Comparing the Future Value of Customer Groups
Jim's Intro: This short tutorial will provide you with all the information
to implement a simple future value scoring process. Use these scores
to compare the future value of customers from various ad
sources, buying certain products, or visiting specific areas of your
site. You will also learn how the concepts of Customer
Lifecycles, LifeTime Value, and ROI
fit into scoring the future value of customer groups. If you can compare future customer value, you
can allocate more spending to higher value customer groups.
This article assumes you have some background in customer marketing; you might want to read the short
articles on customer profiles and customer
models if you don't have experience with these concepts.
Over the past five decades, a lot of research and testing has
been carried out concerning the profiling of customer behavior based
on transactional data. The appearance of computers and
"data-mining" have allowed even more extensive studies to be
The end result? If you had to pick one variable to
predict the likelihood of a customer to repeat an action, Recency,
or the number of days that have gone by since a customer completed an
action (purchase, log-in, download, etc.) is the most powerful
predictor of the customer repeating this action.
As each day goes by after the customer completed the action, the
customer gets less and less likely to repeat it. Plain and
simple. You can run all the fancy data-mining scenarios on
"likelihood to buy" or "likelihood to visit" you
want to - Recency always comes up as the most important variable in
predicting the likelihood of a customer to repeat an action.
Recency is the number one most powerful predictor of
future behavior. The more
recently a customer has done something, the more likely they are to do it again.
Recency can predict the likelihood of purchases, log-ins, game plays, just about any “action-oriented” customer behavior.
Recency is why you receive another catalog from the same company shortly
after you make your first purchase from them.
They know you are most likely to order again immediately after your first
order. Recency is the most powerful predictor of future behavior.
It should not surprise you that Recency is also the most powerful
predictor of a customer to respond to a promotion - after all, the
more likely a customer is to repeat an action, the more likely they are to
respond to a promotion asking for this action (purchase, log-in,
If a Recent customer is more likely to repeat an action, and is
more responsive to promotions for this action, it follows the more Recent a customer is,
higher their future value, because Recent customers are the most
likely to contribute to profits in the future by responding to your
promotions (or simply just coming back by themselves).
Customers who are more Recent have
higher future value than customers who are less Recent, for any given
activity. Customers who made a purchase 15 days ago have higher
future value than customers who made a purchase 60 days ago.
Customers who logged in last week are much more likely to visit than
customers who logged in 30 days ago, and so have higher future
Make sense? Great. But how
is Recency implemented, how do you actually do anything with this information?
Glad you asked. Let's use Recency to compare the future value of
customers coming from two different ads (Ad #1 and Ad # 2) that ran at
the same time, for the same duration.
1. Identify the groups you want to compare for future value. In
this example, it's the customers who clicked on either of two ads, Ad #1 or Ad
#2 (two groups).
2. Decide which activity is most important to you for these groups. If you're a publisher,
log-ins or page views are most important. If you are selling merchandise,
you would use purchases. For this example, we will use purchases. An
example using visits (or log-ins, if you don't track visits) is below.
3. Create a database (or set up for a query of one, depending
on how your resources are structured) of all the purchases
people who clicked on Ad #1 or Ad #2 have made. These
transactions need to be date stamped; most interactive activities are
so this should not be a problem. If an activity you want to
profile for future value has no date stamp, start collecting the
dates of activity.
4. Pick a Recency cut-off. For page views, it might be 1 week;
for purchases, maybe 30 days. It doesn't really matter, because you are
interested in a comparison of the activity between the Ad #1 and Ad #2
groups, not an absolute number. Pick something reasonable
based on what you know about your customers. Let's use 30 days for
5. Query your database and find out what percentage of the people
clicked on Ad #1 and made a purchase have made at least one purchase in the past 30 days.
You might come up with 20%.
6. Run the same analysis for people who clicked on Ad #2 and
made a purchase.
Let's say only 15% of these people have made at least one purchase in
the past 30 days.
7. You're done, and you know the answer. A higher
percentage of people who clicked on Ad #1 are Recent - active and
purchasing - when compared with Ad #2. This means Ad #1
generates customers with a higher future value. You need to take this
into account when analyzing the success of the ads.
Do you understand how powerful this idea is?
If you go through this process for customers grouped by which
products they buy, you can determine which products generate customers
with the highest future value.
Go through this process for customers grouped by which area
of the site they visit most, and you will determine which areas generate
highest future value customers.
If you go through this process for customers grouped by the
demographics or the survey data they provide, you can determine which
data points define customers
with the highest future value.
You can track multiple activities for the same customer
groups. In the first example, you found customers who clicked on Ad #1
and made a purchase are more Recent on purchases, so they have a
higher future value on the activity
"purchases". But what about the Recency of
people who clicked on the ads for visits? If they keep coming
back, they could be of some future value. Let's see how this
Recency study might look.
1. Visits / log-ins example:
Create a database of all the visits (or log-ins if you don't track
visits) people who clicked on Ad #1 or Ad #2 have made (need date
2. Pick a Recency cut-off. Again, we are interested in
a comparison, so the number isn't critical. Let's use 1
3. Query your database and find out what percentage of the people who
clicked on Ad #1 have visited (logged-in) at least once in the past week.
You might come up with 10%.
4. Run the same analysis for people who clicked on Ad #2.
You might come up with 30% who have visited / logged-in at least once
in the past week.
5. You're done, and now you have an interesting
situation. It appears the customers who clicked on Ad #1 have a
higher future value on purchases, but people in general who clicked on
Ad #2 have a higher future value on visits. Maybe they're just
tire kickers, or maybe they're doing research. We'll
take a closer look at finding answers to this situation on the next
page of the tutorial.
Note that this method is based on the actual facts of customer
behavior - not speculation or "best guess"
theories. The behavior of the customer is the most accurate
yardstick you will find for assessing future value.
Once you complete studies like these, you can begin to organize all
your business practices around the future value of the customers they
generate. If you allocate money away from activities generating
low future value customers, and allocate this money to activities
generating higher future value customers, you will become more
profitable over time. It's really as simple as that.
in Tutorial: Adding Customer LifeCycles to the
Read advanced version of
Maximizing the value of customers using Recency in many different
ways is explained in