A few years ago, Paul Feldwick wrote a seminal book called The Anatomy of Humbug. It talks about how, over the last century of marketing, we have always had people convinced that their way of doing marketing was right and others wrong. It’s an excellent reminder that in marketing, unlike many other professions, there are many ways to be right (and, of course, many ways to be wrong). But whilst there have always been arguments about how marketing works, before the advent of the internet and ‘digital marketing’, at least both parties knew what it was the other was talking about. It’s not difficult to understand what mass marketing is, or direct marketing, or retail marketing. They are what they say on the tin so to speak. Brand marketing is somewhat more esoteric to create but, conceptually, also not that hard to understand.
The constant conversation about the right way to market still exists today, however, the industry’s complexity and specialisations now make any arguments virtually impossible to resolve. Jane Austin once said, “one half of the world cannot understand the pleasures of the other”. If only we had one half of a room not understanding the other. We often have a room full of people that have only a token understanding of what any other person does. It is not realistic for anyone to have deep expertise in search engine optimisation, human choice and decision making, crisis management, tech stack setup, econometrics and creative ideation just to name a few of the many areas of specialisation in today’s marketing and advertising world.
In theory, this needn’t be a problem. As long as we have all the specialists needed in a room we can come up with the correct solution, right? In reality this rarely delivers optimal solutions. There are several inter-related issues. The first is finding someone who knows enough to know who to have in such a meeting. This is the easiest to remedy. The other issues are cognitive biases and the use of obfuscatory language. They are subtle and harder to recognise and, therefore, much harder to fix.
Confirmation bias means we all think our area of expertise is more important than the areas in which we don’t have so much, or any, expertise. Magnifying this is a sunk cost bias which subconsciously persuades us to read more articles in our area of ’expertise’ and less articles about other industry areas, especially those that may conflict with our current points of view.
These issues are not unique to our industry by any means but because it has become so complex so quickly the effects are bigger than they are in many other industries. The digital revolution has not created a situation where dentists argue with each other over about how to pull teeth out or engineers who think gravity is a bit over-rated because it’s so last century. Yet in marketing and advertising, the last decade or so of evolutions in the industry have introduced a multitude of new tactics, technologies and conflicting philosophies.
The obvious solution is for all of us to read more about the areas we know less about. But, even if we all wanted to, the reality is most of us do not have the time to read about all the disparate areas our industry now operates in. However, there is something we could do that might help minimise the effects of the complexities of our industry, fix the language we use.
What’s in a sound
Look across any trade publication or company website, they are littered with phrases, words and terms which, to a non-subject matter expert, most people on most subjects in most rooms nowadays, convey meaning that is confusing at best and incorrect at worst.
Some of the more regular offenders are real-time optimisation, performance media, right message right place right time, content, user experience, artificial intelligence and attribution modelling.
These all sound reasonable. In fact, they sound more than reasonable. They sound like the pinnacle of modern marketing, all science and accuracy and precision. And they would be if they worked like they sound but the reality is they are much more grey and imprecise than their names imply.
The poster child of the big data world is Amazon and their recommendation engine. It is good at recommending economics books to me primarily because I’ve bought a few before, making me a bit more likely than someone who hasn’t ever bought economics books on Amazon to buy another one. It’s moderately useful and very profitable but in reality it’s rightish message, some of the time. It is a million miles from right message, right place, right time, reflected quite clearly in their basket abandonment rate of over 75 percent.
This is not to have a crack at Amazon, clearly they are doing a lot of clever things (and they are one of the few unicorn companies that actually makes money), but alongside the new marketing tactics that technology and data provide us, we need to remember how complex and whimsical human nature really is and how difficult it is for even our closest friends to know what we are going to do or buy next.
Like Amazon, Netflix also has a solid pool of first-party data but most people I know scroll about for ages to find something they want to watch. Again, it’s hardly right movie, right time. The problem is if we don’t stop and think about this it’s all too easy to overvalue these tactics because they sound more accurate than they really are.
Is advertising content? And, if it is, do we need to distinguish between an online video helping someone build a fence and an ad selling paint? Similarly I would argue ‘user experience’ is another phrase probably used to cover too many situations. Clearly visiting an Apple store is a user experience but is visiting a website really an experience? Some Apple store visitors are happy to spend hours waiting outside in order to experience a new product instore but, online, most people want to learn about, or buy, whatever they are looking for as quickly as possible. No sane person queues up overnight to pay their power bill online, buy some new shoes or to price up a holiday. They are not really ‘experiences’ yet the industry uses the term interchangeably. It stands to reason if we are not clear with our words and meaning then the chances of poor execution and, somewhat ironically, subpar user experiences, only increases. Amazon made a lot of money when it finally allowed people to buy from them without having to fill in all their personal details before purchase. Would they perhaps have identified that insight a little earlier if the industry hadn’t been calling it a user experience?
Performance media and attribution modelling are mandatory in any modern marketer’s arsenal, but both more or less imply that media that cannot be easily optimized or attributed do not ‘perform’. Most people now know that SEM tends to work more effectively when broadcast channels like TV are running. But TV is not typically included in a list of performance media or attribution modelled channels. Again, if we take the terminology at face value it is highly likely we will over value it.
Real-time optimisation is another potentially useful tactic but it too implies a level of precision that is often too high. A look back inclusion window of up to 90 days (even with a diminishing return model) is almost certainly far too long for most purchases. Furthermore, in the wrong hands and/or left to an algorithm, real-time optimisation can easily result in the ‘optimisation’ of random data patterns rather than actual data signals. Algorithmic optimisation could easily, for example, recommend the solution to excess speeding on the road is to have more car crashes because everyone slows down to look at them. This sounds trite but the issue is very real and covered well by Cathy O’Neill in her book Weapons of Math Destruction.
The scientific community, which embraces measurement strategies such as double-blind to prevent people cherry-picking data patterns that they want to see, suffers from reproducibility. Meaning all too often a scientist’s finding is not really a finding at all but a random incident that no one else can reproduce. Knowing this, perhaps we should be asking ourselves how many of our ‘optimisations’ are not really optimisations but are instead random data patterns. It’s clearly hard to ask this if you don’t work in the technology area and don’t know enough about how it works. Especially when the phrases and naming conventions overtly imply fancy maths and science and accuracy.
Right or wrong?
In 1968 George Box uttered a phrase that our industry would do well to remember.
“All models are wrong, some are useful.”
It’s OK to be imprecise. Almost anything involving human behaviour will be but, in order to promote useful discussion between specialists, we should use nomenclature that are consistent with the reality that our models and tactics are often not anywhere near as accurate as the current terminology implies. Perhaps we should bastardise George’s quote and remind ourselves that all marketing techniques can be wrong but sometimes they’re useful. But knowing when they’re right and when they’re wrong requires cross-discipline and naming conventions that help non-subject matter experts understand what each discipline does and, as importantly, doesn’t do.
In an industry such as ours which continues to get more complex every year the ability to simplify and work cross-discipline gets more important every year. Algorithms, tech stacks and big data are all great but they are much better at efficiency than they are at insight. The source of real competitive advantage is the ability to think creatively and link data that haven’t been linked before or look at the same data pool everyone else has and extract unique insights and it remains, for the foreseeable future, a uniquely human trait. It requires a myriad of skills; analysts, technologists, creative and strategic thinkers amongst others. But in order to achieve this, these cross-discipline experts need to be able to understand what the others are saying and doing.
We need to stop confusing each other with inaccurate words and instead use clearer and more easily understood language.
Maybe it’s the aspiration for our industry to be more scientific that has exacerbated our desire for complex naming conventions but scientists understand well the difference between probabilities and certainties and the elegance of simple solutions. It’s why people like Patrick Winston, professor of AI and Computer Science at MIT, prefer to talk about advancements in computational statistics rather than advancements in artificial intelligence. It’s less grandiose but much more descriptive, especially for those of us without a computer science background. If we do want to be more scientific then perhaps, as our industry continues to develop and the services we offer clients grow both in number and in complexity, let’s borrow the right things from science and take pride in simplicity and clarity.
For, in the words of Issac Newton, “Truth is ever to be found in simplicity, and not in the multiplicity and confusion of things.”
