So, I had a piece in the Guardian published this week talking about the rather ridiculous nature of 'rearview thinking' masquerading as insight that is starting to penetrate the marketing fratenity and its inability to use common sense. It got edited (rightly) so that it upset less people than it informed or entertained. I promised I would publish the un-edited version but edifyingly it has been picked up by a few journalists who want to use bits and pieces so I am somewhat hamstrung in giving the 'full monty' here until those have run.
It is a watered down version of the original rant but a bit more Charlie Brooker perhaps - abeit a bit longer than necessary but bear with me:
Big data can't single-handedly deliver all it promises to marketers
There is a rather disturbing meme doing the rounds, which concerns the value (or not) of "big data". Its impact has been likened to the renaissance, the enlightenment and postmodernism rolled into one.
Unfortunately,
many marketers are bombarded these days with similar dispatches from
the purported outer reaches of cutting-edge science about how they
should be using big data to deliver more value for their customers fueled by those with something to sell or devoid of something useful to say.
What they're really selling is the use of historic data to sell more stuff, more cheaply, while not falling foul of data privacy rules masquerading as actionable future facing 'insight'
I am no longer surprised by the misuse of the word 'insight' btw - hardly anone I know in marketing genuinely knows what an insight is Vs information - but that's another post altogether.
I
think it's time someone told the Emperor "excuse me but you're not
wearing any clothes" before these floppy, ill-conceived notions become
so ingrained they are almost entirely unchallengeable and the digiterati among others start claiming another non-event for their own.
What's the problem?
Some
of the world's finest mathematical minds – including those at NASA –
are swimming in big data and yet something as seemingly simple as
predicting catastrophic weather events is still way beyond our reach.
The earth and its weather patterns occur within a closed system with
hundreds of years of data points and yet we still get it badly wrong. Ever thought about that?
Those
analysing raw data within the marketing sphere can't possibly believe
that they have an advantage over scientists and statisticians struggling
with future predictability across frontiers such as global finance,
medicine and government. So why are we fuelling the belief that we can
now predict future human behaviour?
To be fair most have a vice/voice but not a POV about it. Journalists desperate for a soundbite from those that pay their wages are endlessly spouting nonsense without doing any THINKING. Another 'empty vessel' to pour column-inches into without anyone having to use their brains. Column inches without a POV is lazy - how many real stories are there in the marketing press every week? Count 'em....
My point is not to denigrate
such work of course, far from it. Rather, I think it's time someone pointed out
the audacity and sheer conceit of marketers and agencies and planners who are trying to claim that
they can use big data better than anyone has done in the past. Sound familiar?
The fallibility of human beings
Ostensibly
clever folk in marketing departments across the corporate sphere are
falling foul of two pernicious cognitive biases. Firstly, observational
selection bias – the effect of suddenly noticing things we didn't notice
that much before and therefore believing them to be statistically
significant. Secondly, the illusion of control – the tendency to
overestimate our ability to control events, particularly those that we
have demonstrably no influence over. In terms of filling space in industry mags its a godsend surely?
While cognitive bias has formed the basis of some of our most treasured comedy moments, it can be hugely problematical when applied to human beings and their behaviour in a commercial context.
Comical behaviour
Most comedians understand that replaying our unconcious cognitive biases through our behaviourial heuristics is the most potent shortcut to a laugh. Do a bit of discourse analysis behind all of Micheal McIntyre's entire repertoire and you see what I mean. Clever tho...
Lending increasing credibility
to big data as a silver bullet for personalising and targeting products
and services is dangerous. Those with data services to sell (Google for
example) and those in a position to buy them are vowing to deliver
insight that can predict future behaviour when all they really have is
information about what people did previously. This could all end badly. Except for Nate Silver who rightly enjoys his moment in the spotlight and has some interesting things for us to learn.
But if we've learned anything about human behaviour as marketers, it is that people want to be entertained, surprised and delighted – which means we want the opposite of predictability. You might like beans on toast when you're in the mood, but you don't want to be served it every day.
This is seemingly how big data works, though – it focuses on historic behaviour (with some future propensity to purchase thrown in) – and as a consequence can only deliver ever-diminishing returns because it cannot factor in human serendipity and frailty.
"Why hasn't this person been using up their available headroom on their average basket value with us in the past two weeks following our amazing 2-for-1 offer?" our hypothetical marketer asks her big data consultant. Probably because the big data on the computer can't know if a person has inadvertently locked themselves in the loo by accident, gone on a surprise holiday or just thrown off their mortal coil. There are simply millions of unpredictable paths in the infinite, unknowable array of options.
Niels Bohr, a Danish physicist, once noted that "prediction is very difficult, especially about the future". The burnt out wrecks of numeric-based prediction still litter our cultural highways: the dozens of dotcom startups that sprung up at the start of the noughties, the multitude of ex-music industry executives, enterprising artists and others who rushed to set up cash-generating music streaming sites based on "verified user-data".
Notable start-ups that survived that era, and notable
successors, have shifted to a social networking model, which relies on
real human beings to recommend and share, rather than a computer
generated algorithm because it is more reliable commercially.
Sure, there are patterns in the data, but you
wouldn't bet the farm on them. Unfortunately, that is exactly what the
big financial institutions did when they used their sophisticated data
analysis tools to put us globally into recession. You think we have more insight into patterns of behaviour? Animal spirits have brought down the biggest of houses as we have learnt to our cost.
In many ways, the social networking explosion neatly demonstrates my argument: people-driven networks invariably work better than data-driven formulae, which can't replicate or predict in the same way. And don't fall into the trap of thinking that social networks are predictive – they simply are not. They ebb and flow and are full of anomalies. Just like human beings.
If you want to use data to understand why people do what they do, then you need to accept that predicting human nature can never be solely a left-brained, rational exercise. In simple terms, you have to include emotional, right-brained input to provide a more accurate lens to look at this hazy-at-best picture. Know any Nate Silvers in your data department? I rest my case.
Let's be honest
Generally speaking we buy stuff because of product efficacy, its suits our version of
our self constructed reality and offends the least amount of our friends
and relatives. There are no 'relationships' with brands that in any way
mirror human beings, there are no brand conversations in truth - but there are
people behind them that we pay to offend the least amount of people.
Agency folk and brand folks - be brave and accept it's just an outsourced facsimile of 'customer service' and stop thinking its a valuable agency resource with a future - its just a money maker until everyone finally takes it in-house as they should. And on that point - no-one should ever have a brand as a 'friend' - media guys and PR agencies take note and please strip that bit of self-conceit from your presentations. Real people are a lot more unsantitised and fun. The language is all distorted and unhelpful IMHO.
So...
Systems using rational algorithms fed on a diet of historic information will just rearrange historic information and tell us what has already happened in a new way. I strongly believe we should not rely upon it to design marketing campaigns. Use data to segment and target, but leave out the creative, instinctive process that goes into any successful campaign at your peril.