May 22, 2014

Summer ... And Visits

As you already know, I won't necessarily post every single work day between Memorial Day and Labor Day. So please keep that in mind.

Now, on to today's topic. Here's something from a recent project. I looked at all customers with exactly 3 purchases in the past year, then segmented customers based on the number of measured website visits in the past year. Finally, I calculated the probability of the customer purchasing in the next 30 days. Here are a few data points from that analysis:
  • 3 Orders Last Year.
    • 6-10 Website Visits Last Year = 29% chance of buying next month.
    • 11-25 Website Visits Last Year = 44% chance of buying next month.
    • 26-50 Website Visits Last Year = 61% chance of buying next month.
Of course, this is opposite of everything you've been taught. You've been taught to convert that website visitor RIGHT NOW, or you are a failure. You are supposed to follow this customer all across the internet, flashing silly ads featuring the product the customer recently looked at ("your local hardware store is featuring crescent wrenches at 20% off plus free shipping"), or using cart abandonment programs to give 25% off right now.

Basically, you are using discounts and promotions to encourage behavior that was likely to happen anyway. But because you only measure conversion, you don't ever get to see that the behavior was likely to happen anyway.

Get your web analytics data out of your web analytics tool, and analyze customers longitudinally. Or do it within the advanced web analytics tools. But please do it ... please analyze customer behavior longitudinally. Customers have a natural rhythm that we are constantly spending money to interrupt ... we spend marketing dollars to bring a customer back to the website when the customer was going to come back anyway ... and we give away gross margin dollars to get the customer to purchase when the customer was going to purchase anyway.

We've built an entire marketing ecosystem on top of faulty analytics. We happily choose to not analyze website behavior longitudinally. We're paying a hefty tax for not being willing to analyze website behavior longitudinally.

Contact me (kevinh@minethatdata.com) for your own, customized online/retail dynamics project.