June 21, 2010

Summer Segmentation: Seasonal Buyers

You probably already know this, but seasonal buyers are different than other buyers.

One easy way to understand this is to segment your customers by months since last purchase. Take all of your customers with at least one purchase as of May 1, and then measure repurchase activity during the month of May. Your table might look like this:


Recency = 1 Months, Repurchase Rate = 10.0%.
Recency = 2 Months, Repurchase Rate = 8.3%.
Recency = 3 Months, Repurchase Rate = 7.7%.
Recency = 4 Months, Repurchase Rate = 7.2%.
Recency = 5 Months, Repurchase Rate = 4.8%.
Recency = 6 Months, Repurchase Rate = 4.7%.
Recency = 7 Months, Repurchase Rate = 5.9%.
Recency = 8 Months, Repurchase Rate = 5.5%.
Recency = 9 Months, Repurchase Rate = 5.2%.
Recency = 10 Months, Repurchase Rate = 4.8%.
Recency = 11 Months, Repurchase Rate = 5.1%.
Recency = 12 Months, Repurchase Rate = 5.5%.
Recency = 13 Months, Repurchase Rate = 5.1%.
Recency = 14 Months, Repurchase Rate = 3.9%.
Recency = 15 Months, Repurchase Rate = 3.6%.
Recency = 16 Months, Repurchase Rate = 3.3%.


Look at the red rows, at months five and six. These are Holiday shoppers, and Holiday shoppers often do not repurchase at the same rates that do other customers. Here's a place to reduce marketing expense.

Look at the green rows, at months eleven through thirteen. These are Seasonal shoppers, they last purchased similar product about one year ago. These are customers that require increased marketing expense, because this is the time when you can "win back" the customer, as the vendor community likes to say. These customers are likely to buy comparable product one year later.

This is about as simple as segmentation gets, and yet, it's completely actionable and it makes intuitive sense.

So go use this tip to grow your business!

2 comments:

  1. Would you still recommend this kind of segmenting and changes in promotion strategy if a model is being used to select customers that considers seasonality? Even if the seasonality variables are considered but do not make it into the final model?

    ReplyDelete
  2. If your model doesn't find seasonality variables to be significant, then I do not recommend following this strategy, trust the model as long as the model is built well!

    ReplyDelete