Sounds simple right? Most marketers would say that they use some notion of
value when they target customers. I think there's more to this topic. First,
most marketers use likelihood of response as a predictor far more often than
value. A recent conversation with Tim Suther of Acxiom got me thinking about targeting techniques. Here's my take:
Generally most marketers have three approaches to onsite targeting.

  1. Deterministic targeting. Also
    known as rules based targeting, this is the most commonly used approach
    and takes the form of – "If visitor performs the following actions,
    then offer x, y, or z. Or even simpler "If visitor belongs to segment
    A then offer product x.
  2. Non-deterministic targeting.
    Also known as "self learning", this is when a decision system
    applies advanced analytics to train a model and apply it to making
    decisions based on visitor behavior.
  3. Predictive targeting. This
    takes the second approach further and applies business rules, constraints,
    pre-existing segmentation, and analytical techniques like decision trees
    and neural networks to drive the targeting process. Not surprisingly this
    approach also requires the most complete profile of the customer.

Now
what’s not often discussed is the cost associated with each approach. What
marketers must recognize is that the act of targeting isn’t “free” and has a cost
associated with it. This cost is made up of system costs, data costs, analytical
costs, and experience costs. Of these costs, the hardest to assess is the
experience costs i.e. the cost of presenting an offer to a customer and the
subsequent reaction and impact on future value. This is exactly where customer
value should fit into the equation. Before marketers decide to target a visitor
online they must understand the overall value of the visitor to current and
future business and use that value to help guide the targeting approach.

For
instance, if I am running a  financial
services web site that see hundreds of 
prospects each day, I should be able to assign a “value” to each visitor
based on what I know about them and people like them. This doesn’t have to be a
detailed metric like customer lifetime value but can be a simpler proxy. Once I
can assign value I should use it to guide my approach – a simple (and cheap) one like rules
based targeting or a complex (and expensive) one with a higher likelihood of success like
predictive targeting. Or better still; the value or a value proxy should help
me understand that some customers aren’t worth targeting at all.

So
the question for you is: How many of you use the concept of customer value to
support online visitor targeting?