The Propensity to eat late night chocolate

chocolate bar with REaD Group logo and title 'The propensity to eat chocolate'

By Scott Logie, MD, Insight at REaD Group

Propensity. It’s a funny word, and one we throw around a lot in marketing circles but rarely in everyday use.  The propensity to donate to a certain charity, the propensity for a customer to lapse, the propensity to make a repeat purchase. We can even build propensity models to try and predict these outcomes.

However, you wouldn’t ordinarily hear someone say that they have a propensity to eat chocolate late at night (as I do). Or perhaps that their dog has a propensity to wolf down her food (you can be sure mine does) or even that their wife has a propensity to buy more shoes than could be worn in a lifetime (…no comment).

What is propensity?

So why is the term used so frequently in marketing?  I suppose it satisfies a need we have for wanting to know what our customers are likely to do next. When we talk about propensity, what we really mean is the inclination for someone to do one thing more than any other.

For the purposes of marketing we’re looking for people who are more likely to do the thing we want – or don’t want – them to do.

As with most things in life, this comes down to probability and the likelihood of something happening. So, let’s say that when I stay in London, I buy chocolate two nights out of three when I’m heading back to my hotel. This would make the likelihood of me doing this the next time I’m in London two thirds. It may be that I don’t fancy chocolate so much when the weather is hot, and this likelihood can be increased or decreased accordingly depending on the weather. There may be any number of variables that will affect the probability (but more often that not, I’m going home with a chocolate bar).

But can my propensity to buy late night chocolate help to predict if other people will?

Well yes – if you know they’re Scottish (and are therefore always craving sugar) or that they can never say no to a Snickers after having a few pints…or are convinced that going for a run the next morning justifies the eating of said Snickers. With the right data to understand the driving factors behind the decision, we can predict the likelihood of others doing so.

Individual vs Group

This can be approached in two ways. First of all, at an individual level – what is the probability that I buy some chocolate tonight? And then secondly – in a certain group, who is most likely to buy chocolate later tonight? Having this insight will help to sell more chocolate (or to combat diabetes if the data is used more responsibly).

This doesn’t just apply to chocolate. The same principle applies when selling a product, asking for donations, encouraging someone to renew their car insurance or book an all-inclusive holiday. The trick is finding the people who have a propensity to do these things more than others.

On an individual basis we can use past behaviour (of a person and of others) to work out how each person is likely to interact with a company next – often called next best action. This involves looking at past behaviour, current status (their last purchase – when this was and how much they paid) comparing their behaviour to lookalike customers – and also considering what we would like to sell!

Super models

By taking all of these things into consideration we can construct a model for each person that scores all the probable decisions and chooses the ones that are most likely to happen. This can also be weighted accordingly to areas that are most profitable.

There is a slightly different outcome at group level, but it is a very similar approach. A score is applied to everyone and we then select those with the highest propensity to do that thing.

For instance, if we are organising an event we want to make sure we choose those who are most likely to take part in that event – whether that be high propensity to run an Iron Man, eat an excessive amount of chocolate or to bungee jump from a skyscraper.

Using the right data

Whichever approach is taken, it is essential to have enough data and the right data in order to build the models. The necessary data should exist inside your business – in the form of data you have on existing customers and past behaviour and outcomes (such as who lapsed and why). It can also be useful to incorporate external data – such as online activity, social media and general activity outside your business.

Ensuring that we are harnessing the right data and applying it to our customers and to the outcomes we are interested in will ultimately increase our propensity to make better decisions. Which is what we all want to do, surely? Perhaps I need to remember that the next time I get peckish after a few beers!

Our Voice

The latest data news and editorial from the team at REaD Group

Big Data for Small Businesses: Everything You Need to Know

August 15, 2022   Data

What is Data Quality? Why It’s Important

August 15, 2022   Data Quality

Data Quality Dimensions: What are They? And How to Use Them

August 12, 2022   Data

Data Quality Management: the What, Why and How

July 19, 2022   Data

Speak to one of our friendly advisors today