Measuring Customer Retention - Basic (Recent Repeater Model)
First Published: 03/16/01 DigiTrends
"A Model of
Future Customer Value"
Jim's Intro: Here's an example of a simple yet very effective
customer model for assessing the future profitability of your
customers, explained in 2 parts. Go to
Part 2 (Recency)
Part 1 (Frequency)
In the ongoing search for developing ways to measure the success of websites,
stickiness, or the amount of time a person spends at a web site, has been
put forth as an important concept to consider.
Those in favor of using stickiness to measure success say the longer a person
stays at a web site during a visit, the more interested and satisfied (loyal?) they are
with what the site is offering. Is this always true?
Let's think about that for a minute. Why would a visitor stay at a site
for a long time if they were not satisfied? Perhaps they're confused.
The navigation design is not clear. The content is poorly presented.
They went to make a ham sandwich in the kitchen.
So maybe stickiness (length of visit) is a good measurement; maybe itís
not. After all, if it seems like most people stay for a while, they
canít all be out in the kitchen. But this measurement works best for
very large sites, where you can average the behavior over many visitors, like
e-Bay, one of the current stickiness champions.
Thereís another way to look at stickiness, one more important than looking
at the time spent at a site (and more accurate). Itís the stickiness of
your site to visitors or buyers over time, and involves looking at repeat
visitors or buyers by customer segment. After all, someone who comes to your
site and spends a half hour there the first time but never comes back isnít
contributing very much to your success, particularly if you paid money to
attract him or her the first time.
Most sites track repeat visitors in the aggregate, meaning a visit to any
part of the site is a ďrepeatĒ. Some
track repeat buyers. This type of
reporting doesnít provide any real information; itís too generic.
You have to track the repeat behavior of customers sharing some other
variable, some characteristic that allows you to make judgments about the value
of the different customer groups and take some kind of marketing or design action based on this
For example, let's say you run a pet site. What if you found out the
cat section was pretty sticky over time (lots of repeat visitors) but the dog
section wasn't? What would that mean? Perhaps you need to pay more
attention to your dog content, and handle it like the cat content to encourage
repeat visiting. Or perhaps you shouldn't be a pet site at all; maybe you
should focus on being a cat site because you do cats better than anybody else.
Your visitors think so, because they repeat at a higher rate for the cat section
than the dog section.
Why does all this matter? Because it has been shown over and over
that past consumer behavior is the best predictor of future behavior. Past
behavior is a much better predictor of future behavior than demographics ever
will be. A visitor or buyer who repeats their behavior is more likely
to continue repeating it, meaning their future value to the business is high.
So when you look at a particular segment of customers, if repeating visitors or
buyers are rising, then your future business with this segment of customer will
be stronger than it is today. If repeaters are falling, business from this
customer segment will be weaker in the future.
When you look at the repeating behavior of different groups of customers, you
can make judgments about which customer groups will be most valuable in the
future. You want to do everything
you can to attract customers with high repeat behavior, and reduce or eliminate
any spending or other efforts on attracting customers with low repeat behavior.
Repeat behavior is a strong indicator of customer
loyalty to your site - it's a "likelihood to return" or
"likelihood to buy again" indicator.
So itís very important to understand which groups of customers have
high repeat behavior and which donít.
Hereís a good way to track repeat behavior.
Find the percentage of your customer base with a certain
characteristic that has visited or purchased more than once. For
visits, you might want to set a higher cut-off, maybe the percentage visited at
least 3 times. Then watch this percentage over time.
If the percentage of a certain type of repeating visitor or buyer is
rising, then your future business with this particular type of customer will be
stronger than it is today. If the
percentage of repeaters in a customer segment is falling, this part of your
business will be weaker in the future, and you need to take some kind of action
to get these people to come back or make additional purchases.
Changes in repeat percentage can be graphed and displayed
You can mix and match different behaviors and types of customers
according to what is important to your business model.
Rising and falling repeat rates act like an ďearly warning systemĒ,
providing important information about the future of your business with different
How should you divide up or ďgroupĒ your customers to analyze repeat
rates? There are certain variables
that affect the future value of a customer more than others.
Here are some of the most important groups to look at:
rate by media source of the customer - search engines versus banner ads, or
compare the repeat rate of customers generated by different banner ads.
There will be huge differences, and this metric can help you improve
the long term ROI of ad campaigns.
rate by category or item of customer's first purchase, and category of
ongoing product preference, is another huge differentiator of customer
behavior. This will tell you
which products are most profitable long term to feature or promote to new
and current customers.
rate by price of first purchase, a similar idea to the one above.
This will tell you what price range is most profitable to feature to
new customers, because customers buying in this range tend to repeat.
rate by content area favored by the customer during the first visit, and
repeat rate by ongoing preference to a content area. This is the cats versus
dogs example from the previous section. Which areas of your site create the most loyal (highest
repeat) customers? You should
feature those areas and not feature or eliminate areas that produce low
repeat customers, particularly when looking at first visit behavior.
rate by demographics or other self-reported data.
Grouping by non-behavioral data can sometimes be effective, depending
on how accurate the data really is. Of
all the groupings mentioned here, this is usually the weakest in predicting
There will be huge differences in repeat rate when examining the first
purchase, media source, and content experience the customer has with your site.
You will no doubt find others particular to your business.
These differences can be used to provide clues optimizing site design; you
should increase the opportunity for any experience which leads to a high repeat
rate, and decrease the opportunity for any experience leading to a low repeat
Repeat rate is a strong indicator of future activity.
Looking at repeat rate by customer segment is the first step to building
an effective customer model (profile
of customer behavior over time) you can use to drive higher profitability in
Part 2 (Recency)
While youíve got your hands dirty down there looking at
customer segments and repeat rates, it might be a good time to also take a look
at customer Recency by segment. Repeat
rate is a Frequency Model, the 2nd most powerful predictor of future
behavior, but one of the easiest to track and measure.
If you want to be able to predict the future actions and value of
customers even more accurately, you should also track Recency, the number of
days or weeks since a customer last engaged in a behavior.
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, visits, game plays, or
just about any ďaction-orientedĒ customer behavior.
Recency is why you receive another catalog from the same company shortly
after you make your first purchase from them.
Recency is the most powerful predictor of future behavior.
Think about it this way.
What good is it to have 10,000 people who have bought or visited at least
5 times when 80% have not repeated in the past 6 months?
Repeaters who havenít repeated Recently are former best customers.
Hereís an example of why the idea of Recency is so
important. Letís say youíre
looking at your repeater segments, and noticing the percentages are kind of flat
to slightly down over time, but still pretty good. Everything seems OK. But
what if most of the repeaters in the segment were from 6 months ago, and new
customers coming in, for whatever reason, were repeating less than customers
used to? The sheer number of
repeaters in your database might mask this decline in repeat rate.
The lower quality (lower repeat rate) new customer behavior is
overwhelmed by the behavior of older customers, and you get a false read on the
future health and profitability of your business.
I have seen this happen; and itís not pretty.
Everything seems to be going smoothly, and new customers are ramping, but
all of a sudden, sales or visits get soft.
This happens at the point where the new, lower quality customers finally
overpower the old, higher quality customers in the database.
The future value of your customers has shifted and you didnít know it
until it was too late.
Recency measurement solves this problem, because you are
always looking at whatís happening now with Recency. Using Recency as an additional filter in your
segment tracking clears away the past, and provides you with a headís up view
of the future. Tracking repeats (or
Frequency) by itself is a rear view mirror, because you never know how many of
these people are really current customers without looking at Recency.
The longer a customer has stopped engaging in a behavior, the less likely
they are to repeat the behavior - they become defected customers.
So how does Recency fit into the tracking of repeaters, how
is it implemented? The easiest way
to do it is to set a cut-off, say 30 days.
You want to track the percentage of customers in a segment who are
repeaters, where the last repeat action (purchase, visit, page view, download)
was in the prior 30 days. Watch to
see if the percentage is rising or falling, just as when you were tracking
repeat rate only. These Recent
repeat customers are your future, they are the strongest, most powerful,
most valuable customers, now and in the future.
They are the most likely to repeat whatever behavior you are tracking.
If these customers begin to decline in percentage, you will feel
it down the road. A decline in Recent repeaters means you have to readjust your
customer acquisition plans, because just to stay even, youíre going to have to
start replacing these best customers who are in the process of defecting from
Track the same segments discussed earlier Ė first
purchase, favorite purchase category, media source, first time and ongoing
content experience, and maybe demographics and other self-reported
characteristics. Look at Recent
repeat rate by any customer variable you have in the database!
You will begin to see concrete patterns in repeater Recency, patterns
which will help you make profitable decisions based on future customer value.
You want to do more of anything that creates high Recency repeaters, and
cut back on anything that doesnít create these high future value customer
segments. For a complete example of how to apply this model to ad
spending, see the tutorial Comparing the Potential Value
of Customer Groups.
The Recent Repeater customer behavior model is incredibly
simple, yet very effective in predicting the relative future value and loyalty of customer segments.
How to use this information to create high ROI marketing promotions and site
designs is explained step- by-step in the Drilling