When we can drive Purchase #2, we get another touch point with the customer and another opportunity to forge loyalty. Once we have a list of one-purchase customers, we can segment them in various ways.
This is the easiest segmentation to perform, but can often be irrelevant. There are cases where certain types of products are only applicable to certain age ranges or genders, but in general customers' behaviors aren't determined by their demographics. Especially in this case, where the customer has already made a purchase, we have so much more information from which to draw to perform better segmentation.
Something as simple as “first product purchased” actually gives us a ton of information. In order to actually have a first product purchased, a customer had to come to our store, find the item, put it in their shopping cart, and give us their shipping and credit card information. The number of customers that actually complete all these steps is a small percentage of customers, but purchases are one of the most powerful things you can use for segmentation. Purchases, more than clicks, email opens, and sign ups, are the most powerful predictor of future behavior.
One issue with segmenting on purchase history is that it doesn't give you much to go on in terms of next steps. It helps you driving similar purchases, but it doesn't tell you where to go from there. For customers who first purchased a particular item, should you try cross-selling? If so, what should you cross-sell them on?
Predicted response segmentation takes first product purchase segmentation a step further to the answers of the questions posed above. Suppose you run a drug store that sells three types of items: toothpaste, vitamins, and razors. You can split your customer base into those whose first purchased was in each of those categories. Then, within each of these groups, split them into thirds and send them emails featuring each of the three categories and look at the responses.
You don't only want to look at the revenue per user, but also the conversion rate. If you're going to make decisions based on this type of segmentation, it's also important to do statistical significance testing. This will inform whether the behaviors you observe are actually meaningful, or whether they are the result of random error. This will help you better target your subsequent campaigns. If toothpaste customers respond better to vitamin emails than the other two, then you will want to send them vitamin emails going forward.