Can one metric do the job for ranking customers?
I am admittedly biased. When asked what I think is the best metric for ranking customers, I invariably recommend our proprietary Loyalty Score for two reasons. First, it is highly predictive, corresponding to the probability that the scored customer will make a purchase in the next period. Second, it’s tested and proven through use by many customers over the past few years. We have a lot of confidence in it. Besides, I enjoy tweaking people who think the only good measure is revenue.
Lately however you might hear a different answer from me. It’s not that Loyalty Score is less useful or has deteriorated. The opposite is true: our algorithms are better and Loyalty Score is more powerful than ever. Rather, it is my perspective, my point of view, that is changing. Now that the new tools are in place, I find myself advocating a different strategy to our customers. Instead of merely ranking their customers on a linear scale, I recommend exploring the whole customer set to find customer segments with varying needs, and then marketing to them with appropriate messages. Identifying those segments usually requires looking at more than one metric, so I’m reluctant to offer up a single answer to the ‘what’s the best?’ question.
For example, if you are planning a campaign to win-back customers, you’ll want to rank the customer both on Recency and Revenue and Loyalty Score. You need Recency to find the customers that haven’t bought in a while; Revenue to find the customers who are worthwhile, and Loyalty Score to figure out which ones you have the best chance of recapturing.
Another example is finding customers who are good candidates for cross-sell. Here I would rank customers based on their category score (how many different types of products have they purchased) and the change in their Loyalty Score. I’d pick candidates whose Loyalty Score is changing in a positive direction (they are good customers), but whose category score is low (they are buying deeply but not broadly).
Other possible metrics are Risk Score (likelihood of defection) and an RFM score. What is your favorite? Will one metric do the job? Should I go back to my simple-minded point of view?