CUSTOMERS’ BEHAVIOUR FROM THE INSIDE OUT
I’m not much of a writer but I can’t resist telling you how interesting and inspiring the study of customers’ behaviour can be, so I’ll try to put into words! It’s like a jigsaw puzzle: so many little pieces to put together to get to see the full picture. The difference is that customer behaviour is not static — it’s always evolving and that’s what’s challenging and intriguing at the same time.
In this article, rather than telling you who Trainline customers are or how to solve a mystery about your own customer base, we’ll look at some tips on how to get started in this never-ending discovery journey… are you ready?
The two elements you’ll need are analytical knowledge and business knowledge. Some advanced knowledge in analytics is assumed as the approach goes from the inside out; it starts by analysing internal data that are then compared to those externally provided by market research. And you will need to know your business — data don’t talk, data needs to be interpreted!
Before understanding your customers, you should focus on your products…
For many industries product classification is a straightforward exercise, for example: if you sell furniture you usually have products already divided into kitchen, living room, bedroom, etc… categories; if you work in the food industry you could refer to the supermarket aisles; if you are in the movie/music industry… come on that’s easy!!!
But what if you, like me, work in the travel industry?
While we know some of the most common purposes of travelling (commute, shopping, business, leisure), we don’t really know why people would want to go from St. Albans to St. Pancras: shopping in the city? Going to work/school? Try to reach the Eurostar station to go on holiday in France, or for a business trip? All these options seem plausible. So how do I classify that specific trip?
Identifying the reasons for travelling is anything but straightforward, therefore we need to get some help from our computers. A simple cluster analysis (k-means) can shed some light by identifying similarities among journeys. This analytical technique is called unsupervised as it chooses a set number of groupings (‘segments’), starting from parameters that are apparently unrelated to one another.
What are those parameters? Well you would need to get creative! Time/day of departure, single/multiple passenger journey, advance/on the day booking are all elements that can work as inputs… I’ll leave you the fun of continuing with this list!
I’ve made this sound all very effortless, but you need to spend a lot of time understanding your data, manipulating it and analysing potential correlations between variables in order to get meaningful answers. The clustering technique is a trial-and-error analysis, so be ready to iterate a few times to identify what the segments mean in the real world and keep discussing your results with your commercial stakeholders and people in the research team.
From products to customers’ classification
Once you are happy with your journey segments you can move on with analysing customers’ behaviour.
Your base data should look like the sample matrix below with one unique line per customer.
Your columns will consist of a series of variables describing the specific buying behaviour. A good starting point would be to associate to each consumer the proportion of journey types they make. Yes, the same customer can travel for many purposes!
It’s also important to include some transactional variables, so that the computers can identify similar purchase patterns and distinguish between more frequent travellers and those who might travel once or twice in a year.
My preference is to avoid averages (i.e., average transactions in a year) as customers often show different purchase behaviours from one year to another. I would use actuals within a specific timeframe instead, taking as a reference the day of first purchase, for instance: transactions within first 7days/ first 30days/etc… and the same with sales (i.e., the value of each transaction).
Once again, I’ll leave it to your imagination! Now you can just apply the same logic as with the journey level data.
We have our customers’ segmentation… What now?
I’d be lying if I said that the part of my job I prefer the most is the analysis itself. What I really enjoy is building stories from the data. Those stories form the hypothesis that people in different marketing teams will have to test. And regardless of whether those hypotheses are proved or disproved, there will always be some learning from them.
I can’t speculate in great length about what to test and what way to test as they are really linked to your business strategy, but to summarise some of the areas where segmentation can really make a difference:
- Acquisition: by building lookalike audiences. In other words, you can decide to acquire new customers that “look-alike” a segment of your choice.
- Engagement, by targeting your existing customers with what you have learnt they need/like.
- Commercial/Finance, by looking at a shift in your customer base. Segmentation can be a powerful tool to track business health. Thanks to segmentation we can immediately recognise a frequent high-value customer from someone that doesn’t use your product as regularly. If the percentage of high-value customers decreases in a given period of time, then that would be an immediate warning that something is not quite right.
- Market research, by indexing the segments against industry benchmarks and understanding the current percentage share of your company.
In a nutshell, customer segmentation is
- A fun analytical technique, which proves your ability in translating customers’ attributes into variables and your expertise in data interpretation.
- A simple way to contribute to your business strategy.
- A smart approach for building audiences.
- A quick measure of success!
