Marketing Analytics
RFM and data mining have been used for years by Forune 1000 companies to gain significant advantage over their
competition. Data mining uses a number of different algorithms (e.g., neural networks, logistic regression, decision trees,
K-
Means, etc) to develop response models and customer segmentations that support targeted marketing efforts and
messaging.
Response Modeling
Response models predict which customers will behave in a particular way in the near
future. For instance:
- Who is most likely to respond to my direct mail campaign?
- Who is most likely to attrite/churn?
The notion is, "why spend marketing budget on customers that won't respond?"..."why spend money trying to retain
customers that are happy and don't plan to cancel their relationship with us?" By only spending budget on those
you
should be targeting, you have more funds left over to spend on additional campaigns or contribute to higher
profits.
The graph below shows the result of using a predictive model designed to estimate the probabiliity of responding to
a
campaign for each of our customers. Customers are then ranked from highest to lowest along the horizontal axis
based
on their probability to respond. The vertical axis measures the number of responses expected. In this case,
we see we
can get 94% of the expected responses by contacting only the top 40% of our customers.

If a direct mail piece costs a total of $3.00 and our customer list is 75,000, we are saving $135,000
(.6 x 75,000 x
$3.00)
by only targeting the 40% most likely to respond.
Customer Segmentations
The notion behind customer segmentation is to divide your customer list into separate groups based on their
profiles
(e.g., purchase history, income level, industry, etc). Everyone in a given segment is similar to each other, while each
overall segment is different from other segments. By gaining a better overall understanding of
customers, then
grouping
them into categories, companies are able to better optimize marketing programs and
allocate marketing dollars
more
effectively. Let’s look at the below example.
The below chart shows how customer segmentation, based on “clustering” can help us more effectively commincate
to
our customers. This example is based on data about customers and the drugs they purchase along with
demographic
data we have obtained.
This shows that drugX (green) is really only purchased by customers in cluster 2, 3, 4 and 10. DrugB is only
purchased
by customers in clusters 1, 5, 8 and 9. And drug Y is purchased by everyone. Can you imagine how
much more targeted
this company's marketing efforts can be based on this information? For example, you could
create a marketing message
for clusters 2, 3 and 10 that spoke to the benefits of drugs X and Y. This message
would be considered as being very
relevant to these customer while a message about drugC would not. Such
targeted messaging leads to increased
response rates, sales and customer satisfaction.
A large statistical software company called SAS has a number of case studies on response modeling and customer
segmentation that include the ROI from these types of studies. Click here to go to their site to see real world
examples
across all industries.
The challenge in using data mining is that you'll need to spend tens of thousands of dollars on the software
and hire a
statistician. Small businesses cannot afford this type of investment, even with the expected positive
return. Your real
alternative is to hire a consultant that has the experience and software to perform these types of
analyses for you.
Please call
or email us today to learn more about Marketing Analyltics and Stafford SBSG’s unique capabilities to
kick-start your
marketing capabilities today!
RFM is another analytical technique used by marketers. It is easier and cheaper to utilize than data mining and has yielded
positive ROI for companies for over 40 years. Click here to learn more. |