Algorithmically Determining a Customer's Value

http://jawadonweb.com/?p=1942&

by python_kiss • 15 years ago

2 1 15 years ago

mooism2• 15 years ago

I have problems with that equation.

First, log10 (totalC / totalA) = log10 totalC - log10 totalA. As totalA is constant across customers, we can discard that term. (And think about it: why would a £100/year customer be worth less to you if you had twice as much revenue coming in from other sources?)

Next, the frequency / recency term. "recency" is defined as time since last purchase, so (a) is this measured in days, months, seconds, what? And (b) if your customer is buying from you right now this second, you're dividing by zero and your customer is worth more to you than all the footballers in Europe.

Finally, he doesn't attempt to justify this equation. There's no explanation of what the terms are supposed to mean. It doesn't seem to make sense from a dimensional analysis viewpoint.

Algorithmically Determining a Customer's Value

Related Stories