What's a Customer Model?
Jim's Intro: The difference between a Profile and a Model is the
element of time, making models more powerful predictors of behavior.
Marketers who use customer data often talk about "customer
modeling" instead of customer profiling. Modeling is kind of like
profiling, but it is action oriented. Models are not about a static state,
like "Customer is 50 years old." Models are about action over
time, like "If this customer does not make a purchase in the next 30
days, they are unlikely to come back and make any further purchases with
This kind of model sounds so mystical, and it is. To see a mathematical
model predict customer behavior is astonishing, to say the least. The
model says, "Do this to these people and they will likely do
this." The marketer goes out and does what the model says, and a good bunch of the customers do exactly what the model said they
What is a model? Simply, it looks at customers who are
engaging in a certain behavior and tries to find a commonality in them.
The marketer might say to the modeler, "Hereís a list of our very
best customers, and hereís a list of our former best customers. Is there
any behavioral signal a best customer gives before they stop shopping?
What does the data say?"
Building models is usually expensive, because it requires
an awesome amount of talent and experience. There are many mathematical
techniques used to build models, each with their own pitfalls and gotchas.
So hereís whatís in it for you, what this site and the Drilling
Down book is
about. You can create your own models based on the years of
experience other marketers have with customer data. They wonít be as
accurate as the "real" models done by Ph.D. analysts costing
$50,000 - $100,000. But theyíll
be nearly as good and are proven to work.
This site and the
book will show you how to build these models with a spreadsheet (or use
the business rules to write a simple program). Ph.D. not
required. You will learn how to build a potential
value of customer groups model on the next page of this
You can use your models to answer some basic
marketing questions about your customers. Questions you no doubt have
asked many times, such as the following:
- Who do I e-mail offers to? When do I e-mail them?
- Should I promote to some customers more often than
- How much incentive should I provide to get a customer
to do something?
- How can I tell when Iím losing a customer?
- How can I put a value on my customers and the
business as a whole?
- Is my business strong and healthy, or becoming
- What can I expect in future sales from my existing
And you can also use these behavioral models in
combination with demographics and characteristics to produce an even
richer picture of the customer. Which of the following seems more useful
- Customer is married, has children, lives in an
upscale neighborhood, and reads Time magazine
- Customers who are married, have children, live in
upscale neighborhoods, and read Time magazine appear to be
disappointed with our site, because a high proportion of them havenít
visited the site in the last 30 days
This combination of characteristics with behavior can be
powerful. But without the behavior, demographic
characteristics donít tell you much about the future profitability of a
customer to the business.
Next in Tutorial:
Comparing the Potential Value
of Customer Groups