On Saturday evening, with two other statisticians and my wife, I played a great new board game. Don’t worry, there was also lots of wine, takeaway food and great chat involved. The game is called Borel. The basic concept of the game is pretty straightforward; over a series of experiments using dice, cards and coins, does intuition beat statistical reasoning? Of course, there are conditions applied so that there isn’t really enough time to be too statistical, but the concept works.
For example, if three six-sided dice are rolled six times, will there be any consecutive numbers rolled? Basic stats suggest this should be a strong ‘no’ but we did this experiment on Saturday and the first two numbers rolled were 1. We rolled again and got another 1. So flabbergasted were we at losing our (not earned) cash that we reran the experiment and immediately rolled two 5s. A triumph for intuition it has to be said.
Unfortunately, or maybe not, we don’t get to play fast and loose with lots of money every day on a hunch. In business, ensuring we have adequate data to make decisions, and increasing the probability of those decisions being correct are pretty important.
Imagine looking for a new location for a store and not taking into account the road network, the parking availability, the demographics of those who shop in the area, their disposable income and likelihood to buy the products you sell. Unthinkable right? Except that it isn’t so long ago that sighting locations for stores was very much on instinct. It’s only been in the last 10 to 20 years that all these factors have been applied to ensure that there is every probability of a new location being the most successful location possible.
I recently saw one of Sports Direct’s directors (that’s a mouthful) being interviewed. When asked about which failing store groups they bid to take over, he said it was purely down to gut feel. I’m sure there is an element of truth in this, retailing has a notoriously strong reputation for ‘gut feel’ after all, but I’m also sure they have a formula, a way to evaluate the potential in a business to remove as much risk as possible and optimise the likelihood of success.
This is interesting as a lot of these phrases are statistical by nature: every probability, remove risk, optimise success, increase the likelihood. We use stats every day, even if we don’t realise it. Can you guarantee success, we are often asked? No, but we can increase the probability of it happening.
As marketers, in our day to day jobs, we don’t get to make massive decisions about investments in new stores or which companies to take over. However, we get measured on the success or otherwise of our campaigns. And, as such, we try to make sure that we stack the odds in our own favour as best we can. Every day we use probability to ensure that we are delivering the right message to the right person at the right time.
For example, in outbound comms, we are always looking to maximise returns – contacting the most relevant individuals with the minimum investment. To do this we use profiles, models, segmentations to help us understand as much as we can about our targets, weed out those least likely to respond and find those who are more likely to want to buy our products.
Online we use different ways to find the most relevant targeting, sometimes it is left to machines to help us do this but in the background are similar algorithms, finding people (or cookies of people) who look like they browsed the same sites as those who clicked through. Even the way we find these cookies uses statistics, using probabilistic matching to try to find the same person across numerous machines.
As with all modelling, the more data we have, the more observations of an event, the more variables we can vary, then the better our decisions will be. Our mantra is that the more you know about an individual the better your marketing will be. That’s why we are always looking for data to help provide a more rounded picture of our customers’ customers. Sometimes that data will be at the individual level and sometimes at the household the live in or even the postcode they live in. Ultimately, it is all about probability and giving us a better chance of getting a response to a campaign.
One of the joys of Borel was that the number of observations was kept small, the number of variables was low and the time to make a decision was short. Hopefully by adding more data, building up more history and ensuring that more information is available, we can help our clients make better decisions by providing more information.
Incidentally, I won the game by the slimmest possible margin and my wife, who based everything on informed hunches was right behind me. I’d love to think my stats background gave me the edge. But maybe we need to play again just to be sure.
Scott Logie is managing director of REaD Group Insights
