My favourite story this week comes from an article published in February in The New York Times Magazine entitled How Companies Learn Your Secrets. Click the link and please read it – it’s mind blowing.
Ok, if you didn’t read it, here’s a summary – Target stores in America developed some sexy analytics to interpret shopping transaction data to create personalised individual marketing campaigns and incentives. So far so good. The market that they were targeting was pregnant women (leaving the pregnant men’s market for future growth opportunities). They were able to establish a sufficiently high correlation of shopping behaviour during the early phase of pregnancy that they could predict a future birth and the creation of a new ‘mother’ customer. Targeted advertising of baby and motherhood products in the lead up to the birth combined with incentives appear to have contributed to a massive increase in revenue for the retailer.
At this point, it looks like one of the already well worn feel good sales stories for big data – give us the IT budget and we’ll give you unheard of growth in revenue.
That probably wasn’t the thought foremost in the mind of the father who went in to complain to his local Target store about his daughter receiving promotions which were age inappropriate and unsuitable for a conservative, I’m guessing, Christian, household. It turns out that the daughter hadn’t told her family she was pregnant – instead, Target brought the conversation out in the open through their personalised marketing!
After the extreme impact of this social marketing gaffe, Target commissioned further research and found that – shock – instead of appreciating their offer of ‘help’, a lot of people thought that it was just too creepy to have a retailer stalking them along life’s journey.
The agile response to this consumer feedback was to include irrelevant promotions along with the targeted promotions so that consumers thought that the ‘right’ products had appeared serendipitously. Whoa!
This story alone would make a great movie – or maybe some future HBR case study in what to do or not not do in FMCG social marketing. Even more interesting to me was the underlying psychology and behaviour theory which suggested the goals for this kind of campaign.
As the article explains, with many useful references, buying habits are hard to break. This is another side of previous comments in this blog about the ease of intuitive, recurrent thinking compared to hard consideration of multiple dimensions and possibilities. From Target’s point of view, if someone always buys milk from their local grocery store it would take an extreme impact to change their default behaviour and buy milk from them instead. No matter how great the marketing, the potential milk buyer simply won’t register the possibility of buying milk elsewhere. Target could try to engineer an extreme impact – maybe giving it away for free or tainting the competitors supplies or……..they could wait for mother nature to deliver the event and track that through correlated evidence of behaviour.
Extreme life events such as divorce, death (of someone else – not the customer), and birth, cause extreme impacts and these events create a sufficiently complex environment that the customer will be open to change behaviour and break with old habits. I was already aware of this pattern from my work in crisis management and national security but I never made the connection to buying baby products. It makes perfect sense and best of all, I can speak from personal experience of the complexity introduced with a baby – particularly the first baby.
One of the challenges in educating people about complexity and extreme impacts is the difficulty of conveying that sense of a mind altering experience which sweeps away long held beliefs and assumptions.
Unfortunately, the baby example doesn’t really hold for true extreme risk. As a metaphor it would require that the woman wakes up one morning to unexpectedly find herself 6 months pregnant. In other words, if terrorists bought the same range of products from their local store a few months before an attack, it would make life a lot more simple for the intelligence community (this does kind of happen in a way, but we’ll come back to that later).
So let’s not confuse the discovery of correlation over a large population sample experiencing a regular (although at least initially quite private) event with a low likelihood / high consequence event where you don’t know the population and the events are emergent.
The real extreme impact here was the sudden appearance not of a black swan, but of a discount retailer playing the role of a stork.
Where does ‘convenient’ stop and ‘creepy’ start? I guess we can thank the commercial imperative to help provide us with some of the experiments. In any case, it’s worth considering how your own big data initiatives will play out and whether you will be perceived as using this power for good or evil. That’s an extreme risk.