Skip to main content

Predictive analysis can be surprisingly accurate.

So accurate that it has some people suspecting they’re under surveillance! We look at how predictive analysis can assist with finding the right reward at the right time.

A fascinating hypothesis from the fields of robotics and animation called the ‘Uncanny Valley’ states that “human replicas which appear almost but not exactly, like real human beings elicit feelings of eeriness and revulsion.” Apparently, the closer a replica gets to being perfectly life-like, the creepier it becomes. People will happily interact with an anthropomorphic robot because it’s obviously not trying to look like a human, however when the robot looks very human, people always get a sense that something is … off.

UncannyValley_x2The Uncanny Valley – proposed by Japanese roboticist Masahiro Mori in 1970

So what does this have to do with advanced data analytics and reward personalisation? Predictive analytics are sophisticated enough that, given enough raw data, their predictions can be uncannily accurate. Possibly even unnervingly so and that’s what needs to be avoided.

The fine line between personal and overly personal

It can be argued that when it comes to personalisation there is also an uncanny valley, or at the very least a line that should not be crossed. Take care to be personal without becoming overly familiar, as this can have completely the opposite effect and alienate your customers.

For example, some customers will not appreciate being addressed by their given names and would prefer a more cordial business relationship. In this case you would not start an email with ‘Hi Mike’ and rather opt for a formal salutation. Data analytics can provide all these insights.

When ad targeting gets creepy

It is a testament to the sophistication of predictive analysis that it’s common, these days, to have an offline, human to human conversation about, say, the best way to bake broccoli (or something equally obscure) and in the days that follow, receive targeted web ads all about broccoli!

The odds against this being a lucky guess are staggering, so it’s natural to feel ‘creeped out’ by what appear to be (almost) overly accurate predictions. This phenomenon has even led some people to believe that their cell phones are listening in on their conversations. How else could you possibly explain the sudden proliferation of broccoli related communications in your online activities? If you’re not paranoid enough, quickly put on your tinfoil hat and read how the sound of a zipper activates Siri!

Enter the data

The answer to ‘how is all this accuracy possible?’ is of course predictive analytics. There were clearly many antecedent thoughts and actions that led up to your curiosity about baking broccoli, yes?

Perhaps you were at a farmer’s market the week before and either checked in on social media or used some other map or location function. Perhaps you’ve recently bought a small oven. Most likely, you’ve done a Google search along the lines of ‘most nutritious green vegetable’. Big data would know about this.

Could it be that the predictive model that analysed your behaviour knew you’d be thinking about baking broccoli long before you did?

Sensible and sensitive use of data

By now you may be worried that the robots are taking over, but you needn’t be. Predictive analytics, as an element of human-focused design, is on our human side, but it falls to us to use it responsibly (read: not creepily). Don’t cross that uncanny, overly familiar line.

There have been many instances where unguided machine analytics have inadvertently produced insensitive results (or things just went horribly wrong). This happens a lot with online ad placements when a keyword is recognised in analysis but ‘the machine’ has no context. For example, a page containing a news video of a police deputy beating a civilian with the banner ad “Become a policeman” running underneath it. Not good!

(You can find similar examples, and enjoy a bit of a laugh, here: 23 of the worst online advertising fails ever)

Finding the right rewards

When it comes to rewards, the huge amount of data generated by a loyalty program can be analysed, mined if you will, for patterns and dynamics that can then be applied on the fly to reward schemes. This lets you provide the right customer incentives and benefits at exactly the right time for maximum surprise and delight value.

Analysing purchase patterns and comparing historical data to real-time purchasing information will allow you to predict future purchases with a relatively high degree of accuracy. By curating these potential purchases into an offering, you are personalising the interaction.

A personalised rewards experience

And that’s what we’re after. A personalised experience demonstrates the appreciation a company has for its customers; simply by delivering the right incentives and rewards. And research by McKinsey demonstrates that companies that build predictive models based on customer data experience a 126% increase in profit margins. That’s massive!

Build trust through personalisation

Numerous studies have shown that people have come to expect the convenience that comes from personalised experiences, but are simultaneously worried about the security of their online data and invasion of privacy – the creepy factor we’ve talked about. Advanced, predictive analytics are extremely effective at identifying the best way to deliver a personalised rewards experience and are not nearly as Orwellian / Big Brother as you might imagine.

It’s worth your time and effort to find the right balance between a personalised and overly personal approach. And the stakes are high. McKinsey reports that “Personalisation can deliver five to eight times the ROI on marketing spend and can lift sales by 10% or more”.

So yes, you can rely on the machines to give you useful customer information – but that will only take you so far. Don’t just ‘fire and forget’. Stay sensitive to the humans behind the data and get personalisation right with advanced data analytics.


[sc name=”blog_cta_download-eguide_01″ ]


Leave a Reply