The Sandwich method: Real Time Marketing in 3 simple steps

Even not being aware of it, Real Time marketing has been the Holy Grail of every CMO since the paleolithic. Being able to communicate the right message, at the right moment, in the right place (or channel) to the right consumer, and deciding so in real-time, based on past and recent behaviors and a number of attributes, is the ideal way of maximize the effectiveness of our marketing investments. There are some technologies and strategies that make this fully possible today. Hungry for a Sandwich?

We all remember Steven Spielberg’s Minority Report film (must see if you didn’t).  Chief John Anderton (Tom Cruise) had to change his eyes, the identification mechanism for every citizen. With his new eyes, he started to receive personalized messages based on the profile and past behavior of the previous eyes owner…

Well, it’s not that the eye scanner technology is not available in the market now or in the near future (it is said the new Samsung Galaxy S5 will come with it), but there is another way, definitely. In the movie this works because there is:

• An identification system (eye tracking)
• A stored consumer profile (They know your name, size, weight, etc.)
• A tracking system (They know you bought a couple of jeans last time you visited)
• A communication system (the holographic projector)
• A message or campaign ( “How’d those assorted tank tops work out for you?”)
• A real-time decision engine that mixes everything

I implemented my first real-time decision engine for a client (a banking one) back in 2002. At that time, this could only be done within the web, where we could identify the consumers, we had their profile, their web behavior and deliver the message (it was a banner…). Today, we call that Behavioral Targeting.

Now there is a mobile phone that consumers always carry with them, always connected and always located, so what could only be done on the web now can also be done everywhere else. The web is now the world.

Now let’s go with the lower bread slice of the sandwich. Big Data is a top trend now, everybody is speaking about it. In the current world, we can store information about everything: profiles, behaviors, campaign responses… This is the base of your decisions. If you build bottom-up, you will be able to analyze every behavior, every response, every attribute of a single identified consumer. But processing such amount of information for every single consumer is not feasible by a human team, then you need computing. For instance, if you are a coffee chain with a mobile app-based loyalty program, you collect tons of information about every consumer: who they are (age, gender, residence), how they behave (what they buy, where they buy it, when they buy it), and how they relate with others (who are their friends on Facebook, how many followers do they have on twitter). Ideally our machine would read all that information and know that on Sundays, if the weather is over a certain temperature, people around Central Park use to buy Frapuccinos. Then, would automatically launch a push message into the phones of the users around Central Park that like Frapuccinos and personalize the offer based on their usual choices. Some of them like caramel Frapuccino without whipped cream, some others like green tea Frapuccino with it. We would also let them know where is the closest store and/or maybe their usual one if it’s close enough although it’s not the closest. Now imagine how many products, offers, consumers and stores a company like that could have, and the combinations of all of them in order to try to predict what to do with every possible situation. Maybe Watson could process all the data and get actionable insights from it, but that’s not feasible (today) for the computing capabilities at most of the companies.

If, on the contrary, we take the top-down approach (the upper bread slice of the sandwich) , we would take our products, generate campaigns, and try to deliver them at interaction points where we know the user buys the product. Following the previous example, the Frapuccino brand manager would create a summer opening campaign, sending a Frapuccino discount offer to every user that bought a Frapuccino in the last month that is close to a store (I know some of you are right now mentally building the SQL sentence of that query). But maybe that day the temperature is below 15ºC or the users receiving this offer are not close from Central Park. The result: lots of coupons sent, low conversion, high cost (assume there is a cost in sending the messages, for example  a display campaign in 3rd party mobile apps with a CPM cost schema).

How to combine the two without huge computing and wasting money? With the sandwich method, in three not so complicated steps:

1. Gather all the data (bottom-up). Storage is far cheaper than in the past (and far more expensive than in the future). Even if you are not going to use it, maybe in a couple of years you will have the computing capabilities to process it and discover a correlation between outside temperature and product consumption. Particularly important for most of the categories are consumers, products, places and time.
2. Work from products to discover behaviors based on hypothesis (top-down). It would be good to have a better understanding of the Consumer Journeys through some research. For example: certain people like to consume frozen coffee outdoors on the weekends.  Go to the data, dig into it, and verify your hypothesis. If you succeed, you might want to create some data aggregations in order to better manage your information. For example you could classify the Frapuccino consumers into three levels (intensive-average-low), so you don’t need to count how many times did every consumer buy the product the last month, but just say you want to address intensive Frapuccino consumers.
3. Define a rule for the validated hypothesis that will be automatically launched to the users that fulfill the conditions. Be sure you personalize the message too (because the product can be configured or because you want to do A/B testing to see which creative message works better). A rule is something very simple (IF…. THEN…. ELSE) that is executed every so often, or based on an event, triggered by a specific consumer (oh yes! event-driven marketing also fits here…), for example: “John Doe is close to his favorite store, find me something for him”.

And start back again. If you’re doing things right, you will start having lots of rules active at the same time. And more than one can qualify to be launched to a specific consumer at the same time. You need to define the criteria that allows to choose from the available rules (/campaigns). For example, the one with lower execution cost or the one selling the product with higher margin.

Needless to say that you need to measure results and readjust every one of your activities, but that is a quite common thing to do these days, the same we did back in 2004….