Monthly Archives: January 2012

You are browsing the site archives by month.

Algorithmically Determining a Customer’s Value

Data crunching can reveal a treasure trove of useful information.

For example, popular news websites, such as Reddit, use nifty metrics to decide which stories to promote on their homepage (Number of votes vs. Time since the story was posted).

Similar algorithms can be used by businesses to discover loyal customers. Special offers could then be extended to strengthen relationship with customers.

So, using a dataset from my own customers, I conjured four metrics that help determine the value of a customer:

1) Time since their last purchase (recency)
2) Number of times they have bought from us (frequency)
3) Total amount they have spent with us (totalC)
4) Proportion of how much they spent with us compared to other customers (totalA)

These four metrics act as a voting mechanism. The more a customer spends, the higher their score. Moreover, the value of a customer is depreciated if they haven’t made any recent purchases.

I pieced together the following equation to represent the above metrics:

Using this equation I was able to determine any customer’s value and scale it on a Netflix style star rating system within our customer management system (CMS):

Data mining is an untapped opportunity for businesses to provide customized solutions to their customers (i.e. at retail stores, call centers, etc). Some tweaking of the above metrics will help you extend these benefits to your own client base. Now that’s a win-win!