In November last year it was announced that Tesco would place face tracking technology at the tills in 450 petrol stations through the UK. The press spoke of Big Brother and Minority Report style intrusion. The tone was one of “how dare they” our privacy it seemed was under threat. All this in the world capital for CCTVs with an estimated one camera for every 11 people.
Tesco are putting camera’s at the tills in the petrol stations, so that they can perform some simple segmentation of the audience and display “targeted” adverts on displays beside the tills. In the land of the CCTV and in terms of snooping or invading privacy it is hardly Sherlock Holmes material.
What Tesco is doing is just scratching the surface. There is so much that can be done without the cameras at the pumps. Knowing the gender and height of the user is the cherry on the cake. Depending on your perspective, you should either enjoy or fear the cake.
Lets start from the bottom up and look at what information we need to create more and more targeted advertising.
For Tesco to construct targeted adverts they need to pull in information from a range of sources, a pyramid of data. Some are public and free others have to be paid for and some are data collections that Tesco already own.
Ambient data is all about our environment, our culture, time and place. It is the public data that we all swim in. Ambient data fills the stores with hamburgers and ice creams when a hot weekend is anticipated. The shelves groan under crates of lager when a cup final is played. The umbrella sellers come out on Oxford street when it rains. The perfume adverts clog our ad breaks as we run up to Christmas.
In the digital world where adverts can be selected and constructed in real time it we can go further. Tesco know the time of day you are filling up and the location of the petrol station. They can do simple things like use the time to alter the message, lunchtime snacks, late night munchies, early morning lattes. They know from the data collected from the millions of previous customers, the kind of products that sell most in different parts of the country. The purchase patterns in a forecourt close to a motorway, may be different from that in a city centre. They may even know that particular pumps in a forecourt are more frequently used by larger vehicles and HGVs.
They know the weather at each forecourt, and the impact that will have on the feelings and needs of the customers. A wet, windy February night makes one dream of being away somewhere hot so holiday adverts may be appropriate. Alternatively a warming soup or offers of wellies for the kids etc.
One more item of location data is whether the customer has driven into the forecourt before or after doing their shop. Sensors on the ground can provide insight into which route they have taken, or surveillance cameras can track route to pump. Aggregated data about customer purchases can give insights about what items are most likely to have been forgotten off a shopping list and adverts can promote these to the post shopping visitor. While pre shopping visitors could receive messages of offers and promotions to tempt them into the store or to nudge them into making purchases.
Lets step up the pyramid. Some of the more valuable elements of the data mix are not the ones about you and me personally but the aggregated data about us all. When you serve millions of customer who purchase billions of items and a large proportion of them are Clubcard customers, then you can gain very detailed insights into shopping. You can find correlations that make it easier to predict what consumers are likely to purchase. It is not about you, but people like you.
We can infer a lot of data about people like you from the car that you drive. Firstly what category is it, a small city car, a family car, and MPV, a sports car, a van etc. Then there is the colour, the colour of the car you drive says something about your personality and about what you want to say to the world. Maybe it is not a conscious decision to broadcast this to the world but on the scale of millions of consumers it is possible to identify correlations between car colour and purchase patterns.
Then we come to the actual make and model of the car. Would you serve the same marketing message to the driver of a ten year old Fiat Panda as to a brand new Mercedes S Class? Maybe other factors in the data mix might trump this element but it is still a strong differentiator and indicator of likely income, social demographics, spending patterns etc.
The cake is almost read for baking, We now finally, arrive at the point where we look at you. Firstly your number plate. A DVLA look up and we can find the name and address of the owner and cross reference this with house prices and home income data. In Tesco’s case they can go further and cross reference their Clubcard membership, home/travel/pet/car insurance and telco, banking and most likely other proprietary data sets, to see if they actually know you or your household. If they do find a match then they could pull on all applicable past purchase behaviour and feed it into the mix.
Assuming they do not strike gold, they can use image detection to identify how many people are in the car, is there a child car seat, bike rack, etc.
Finally, finally the cherry steps out of the car and image analysis discovers the gender, age range, body type of the individual.
There is a lot of data that could be used to make the adverts you see while standing in the queue to pay – or better still while filling the tank. Masses of business logic created to determine what should be seen. It begs the question is this application of big data all worth it? It depends, on how many people you see every day, what you are selling and how much of a difference it makes to actual sales. It is a leaning game where the algorithms would be tested and retested, tweaked and changed, hypothesis challenged and reformulated.
Big data is often seen in a purely digital context: the digital traces we leave from browsing, searching and purchasing. But the data analysed does not have to be digital in origin. When you marry data form many sources you can find correlations which may provide profitable insights. Tesco is an old master at this game. You can be assured that it is highly unlikely that Tesco – or another retailer - is doing all of the above, but I would be surprised if they will not go beyond just fuel pump cameras.