of what economist Professor Sir John Kay calls ‘obliquity’. It arises from the theory, heavily borne out by empirical observation, that those companies which are most profitable in the long run are those which pursue profit laterally rather than literally. (It is, among other things, a devastating critique of the shareholder-value movement).

Note that Ogilvy did not remove profitability from his list: he simply reordered it to better reflect this principle. Moreover, Ogilvy intuitively knew, as did Bower, that order-effects matter. Nothing on the list suggested explicitly that what came first should be prioritised. Yet, by placing profit at the top, there was a strong implicit implication in Ogilvy’s original that profit-seeking was the primary purpose of the business.

Order effects – the tendency for the sequence in which information or options are presented to shape our judgements, independent of the underlying content – are among the most robust findings in behavioural science. Many classic experiments illustrate just how powerful they can be. It is a perfect example of one of the primary findings of behavioural science, which is that while we are consciously aware of content, we are unconsciously influenced by context.

One such example is Solomon Asch’s 1946 impression-formation study, in which participants were read identical lists of personality traits describing a stranger, but in different orders. When traits were presented in a positive-to-negative sequence, participants formed a much more favourable overall impression than when the same traits were presented negative-to-positive – a phenomenon later termed the ‘primacy effect’ in person perception. Asch argued that early traits act as an interpretive anchor, colouring how subsequent, sometimes contradictory, information gets absorbed into the overall impression.

Later work has debated exactly which traits drive the effect and how reliably it replicates, but the core insight – that the order of otherwise identical information changes the resulting judgement – has held up across decades of replication and extension.

The elimination problem

But that’s not all. In sequential decision-making – where options are reduced by a process of ‘elimination by attribute’ – the biases created by order effects are magnified enormously. That is because at every stage you are necessarily eliminating many other possibilities with compensatory strengths without even being aware of their existence.

Imagine you search on Airbnb for “Tuscany” and “hot tub”. You will never see any properties in Umbria or, indeed, properties in Tuscany or Umbria or anywhere else which have a swimming pool but no hot tub. It’s for this reason that the property-search website Rightmove strikes me as fundamentally flawed. I would much prefer an AI-powered version where, among the properties you have actively selected, a number of intelligent wildcard suggestions are shown. In the same way that music sites recommend tracks by bands or composers I may never have actively sought out – or even been aware of. Yes, I have decided to search for detached houses in Margate, but that does not mean that a gorgeous Georgian terraced house in Ramsgate should be allowed to escape my notice.

Indeed, it now occurs to me that if the owners of Chatsworth House were to split the house in two and attempt to sell each wing for a mere £900,000, internet searches would almost certainly find no buyers. This is because the number of people searching for semi-detached houses in Derbyshire over £900,000 must be vanishingly small. It does not mean the properties would not be good value. It simply means they’re interesting outliers and, consequently, de-rated by any process of choice by sequential elimination.

Or consider the standard recruitment default of only interviewing candidates with a 2:1 degree or above. In accepting the primacy of that rule, you are denying yourself the opportunity to interview extraordinarily talented candidates who may ace every other criterion but fail at your somewhat arbitrary first hurdle. (David Ogilvy himself was a university dropout).

It’s a very dangerous practice because academic attainment is a one-way indicator. Yes, it is often reasonable to infer that someone with a distinguished academic record is both reasonably intelligent and assiduous. However, the reverse is not necessarily true… it is completely unsafe to infer the opposite: that someone with a lower-second degree has nothing whatsoever to offer your organisation.

As a thought experiment, let’s consider chess-playing ability used as a proxy for recruitment. It is a plausible assumption that someone who is very good at chess is reasonably intelligent, at least within some narrow framework. However, it is completely unsafe to assume that bad chess players are stupid. There are Nobel Prize winners who can’t play chess to save their lives.

One-way indicators of this kind are probably harmless if only a few people use them. When they become universal, however, they’re catastrophic. If McKinsey & Company and Goldman Sachs demand a 2:1 or above, that’s not too much of a problem. When every employer does, it leads to an extraordinary over-concentration of recruitment – and a consequent waste of interesting talent. Never forget that Damien Hirst was awarded an E in A-level art. It was only through the impassioned pleas of an intrigued teacher that he was admitted to the sixth form in the first place. (DEI initiatives similarly led to absurd collective behaviours when everybody adopted them – with Nigerian Old Etonians from wealthy families being prioritised for promotion over council-house kids).

Lost options

Perhaps the best way to describe this problem is the “invisible opportunity cost of sequential decision-making”. By actively ordering decision-making, we accord a disproportionate importance to the first item on any list when pursuing a process of elimination. One consequence of this is that options with enormous strengths in attributes we have de-prioritised – which are often the most interesting and disruptive options in the marketplace – are never even considered.

I keep arguing that the Heathrow Pod is a significant transport innovation. My case is empirically based and derived from the dynamic pricing of Heathrow Pod parking:  people like riding on the pod so much that they are prepared to pay almost as much to park in the pod carpark a mile north of Terminal 5 as they are to park in the short-stay carpark adjacent to the terminal. Passenger preference, however, probably features very low on the list of transport procurement metrics – certainly far below tedious numerical attributes such as speed, capacity and cost. Thus, any consideration of pods as a widespread transport solution is doomed.

You can see why this mode of decision-making is deeply inappropriate in so-called ‘fat-tailed’ environments: where being utterly outstanding in one dimension is more important than being averagely good in several.

New order

By contrast, many successful innovations have emerged precisely because someone has been given the freedom to reorder the accepted list of priorities. In the Silicon Valley of the 1970s, everybody was concerned with what a computer could do: its processing power, its RAM, and so on. It was one Steve Jobs who decided to look at the ranking order through the opposite end of the telescope: “Never mind what a computer can do, what does it feel like while you’re doing it?” He prioritised usability and aesthetics over functionality. The rest is history.

One way to reduce this problem significantly is to consider using selection procedures according to a checklist rather than a sequential process. In other words, judging things in parallel rather than in series. If that isn’t possible, another alternative is to run the same process of elimination several times, but in random orders, noting what discrepancies emerge.

Where else do we see this problem happening? Data-driven rationality has now elbowed its way to the top of the list in any form of corporate, institutional or government decision-making. Perhaps it is inevitable that this happens, since any course of action now requires an argument, and rationality is usually the easiest way to win an argument, even if it fails to present the best solution to a problem.

But the widely accepted idea that all decisions must be based on pre-existing available and numerical data – however irrelevant this data may be to the wider question – has undoubtedly compounded this effect. It has led to the absurd primacy of corporate finance departments (or, if we’re talking about government, HM Treasury) in decision-making.

It all looks perfectly sensible until you realise the invisible opportunity-cost – namely an extraordinary reduction in innovation, experimentation and long-term thinking. This will always emerge when you award power of veto to people whose primary concern is predictability and certainty. The invisible price you pay for this certainty is the loss of the discovery layer.

Micro-measurement, micro-management

Once you have handed power to these people, the problem will compound. My observation of finance departments is that their principal motivation is the tireless and brutal pursuit of the reduction of headcount in every department other than their own. But they are also deeply averse to the very kinds of uncertainty and ambiguity from which the most interesting innovations and breakthroughs emerge. Theirs is a thin-tailed mentality, desperate to avoid the vagaries of a fat-tailed world.

It was not always like this. Thirty years ago, the finance department was there to monitor and to warn, not to exercise the right of veto. A friend of mine – who is a government minister – was bemused by the fact that he needed permission from the Treasury to move £60,000 from one budget to another. Note that he was not spending any extra money. He was merely reapportioning it. He approached a former chancellor of the exchequer from the 1990s, the great Kenneth Clarke, and asked him what the hell was happening. “I don’t understand this at all,” Clarke told him. “In my day, we gave people a budget and, after a year or so, we’d look at how well they’d spent it. If we thought they’d done this well, they got to keep their job.”

But what gets measured gets managed. And, unfortunately, what gets micro-measured gets micro-managed. Consequently, the variation, serendipity and discretionary judgement from which progress emerges has been engineered out. Consider the following list of desirable characteristics in any business:

• Efficiency
• Growth
• Innovation and experimentation
• Intuition
• Distinctiveness and differentiation
• Resilience and adaptation
• Internal culture
• Customer loyalty
• Brand value
• Sheer luck
• Humour

Efficiency patently belongs somewhere on this list. A very large part of the success of any business consists in the exploitation of what it already knows and what it already does. But by placing efficiency at the very top, what you notice is a massive diminution in all the qualities lower down. Because to an efficiency mindset, every one of these qualities seems either unsafe, uncertain or, in the short term, costly. Viewed through the mindset of an efficiency-first paradigm, anything else of interest can be portrayed as inefficient.

Order effects really matter. Bower and Ogilvy knew this. Whether their successors do is a different question.