Predicting Customer Spending Habits: Where machine learning meets retail

Predictive modelling can help us anticipate customer purchases

How Well Can You Predict an Individual Customer’s Spending Habits?

In my previous post, I tackled the issue of predicting customer churn. However, another equally important concern for businesses is predicting an individual customer’s spending habits.

Imagine this scenario; you’re a manager in a large retail company that operates a loyalty card scheme. One of your main tasks is predicting how much a specific customer is likely to spend within the next week. Sounds pretty straightforward? Not necessarily.

Typically, you’d notice some patterns:

• Customer spending often peaks at the start of the week (weekly cycle)
• Some monthly and yearly cycles might be noticeable
• Holidays like Christmas, Easter, and bank holidays usually trigger an increase in customer spending

However, issues arise when you focus on individual customers:

• Infrequent customers: Some customers may only visit your store once
• Frequent customers: Others might be regulars and have hundreds of visits under their belt
• Dormant customers: A customer might go dormant for months and return, causing unpredictable fluctuations in spending

When you break it down to the individual level, it’s challenging to notice any identifiable patterns due to this noise. Weekly and yearly trends are only evident when studied at an aggregate level.

So, how then do you predict future expenditures of a certain customer? Well, there are two primary methods, each stemming from distinct traditional disciplines:

1. Predictive modeling: This approach, rooted in machine learning, targets individual customers
2. Time series analysis: This, on the other hand, deals with groups of customers and has its origins in statistical analysis

Depending on the professional background of the person you consult, you might receive different answers. If this is the case, this introductory video should help clarify things.

In my article Fast Data Science — How Well Can You Predict a Customer’s Spending Habits, I will describe how predictive modelling can help us predict customer spend. If you’re interested in time series analysis, you can read my next post on predicting how much a group of customers will spend.