If creative work was deterministically predictable then movie studios would only release hits. In modern marketing, signals and statistics are both abundant and immediate and suggest that marketing outcomes are predictable - this is a deeply flawed idea, writes Faris Yakob: if there is a scientific framing to marketing, however, it is likely to be quantum.
There is a quotation, attributed to management legend Peter Drucker, that has become part of the common knowledge, or folklore, of corporations. It goes “what gets measured gets managed”.
The first thing I do when considering a quotation, which regular readers of this column will know well, is check to see if it's real. It’s usually not if you found it online, moreso nowadays with AI hallucinations filling the open web with dreck for dollars. It turns out that the aforementioned quote is not from Peter Drucker, but that’s how ‘fauxtations’ work. They free themselves from the constraints of any textual context and fly into the world as winged words of wisdom, attaching themselves to the most recognizable avatar that gives the quote authority.
The quote is usually read as a law but it was intended as a warning. It is derived from a paper by V. F. Ridgway criticising management by measurement. A columnist summed up the argument thus: “What gets measured gets managed — even when it’s pointless to measure and manage it, and even if it harms the purpose of the organisation to do so”.
This happens all the time, both linguistically and commercially. “Jack of all trades, master of none” is only an insult when it lives incomplete. “Jack of all trades is a master of none, but oftentimes better than a master of one” is the full couplet, which inverts the meaning. What the paper by Ridgeway actually said was “Quantitative measures of performance are tools and are undoubtedly useful. But research indicates that … undue confidence and reliance in them result from insufficient knowledge of the full effects and consequences. Judicious use of a tool requires awareness of possible side effects and reactions. Otherwise, indiscriminate use may result in side effects and reactions outweighing the benefits … The cure is sometimes worse than the disease.”
We have a wicked problem to explore. This management ideology was inculcated by McKinsey in light of Taylorism. Simply (as simply as possible) measure every input variable, focused on cost, to optimize the amount of time and money it takes to make things. It is also implicitly a model for how companies create value.
We manage what we can measure and we optimize against those measurements. This means, unless the measurements are financial, Goodhart’s Law applies, which says a measure ceases to be useful when it becomes a target. The intermediate targets will distort the entire system around it to ensure they are being optimized, based on a usually assumptive model of how advertising works.
Recently, Meta (much valued occasional client) has been on tour with their AI pitch. After reassuring the industry that AI will create “more need for agencies” their VP went on to say “I think definitely marketing is becoming more science. Now, even creative is becoming science."
Having dealt with the naive binary of ‘science versus art’ in an earlier column, today I want to explore what they are trying to communicate here, what the implications are and how measurement both can improve and distort our view of effectiveness.
The most recent Value of Measurement report from the DMA points out that: “There is a 67% uplift in the number of business effects for those campaigns that avoid using campaign delivery effects in their reporting. Those marketers who steer away from delivery effects in reporting invest in best practice measurement solutions and have a culture of effectiveness measurement that results in a stronger overall business performance.”
Often we ‘measure’ what is easiest and cheapest and thus rely on metrics from AI powered partners that provide blackboxes of media allocation and, increasingly, creative construction. The proposition these platforms are making is that advertising is predictable, just as Claude Hopkins attempted to ‘prove’ a century ago [Spoiler: he didn’t. The book is hilariously light on evidence].
In essence, by saying “marketing" and “creative” are “science”, they are communicating confidence that this will create predictable returns. This is inherently wrong, and leverages a cultural understanding of ‘science’ as something that is predictable. That is not what science is.
Science is the process of endlessly attempting to falsify hypotheses through experimentation. Those experiments, once demonstrated to hold true within certain confidence levels, allowed certain predictions if enough was known about the system. Then quantum mechanics happened and overturned Newton’s clockwork conception of reality. The implications made people unhappy - Einstein famously didn’t believe God played dice - but the experiments hold true. At a quantum level, deterministic causality doesn’t play a primary role. It’s weird but at the smallest level we currently know about, the universe is stochastic and the observer creates the collapse in the superposition of particles, forcing them into a defined state. There is also entanglement, which is ‘spooky action at a distance’ [Einstein], which breaks all known rules for how information propagates through a system because it seems to pass information faster than the speed of light.
Anyway, if advertising is to be considered scientifically, the appropriate framing here is likely quantum, at whatever level of metaphorical abstraction you are comfortable with. Customers are constantly coming in and out of the market randomly from within a total universe (as do virtual particles, which aren’t virtual like VR but because they exist for a miniscule amount of time, and yet they can interact with the existing system in measurable ways before they vanish).
Advertising almost always ‘works’ to create some positive effect for a marketer as a function of media spend but there is significant upside potential. Multiple profitability studies show that the creative that works best can provide a 10x multiplier on efficacy, which is a variance of one order of magnitude. That’s massive and creates inherent problems in predicting returns. Creative testing, whatever technology it uses, observes patterns but does not provide specific recipes, more broad combinations of sugar, salt and fat [attention, emotion, assets].
What we choose to observe will change the system. If creative work was deterministically predictable then movie studios would only release hits. It cannot work that way. Advertising is a complex system that various vested parties have tried to position as linear causality - but it’s not. The multiplicity of factors surrounding any individual purchase decision are impossible to enumerate. Instead we tend to look at recent interventions as primary triggers of buying behavior but they can only work if the person is in-market and the product is relevant to their needs. Therefore, this inverted temporal discounting overweights the impact of the intervention and is inherently misleading, which is what creates the performance plateau.
Historian Marc Bloch points out that when reaching for a deeper understanding of complex phenomena, linear chronological prioritisations don’t always make the most sense. The things that happen most proximately to a purchase are not necessarily the most impactful drivers thereof. This is obvious to any regular buyer of things - and those practitioners that have read Bullmore’s classic Aston Martin anecdote.
Since we operate in real time and the architectures of demand are being continually re-shuffled by culture, we should indeed take a scientific approach. Marketers are creating their bespoke LLM and measurement models but seem oddly reluctant to take the time and budget to run controlled experiments. Experiments are needed to baseline the data received from partners against their own specific business KPIs.
The inherent tension of AI-powered-outcome-based-media-buying is ‘incrementality’. By definition, the algorithm finds the audiences - using whatever signals from across their digital behavior - that are most likely to buy now. There is a distinct possibility that this purchase would have happened anyway, probably in line with overall market share, and you are attributing outsized credit. The argument seems to be that by inserting an advert at the ‘right time/place/person/data’’ you can shift that purchase at the last mile to your brand. Perhaps. Since the Duplication of Purchase Law explains that all consumers buy from a basket of brands, we can understand brands in the consideration set as being in superposition.
Until the interaction of need and availability intersect, possibly influenced by various moments of truth, nothing is decided. Thus, the real work of most advertising is getting into that consideration set, not triggering a purchase right now. As Heisenberg pointed out, how we measure something matters a great deal: What we observe is not nature itself, but nature exposed to our method of questioning.
