What’s next for your customers – it’s in the math!
Once we have completed a purchase or transaction with our customer, we should immediately begin thinking about how to continue our relationship so they become a raving fan and lifetime customer. Some of the ways in which we do this are obvious, great product and customer support, for example. However, now that we’ve got some ability to see the future (predictive analytics), we can enhance that relationship in powerful new ways.
Most companies are similar. On its face, that statement feels wrong. What does on ice-cream shop have in common with a furniture manufacturer? It turns out when we use the right lens, quite a lot. Buying patterns are often highly correlated across businesses when we take into account enough information about them. For example, we may see that the furniture company buys 30 new sets of printer ink every six months, while the ice cream shop buys one every six months. That doesn’t seem similar at all. However, if we look at the number of people who use the printer, we might find that there are 30 times as many people working at the furniture manufacturer. So the similarity becomes, we buy one new set of ink per person, every six months. It’s a very simplistic example but illustrates an important point. When we find the right things to look at (number of printer users), identifying similar patterns becomes much easier to understand.
Predictive analytics is, at its core, just a way of finding those relationships and using them to make guesses about what’s going to happen. If you want to know the future, wait until tomorrow. If you want to know what’s likely to happen, use predictive analytics. Weather forecasting is a great example, it’s not right all the time, but it is invaluable in helping you plan your day.
So, back to the core question, how do I know what my customer is likely to buy next? Turns out that this is a problem that is an excellent fit for predictive analytics because we have lots of data (customer, product, price) about each transaction and lots of transactions. So, we build a mathematical model or formula which gives us the right answer for an observed behavior most of the time. So that’s neat, we can take a customer’s information and purchase history, pump it into our formula and it will tell us what the customer actually did (at some level of accuracy). The real leap here is that we can also predict behavior which hasn’t happened yet. That’s some information we can really use.
Predictive analytics lets us take our observations about a customer’s previous behavior, combine it with lots of other customers, and predict both what we think will happen and how likely it is to happen. Customer 1000 will buy a new table after buying eight chairs with 80% likelihood, based on the behavior of customers 0-999.
Think about what changes we could make to the way we market and sell if we knew who needed what — before they did.