Measuring and

Managing Customer
Lifetime Value:
A CLV Scorecard and
Cohort Analysis in a
Subscription-based Enterprise
BY MASSIMILIANO BONACCHI, PH.D., AND PA O L O P E R E G O , P H . D .

This article is based on research supported by the IMA® Research Foundation

COMPANIES FREQUENTLY ENCOUNTER DIFFICULTY MEASURING CUSTOMER LIFETIME

VALUE (CLV) AND CUSTOMER EQUITY (CE) BECAUSE OF THE LACK OF A PRACTICAL

FRAMEWORK AND THE ANALYTICAL CHALLENGES BEHIND THE CLV PARADIGM.

THIS ARTICLE PROPOSES STEP-BY-STEP GUIDELINES TO MEASURE AND MANAGE CLV IN A

SUBSCRIPTION-BASED ENTERPRISE (SBE). A CASE STUDY DEMONSTRATES HOW A

CLV SCORECARD AND COHORT ANALYSIS ARE PRACTICAL TOOLS THAT CAN

SUPPORT DECISION MAKING.

he last few years, companies have increas- HomeAway, LinkedIn, Pandora, Skype, and Zynga.

T ingly turned to subscription-based models for

offering services such as mobile telephones,
cable television, software, and e-banking.
The growth of the Internet and mobile
devices has expanded the types of innovative services
that are offered in a contractual setting (e.g., music,
games, movies, and e-books) and has brought about a
In subscription-based enterprises (SBEs), a customer
pays a fee to have access to the firm’s products or ser-
vices for a certain period of time. For an SBE to suc-
ceed, managers need to gain a more nuanced
understanding of the strategic, financial, and operational
implications of a subscription-based model. As high-
lighted by V. Kumar and Bharath Rajan, managers need
new wave of Web start-ups, including FriendFinder, performance measurement reports that are able to con-

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 27 FALL 2012, VOL. 14, NO. 1

vey information they can use to diagnose the health of such as “free ringtones” or “free music.”3 Figure 1 pre-
their business and that can assist them in making strate- sents the industry value chain. It typically is character-
gic and tactical decisions such as: ized by the following activities: content origination,
◆ Which type of actual customer or future prospect service management, marketing and display, network
should we retain, grow, or acquire? delivery, customer relationship management, and
◆ How much should be spent on various customer seg- billing.
ments to retain, grow, acquire, or win back customers? Company.net follows a common SBE business strate-
◆ Which advertising channels are most effective and gy: (1) acquire new customers through aggressive mar-
efficient? keting techniques of customer acquisition aimed at
◆ What is the value of the customer base (i.e., the building up a new user base; (2) retain existing cus-
most important asset for this kind of company)?1 tomers by measuring the lifetime of users and their
value with techniques that typically stimulate user
To address these issues, we refer to the notion of cus- retention and minimize churn rate; and (3) once the
tomer equity (CE), which is the sum of the customer user base is built and the churn rate is under control,
lifetime value (CLV) across a firm’s entire customer the focus shifts toward planning for the organic growth
base. CLV can be enormously informative when calcu- of the user base and defining the new client target
lated correctly, but most companies today calculate CLV acquisition number for each period (e.g., month or quar-
improperly either because they do not connect the ter) in order to balance churn rate and reach the target
measure with its drivers or because they do not growth rate.
acknowledge customer lifetime heterogeneity.2
Academic and practitioner literature has repeatedly T H E P E R F O R M A N C E M E AS U R E M E N T SYS T E M
demonstrated the beneficial effects of a customer- A company’s performance measurement system shapes,
centric strategy, but anecdotal evidence suggests that and is shaped by, its strategy.4 At Company.net, the per-
few companies practice such an approach in a system- formance measurement system has been built around
atic and effective manner. We believe one of the rea- CLV and CE. CLV is the value of future cash flow/prof-
sons is a lack of practical guidance. In this article, we it attributed to a single customer or a group of cus-
attempt to fill this void by presenting a case study of a tomers discounted using the company’s average cost of
mobile content provider. We focus our attention on a capital. It is a forward-looking metric that drives cus-
CLV scorecard and cohort analysis—two tools that man- tomer profitability, particularly when a firm has to
agers and management accountants in SBEs can use to decide which customers to acquire (CLV is the upper
gain a better understanding of the way they measure bound of expenses that a firm should incur to acquire a
and manage customer profitability. customer), what customers to nurture (managers should
focus on customers with high CLV), and the quantity of
C O M PA N Y B A C K G R O U N D resources to allocate (marketing resources should be
Company.net (the firm’s real name has been disguised allocated so as to maximize CLV).5 As stated previously,
for confidentiality purposes) is a typical example of a CE is the sum of CLV across all customers of a com-
fast-growing SBE. A customer pays a subscription fee to pany, both existing customers and future customers. CE
receive a specified number of downloads of content, is the most important asset for an SBE, and it is influ-
namely ringtones and music MP3s, to a mobile phone enced by the ability to acquire, retain, and increase the
or device. Company.net works with all the major record customer base. While several methods of computing
labels, which supply content, and it has agreements CE have been suggested, we will focus on two metrics
with the most important phone carriers to deliver con- that can be used when evaluating the expected prof-
tent to the network and bill customers through their itability of a company’s customer base:
mobile accounts. Customers are reached predominantly 1. Current customer equity (CEcur), which is the
through paid search advertisements using keywords sum of the future profit margins generated from

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 28 FALL 2012, VOL. 14, NO. 1

 Figure 1: Entertainment Industry Value Chain of Company.net

Company.net

Mobile
Content Content & Media
Telcos
Owners Service Companies
Providers

Content Service Marketing Network

Origination Management & Display Delivery User

Research h and Aggregation and Marketing and Network k used

use
ed
development
m publishing of promotion of to distribute
distrib
butee
activity content and VAS/music VAS/mu usic tto
VAS/music Revenues
services in through media: end users
useers
r
a
Design and ect
project
di erent formats TV, print, or (Mobile Cost of Delivery
of the product/
(www, text, MMS, online operators)
t )
service
WAP, etc.) networks Contribution
Margin
Cost of Acquisition
Customer Margin
Fixed Cost
Earnings Before
Interest and Taxes

the customers who have already been acquired by drivers are cost of customer acquisition (CoA) and life-
the end of the period.6 time value (LTV).
2. Total customer equity (CEtot), which is the sum
of the future profit margins generated from cur- Cost of Customer Acquisition (CoA)
rent (CEcur) and future (CEfut) customers of the CoA is a straightforward metric. In this setting, it can be
firm.7 estimated in a very precise way for every single cus-
tomer. Company.net uses search engine advertising
From a management accounting perspective, the that measures cost per click and cost per acquisition
challenge is not only to measure CLV—and thus CE as metrics—as opposed to banner advertising (measured
a summation of each customer’s CLV—but also to man- by a cost per impression model) or traditional channel
age its drivers. Looking at the relationships among advertising such as magazine and TV ads. Because CoA
these drivers is crucial because the variables of the CLV is influenced by the type of marketing policies, its
formula are interdependent. When a metric is pushed analysis is the first step for managerial actions that will
in one direction, it becomes more difficult to ensure impact the company’s attraction and conversion rate. In
that the other keeps the same pace. A managerial tool other words, it can lead to better targeting, better adver-
to consider the interaction among CLV and its drivers is tising, better landing pages, optimization of the flow
the CLV scorecard.8 Figure 2 shows the CLV scorecard through to checkout, more and better payment options,
Company.net used. As you can see, the two key CLV and so forth.

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 29 FALL 2012, VOL. 14, NO. 1

 Figure 2: Customer Lifetime Value Scorecard at Company.net

Measures
Average
Revenue
per User (ARPU)
CONTRIBUTION
Yield
MARGIN
Cost of
service
LIFETIME VALUE
(LTV)

Instant churn

LIFETIME
CUSTOMER Historic churn
LIFETIME VALUE
(CLV)

Attraction
COST OF
ACQUISITION Conversion
(CoA) Cost of contact
(cost per click,
cost per acquisition)

Feedback

Lifetime Value (LTV) Company.net uses two approaches:

Determining LTV requires a more sophisticated analy- 1. Historic churn is the number of subscribers who
sis of its main drivers, which are lifetime, margin, and canceled during the period N (day, week, or
yield. In order to manage LTV, a company needs to month) who initiated their subscription before that
carefully examine the relationship between the metric period (i.e., their join date < quit date). Here the
and its drivers. This relationship can be synthetized by ratio is computed as the number of subscribers
the equation: who quit in period N divided by the subscribers
Lifetime Value = Lifetime ✕ Margin ✕ Yield.9 active at the end of period N+1. Historic churn is
Lifetime is the period during which a customer stays the figure used to transform churn in lifetime.
with the company. This metric is a function of the rate 2. Instant churn is the number of subscribers who
of attrition (cancellations/average users per period) over canceled the service during the period (day, week,
a period of time that subscription-based customers or month) in which they had initiated their sub-
“churn out” (unsubscribe) from the customer base. scription (i.e., join date = quit date). This is calcu-
Churn is a proxy of the customer satisfaction of the ser- lated using gross addition, which is the number of
vice. The underlying rationale is that there is a fairly new subscribers in the period. The ratio is calcu-
simple relationship between churn per month and the lated as: Subscribers who quit in period N/Gross
number of months that customers stay with the addition of period N. This metric is devised to
company—in other words, a ratio of 1/churn. Thus, a judge the quality of specific marketing and adver-
2% churn means 1/0.02, or an average customer dura- tising decisions.
tion of 50 months. To get a good grasp of churn,

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 30 FALL 2012, VOL. 14, NO. 1

 Table 1: Company.net Internal Report on LTV Drivers

Metric INSTANT HISTORIC YIELD MARGIN

Dimension CHURN CHURN

COUNTRY

SUBSCRIPTION FLIGHT

TELECOM OPERATOR

Margin is the contribution margin per customer. It is customers in order to gauge the effectiveness of a sub-
equal to the average revenue per user (ARPU) minus scription campaign on a specific product or service. For
the cost of service per customer. For example, a $10 example, a customer acquired through Google AdWords
subscription to a mobile service (such as ringtones) at a could have a different churn rate than a customer
per-customer service cost for content delivery of $4 acquired through Yahoo! My Display Ads or Facebook.
would generate a margin of $6. Furthermore, customers who are acquired from a specif-
Yield is defined as the ratio between subscribers suc- ic mobile operator (such as AT&T, Sprint, or T-Mobile)
cessfully billed and customer candidates to be billed. could have different attitudes toward churn and differ-
For instance, if 1,000 customers signed up but only 550 ent yield.
actually pay, the yield would be 55%. These drivers are leading, forward-looking indicators
and must be monitored on a daily basis to ensure the
Monitoring LTV company is heading in the right direction. Company
Monitoring these metrics is useful for evaluating strate- presentations often show all customer metrics improv-
gic and tactical choices, including different advertising ing as they look toward the future. This is unlikely to
campaigns, the launch of a new product, the launch of play out in reality. For example, if a firm attempts to
an existing product in a new country, cross-sell and raise ARPU (price), it will necessarily increase churn.
upsell campaigns to increase ARPU and improve mar- Similarly, if a firm aims to grow faster by spending more
gin, and so forth. on marketing, CoA will likely rise. Churn may increase
The typical Company.net report on LTV drivers as well because a more aggressive marketing campaign
focuses on a product or service (e.g., music, ringtones, will likely capture customers of a lower quality.
and other value-added services) and includes country, The availability of customer data and the particular
subscription flight (i.e., the scheduling of advertising for nature of the business—in which all transactions are
a period of time), and telecom operator used to deliver made online and the log files of each transaction are
the service (see Table 1). Comparisons at the country constantly registered—is a prerequisite for timely
level are used to evaluate the success of a product that management control of these metrics. This allows
was already launched in other countries. The type of Company.net to judge the effectiveness of its marketing
marketing campaign is linked to the churn of acquired investments.

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 31 FALL 2012, VOL. 14, NO. 1

BENEFITS OF A C LV S C O R E C A R D ing also ensures that a firm is not paying excessively
A CLV scorecard makes it possible to examine past cus- for clicks. For instance, where the listing position is
tomer behavior and understand the drivers of CLV. For determined by how much an advertiser is prepared
example, it can be useful to study how the CLV or CoA to pay for a keyword or phrase, being in second or
of an average customer acquired through Facebook dif- third position on a search page may generate as
fers from an average customer acquired through Google much business as being in first position.
ad campaigns. In other words, the analysis of the CLV ◆ Different customer characteristics. Although
drivers allows managers to test cause-and-effect rela- Company.net managers are interested in computing
tionships between managerial actions and results in the CLV of their customers, they are similarly keen
terms of both customer metrics (churn, CoA, margin, on identifying the drivers of a profitable duration in
and yield) and financial results (CLV and CE). their customer-company relationships.12 Specifically,
Company.net, for instance, regularly monitors these the telecommunication carrier can be considered a
objects or units of analysis: proxy of income. Moreover, a longer subscription
◆ Campaigns currently active. Different channels give length is a proxy of customer satisfaction because it
different results in terms of CLV. For instance, a cus- is well documented in the literature that the longer
tomer acquired through Google or Facebook could subscribers remain with a company, the lower the
have a tendency to churn that is different from a probability that they are going to churn.13
Yahoo! or Bing customer. The analysis can be con-
ducted by examining the retention behavior (instant Based on the analysis of Company.net log files, the
and historic churn) of customers acquired through main determinants of a profitable lifetime duration are
these channels in order to understand which lifetime the type of telephone operator, the length of subscrip-
those users project. tion, and the type of advertising campaign that has con-
◆ Country breakdown (if the same product or service is vinced the person to become a new customer. Stated in
provided internationally). This information is quite mathematical terms, we can say that a profitable life-
useful when a service is introduced in a new country time duration for Company.net is a function of tele-
because data gathered in a similar country provides phone operator, length of subscription, and advertising
an appropriate benchmark. For example, the pay- campaign.
back of marketing investments in Brazil could be a
good proxy to project what will happen if the firm C O H O R T A N A LY S I S
invests in offering the same services in Chile. Although churn is the main driver of lifetime, and
◆ Different time frames to monitor different consumer churn data is crucial for judging the success of a cus-
behavior. In this way, a daily or hourly fine-tuning of tomer acquisition campaign once it is complete, market-
a marketing campaign could be performed, for ing literature raises serious concerns regarding the
example, by changing the bid price on the keyword typical approach of projecting an historic churn rate into
advertising.10 In the paid search model described by the future in order to get the lifetime of acquired cus-
Des Laffey, monitoring the behavior of click- tomers and then their profit streams (i.e., LTV). Sunil
throughs via paid search is essential as it provides a Gupta, Donald Lehman, and Jennifer Stuart demon-
precise measurement of the success of the advertis- strate that the widespread method of converting reten-
ing method in terms of achieving the objectives set tion rate to expected lifetime (1/churn rate) and then
forth.11 Data collected from such tracking should calculating the present value over that finite time peri-
then be fed back into the process to make perfor- od ultimately overestimates LTV.14 Peter Fader and
mance reviews more effective. It can help identify if Bruce Hardie demonstrate that the retention rate
poor-quality prospects are being attracted (perhaps (defined as the opposite of the churn rate) is an increas-
through the use of the wrong keywords) or if clicks ing function of time.15 Therefore, the longer sub-
from some sources work better than others. Monitor- scribers are with the company, the lower the probability

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 32 FALL 2012, VOL. 14, NO. 1

that they are going to churn. Researchers explain this included as part of CLV. Yet customer acquisition cost is
phenomenon with two main reasons.16 The customer’s often displayed alongside a customer’s LTV. In this way,
preference or satisfaction with a product or service Company.net gains insights on whether an unprofitable
increases over time as he or she uses it, thus decreasing customer’s CLV (LTV minus CoA) is negative because
churn. Alternatively, an increasing retention rate may be of high costs of acquisition rather than a low LTV.
because of cross-sectional heterogeneity in individual To compute MpK, take the following steps (see the
retention probabilities across customers in their prefer- sidebar, “Cohort Analysis at Company.net,” for an
ence for the product or service.17 The heterogeniety example):
problem can be particulary severe if the company is 1. Collect the gross additions (acquired subscribers)
showing substantial growth. By definition, growth for each cohort and the corresponding number of
implies many first-time customers, and their churn can billings (subscribers who effectively pay) in each
be different from the churn of older customers, thus period. Notice that gross addition is always higher
distorting the view of how much repeat purchase than the number of subscribers billed for two rea-
behavior will occur in the future. sons: First, some of the subscribers quit (this phe-
Company.net’s business intelligence unit recognized nomenon is measured by churn rate). Second, it is
and addressed these issues. When it evaluates the not possible to charge all the candidate sub-
potential profit from future customers, the company scribers. This is measured by the yield.
uses a heuristic methodology to estimate LTV from 2. Multiply the number of billed subscribers by the
future acquired customers. The underlying rationale is margin per customer in order to get MpK.
to use a data-mining approach to project the data of 3. Estimate the average MpK among the different
each acquired customer’s cohort. Cohort analysis has cohorts for each period of time considered. Averag-
been used by statisticians for decades, and recent ing across cohorts provides an average MpK at the
advancements in data collection and processing power end of one month, two months, and so forth
have made it a viable technique for online businesses to (Equation 1).
study customer loyalty trends, predict future revenues,
∑ t=1 #_Billingt
1,000 1 n
and monitor churn. The most popular cohort analysis Equation 1: MpKt = margin gross_additiont n–(t–1)
(which we present here) involves segmenting customer
groups based on a join date. The month, week, or day Where: t = period of time
of that date then becomes the user’s cohort, meaning n = total periods where data is available
each cohort is the set of users who joined during the margin = ARPU – COGS
same time period. #_Billing = paid subscribers for each cohort
The pivotal metric used in this analysis is called
margin per thousand customers (MpK), which focuses The estimate also requires a normalization of
on the projected margin.18 Basically, Company.net esti- the data (i.e., outliers are excluded from the aver-
mates across cohorts an average MpK for the last N age). In most cases, outliers are easily identified
months of actual data. After normalizing the data based based on past experience. Most are the result of
on previous observations, it projects the average MpK problems in the firm’s information system that
in the future months. The metric is calculated with dif- produces the log files.
ferent windows, namely daily, weekly (preceding seven 4. Project the average MpK on the lifetime of the
days), and monthly (preceding 30 days). In all three customer in order to estimate how much margin
cases, a notion of CLV with the following characteristics can be obtained from the acquisition of 1,000
is considered: (1) LTV is based on future profit,19 and customers today. This step makes it possible to
(2) CLV does not directly take into consideration the estimate the distribution of the MpK across cus-
CoA.20 Formally, CLV is computed by subtracting CoA tomers’ lifetime (Figure 3 in sidebar). As the
from LTV. In other words, acquisition costs are not cohorts mature, there are fewer data points to

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 33 FALL 2012, VOL. 14, NO. 1

Cohort Analysis at Company.net
This example hypothesizes the launch of an online service in a new country at the beginning of the year. The customer
acquisition and collection are described in Table 2. The firm acquired an initial cohort of 25,016 customers in the first
month. These customers produced the following stream of payment: 8,614 billing events in the first month, 13,437 in
the second month, and so on. Another cohort of 38,862 customers was acquired in the second month, which gener-
ated a separate stream of billing events (15,086 in the first period, 27,082 in the second, and so on).

Table 2: Customer Acquisition (Gross Addition) and Collection (Billing Events)

Cohort Period of Time January February March April May

Gross Addition Number of Customers Billed

Cohort 1 25,016 8,614 13,437 15,516 12,658 9,316

Cohort 2 38,862 15,086 27,082 22,085 15,996
Cohort 3 54,985 32,380 35,906 26,036
Cohort 4 68,099 36,220 35,276
Cohort 5 51,851 19,798

Note: Figures were disguised by a constant multiplier for confidentiality reasons.

For the calculation of MpK, it is necessary to complete the scenario with ARPU as the input variable. The example
covered here uses an ARPU of $10, which is considered to be booked at the beginning of the contract period, and a
cost of service of $3.50, which makes a contribution margin of $6.50. The customer acquisition cost is $11 per customer
(this figure is estimated looking at the ratio of marketing cost/gross addition).
At the end of May, managers raise the following questions:
1. All else being equal, what is the expected contribution of the next 1,000 customers we will acquire?
2. What is the marketing investment’s payback period?
3. If we stop investing in this market or service, what is the expected residual value of the customers currently on the
firm’s books (i.e., customer equity of current customer base)?
To answer these questions, a report can be developed that follows the MpK rationale. The first step is to transform
billing events in monetary terms by multiplying billing events by margin and scaling the result by 1,000 customers
(Table 3). For example, the January figure ($2,238) has been obtained by multiplying 8,614 (January’s billing) by $6.50
(Margin). This is then multiplied by (1,000/25,016) in order to scale by 1,000 customers.

Table 3: Evolution of Gross Margin for Each Cohort

Cohort Period of Time January February March April May

Gross Addition Gross Margin X # customers billed scaled by 1,000 customers

Cohort 1 25,016 $2,238 $3,491 $4,032 $3,289 $2,421

Cohort 2 38,862 $2,523 $4,530 $3,694 $2,675
Cohort 3 54,985 $3,828 $4,245 $3,078
Cohort 4 68,099 $3,457 $3,367
Cohort 5 51,851 $2,482

Note: Figures were disguised by a constant multiplier for confidentiality reasons.

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 34 FALL 2012, VOL. 14, NO. 1

 The next crucial step requires computing an average contribution margin estimated for each period. The question is:
“How much margin should be expected from 1,000 customers acquired today?” Given Equation 1, it follows that 1,000
acquired today are expected to provide a gross margin of $2,905.65 in the next month (average of $2,238; $2,523;
$3,828; $3,457; $2,482); $3,908 in the second month (average of $3,491; $4,530; $4,245; $3,367), and so forth (see Table
4). As the cohorts age, there are fewer data points to average. Hence, for the subscribers who began in January, there
are five months of data; for the subscribers who began in February, there are four months of retention data, and so
forth. The number of actual data (i.e., number of cohorts) depends on data availability, with the usual range from a
minimum of two to a maximum of 12 months of past data. Regarding the normalization of the data, the analysis of the
standard deviation allows detection of cases in which the average is biased from outliers. In those cases, the outliers
are removed from the average.

Table 4: Projection of MpK in the Future

Time Period MpK Standard Deviation Rate

1 $2,906 $694 35%

Average

2 $3,908 $567 - 8%
of past
data

3 $3,601 $484 - 17%

4 $2,982 $434 - 19%
5 $2,421 - - 17.9%
6 $1,988 - 17.0%
(rate x 95%)

7 $1,650 - 16.1%
decrease
Forecast

t … …
35 $123 - 3.8%
36 $118 - 3.6%

A typical pattern found in Company.net is that, following an initial period of time, MpK tends to level off month by
month. With such a pattern, the company can extrapolate forward using the same month-by-month decrease across
several months. To forecast the evolution of MpK for the cohort for which data is not available (Cohort 6 and beyond in
this example), the following algorithm is applied: The drop from the previous month is multiplied by the percentage,
an x% of decrease, to account for a decrease in churn rate (in this example, the percentage of decrease is 95% for each
month). This example includes five months of data. That is extrapolated forward using the same month-by-month
decrease (on the basis of previous experience) for the subsequent 31 months. Stated differently, MpK of period 6
(1,988) is obtained as follows: 2,420.61 x (- 19% x 95%). This method allows Company.net to project a cohort’s MpK in
the future, estimating the distribution’s tail (as exhibited on the right side of Figure 3). At Company.net, this estimation
procedure usually does not exceed 36 months when forecasting MpK.
Table 4 provides an answer to the managers’ first two questions. Namely, 1,000 customers acquired today would
produce a margin of $31,121 ($2,906 + $3,908 +…+ $123 + $118) during the next 36 months. This result is obtained with

∑ MpK
the following formula:

36

MpK36 months = t
t=1

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 35 FALL 2012, VOL. 14, NO. 1

This figure must then be compared with CoA to obtain a CLV of $20,221 ($31,121 – $11,000), as illustrated in
Table 5.
The payback of the marketing campaign is obtained from Table 4 by summing up MpK until $11,000 (i.e., $2,906 +
$3,908 + $3,601 + $2,982). This is quite relevant information when firms have to optimize the resources spent on cus-
tomer acquisition and evaluate the risk of a marketing investment.

Figure 3: Distribution of MpK for the Average Cohort

$4,500

$4,000

$3,500

$3,000

$2,500

$2,000

$1,500

$1,000

actual future
$500

$0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

GMpK

Note: The graph plots the distribution of the monthly margin of the “next” 1,000 customers being acquired.
The vertical line at the fifth month separates the estimation based on actual data (first five months) from the
projection of the tail applying the previous monthly change and a certain rate of decrease (e.g., 95%).

Table 5: Typical Report to Make Decisions

INPUT

ARPU COGS incidence Margin CoA estimate Payback period

$10 35% $6.50 $11 4.0

OUTPUT

MpK6months $17,805
MpK12months $24,621
MpK36months $31,121
Cost of Acquisition $11,000

Note: The report highlights the relation between CoA and CLV (estimated using MpK). These kinds
of reports are crucial in order to judge the contribution of new subscribers to value creation. The
value of future customers and the payback period alone, however, are not sufficient to obtain a full
picture. The other piece of information required to make a rational decision in this business setting
is customer equity (CE).

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 36 FALL 2012, VOL. 14, NO. 1

 To answer the third question, a further step is needed to estimate the CE of current customers, extending the actual
data of each acquired cohort (from 1 to 5) with the estimation for the future. This means that data in Table 6, column A
must be projected into the future using data from Table 6, column B. Hence, for each cohort we sum up the future peri-
ods (i.e., for Cohort 1, the period 6 to 36; for Cohort 2, period 5 to period 36; for Cohort 3, period 4 to 36; for Cohort 4,
period 3 to 36; for Cohort 5, period 2 to 36). The last step is to unscale the data in order to consider the real gross addi-
tion for each cohort because all the calculations made so far refer to 1,000 customers. The result is a CE of the current
customer base of $5,328,426 (Table 6). A CE of $5,328,426 can be considered as the value embedded in the acquired
customer base.

Table 6: Estimation of Customer Equity for the Current Customer Base

C=
A January February March April May B (A/1,000)*B

Customer Customer
equity equity of
scaled per current
Gross Margin X # customers billed scaled 1,000 customer
Addition by 1,000 customers customers base

Cohort 1 25,016 $2,238 $3,491 $4,032 $3,289 $2,421 $15,303 $ 382,820

Cohort 2 38,862 $2,523 $4,530 $3,694 $2,675 $17,724 $ 688,790
Cohort 3 54,985 $3,828 $4,245 $3,078 $20,706 $1,138,519
Cohort 4 68,099 $3,457 $3,367 $24,307 $1,655,282
Cohort 5 51,851 $2,482 $28,215 $1,462,976
Customer 238,813 $5,328,387
base

Note: Customer equity scaled per 1,000 customers is the value at the end of May of 1,000 customers
acquired in each of the five cohorts. Customer equity of current customer base is obtained from the sum of
the unscaled customer equity of each cohort.

Because of the normalization of the data, the performance measurement system of Company.net is also able to sim-
ulate the effect of other factors that are not under the control of the company, including competitor actions, regulation
changes, or technological discontinuities—any of which may affect the consumer behavior in both conversions and
retention. This is particularly important when it is necessary to neutralize the effect of these factors from the metric
utilized as the base for managers’ bonuses. For example, consider the case of a problem in the log files transmitted
from the telecommunications carriers to Company.net for a specific cohort.

(NOTE: A copy of the Excel file used in the analyses is available from the authors by request.)

average across, so the potential for error increases. metrics are computed across different time
Nevertheless, it is still a useful exercise to get an ranges:
idea of future margin associated with the acquisi- ● MpK1month = MpK1: expected margin of the next
tion of 1,000 customers. How far into the future 1,000 customers in the first month.
MpK6months = ∑ t=1MpKt : expected margin of the
6
the estimation can be extended depends on the ●

type of business. At Company.net, the following next 1,000 customers after six months.

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 37 FALL 2012, VOL. 14, NO. 1

 MpK36months = ∑ t=1MpKt : expected margin of the
36
● Paolo Perego, Ph.D., is an assistant professor at the Rotter-
next 1,000 customers after 36 months dam School of Management, Erasmus University Rotter-
dam. He can be reached at pperego@rsm.nl.
In theory, CLV models should estimate the value of
a customer over the customer’s lifetime, but many The authors gratefully acknowledge the financial support by
firms, including Company.net, consider three years to the IMA® Research Foundation. The authors are indebted to
be a reasonable estimate for the horizon over which the Massimiliano Pellegrini (CEO of Dada Entertainment Inc.)
current business environment (with regard to tech- for his insights on the subscription-based business model and
nology, competition, and so forth) would not change to several managers of the case company for their valuable
substantially.21 feedback and for making the study possible.

TA K E A W AY E N D N OT E S
1 V. Kumar and Bharath Rajan, “Profitable Customer Manage-
The goal of this case study was to provide management
ment: Measuring and Maximizing Customer Lifetime Value,”
accountants with a better understanding of how to Management Accounting Quarterly, Spring 2009, pp. 1-18.
ascertain the preferences of new and latent customers 2 Peter S. Fader, Customer Centricity: Focus on the Right Customers
for Strategic Advantage, Wharton Digital Press, New York, N.Y.,
in a typical subscription-based business model. It 2012.
involves directly observing the customer’s purchase 3 Paid search advertising entails advertisers competing for top
behavior and subsequently linking this data to customer listing position by bidding in ongoing auctions and then pay-
ing when users click on the advertisements, making paid
value and firm performance. search a flexible and accountable form of advertising (see Des
Calculating CLV (and, by extension, CE) is not Laffey, “Paid Search: The Innovation that Changed the Web,”
Business Horizons, 2007, pp. 211-218). Pay-per-click ads are one
enough because managers need to examine CLV dri- of the most cost-effective advertisement tools. They are the
vers using a framework that makes cause-and-effect “sponsored ads” that are displayed at the top and right of the
relationships between managerial actions and key cus- search results on Google, Facebook, Yahoo!, Bing, and similar
websites. Payment for these ads occurs whenever someone
tomer metrics visible. There is no such thing as an aver- actually clicks on an ad, not when it gets displayed. The cost
age customer lifetime. Because the typical survival per click can range anywhere from a few cents to several dol-
lars, depending on the type of industry and keywords.
curve drops quickly and then levels off, better-informed 4 Christopher S. Chapman, Controlling Strategy: Management,
decisions can be made if customer groups are seg- Accounting, and Performance Measurement, Oxford University
mented based on a “join date.” Failing to do so will Press, New York, N.Y., 2005; Robert H. Chenhall, “Manage-
ment Control Systems Design within Its Organizational Con-
undervalue a firm’s CE. text: Findings from Contingency-based Research and
For every decision, such as the launch of a service in Directions for the Future,” Accounting, Organizations and Soci-
ety, 2003, pp. 127-168; Jean-François Henri, “The Periodic
a different country, a new advertising campaign, and so Review of Performance Indicators: An Empirical Investigation
forth, managers need to rely on reports showing how of the Dynamism of Performance Measurement Systems,”
that decision will impact future CLV and CE. This will European Accounting Review, 2010, pp. 73-96; Ralph Kober,
Juliana Ng, and Byron J. Paul, “The Interrelationship
allow decision makers to better answer questions about Between Management Control Mechanisms and Strategy,”
the types of customers to retain, grow, or acquire, as Management Accounting Research, 2007, pp. 425-452.
5 Kumar and Rajan, Spring 2009.
well as how much should be invested to do so. The use 6 Julian Villanueva and Dominique M. Hanssens, “Customer
of CLV and CE also helps managers understand which Equity: Measurement, Management, and Research Oppor-
tunities,” Foundations and Trends in Marketing, 2007,
advertising channels are more effective and efficient
pp. 1-95.
and how to increase the value of the customer base. ■ 7 CEcur can be considered a special case of CEtot, where the
acquisition of future customers is a project with a net present
value of zero, as seen in Massimilliano Bonacchi, Kalin Kolev,
Massimiliano Bonacchi, Ph.D., is an associate professor at and Baruch Lev, Customer Franchise—A Hidden, Yet Crucial
the University of Naples “Parthenope” and an adjunct pro- Asset, Canadian Academic Accounting Association, Ontario,
Canada, 2012. See also John E. Hogan, Donald R. Lehmann,
fessor of accounting at NYU Stern School of Business. He
Maria Merino, Rajendra K. Srivastava, Jacquelyn S. Thomas,
can be reached at massimiliano.bonacchi@uniparthenope.it. and Peter C. Verhoef, “Linking Customer Assets to Financial
Performance,” Journal of Service Research, 2002, pp. 26-38; and

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 38 FALL 2012, VOL. 14, NO. 1

 V. Kumar and Denish Shah, “Expanding the Role of Market- 15 Fader and Hardie, 2007, and Fader and Hardie, 2010.
ing: From Customer Equity to Market Capitalization,” Journal 16 Robert C. Blattberg, Byung-Do Kim, and Scott A. Neslin,
of Marketing, 2009, pp. 119, for more about CEtot. Database Marketing: Analyzing and Managing Customers, Springer
8 Massimiliano Bonacchi, Mascia Ferrari, and Massimiliano Pel- Science + Business Media, New York, N.Y., 2008.
legrini, “The Lifetime Value Scorecard: From E-Metrics to 17 For example, see Gupta and Lehmann, 2005.
Internet Customer Value,” Performance Measurement and Man- 18 Using the same rationale, Company.net also computes revenue
agement Control: Measuring and Rewarding Performance, 2008, per thousand customers (RpK), which focuses on projected
pp. 193-226. revenues instead of margin. In the example, we concentrate on
9 In theory, the issue of the time-value of money must be also the MpK evolution, but the same report can be adopted for
considered, usually discounted in the CLV formula, in order to RpK because gross margin is a percentage of revenue (in this
get an estimate of future profits. Discounting may be very example, 65%), and the difference between RpK and MpK is
appropriate in some businesses, particularly for firms operating merely a question of scale. Our choice to present MpK is
in markets with a very long cycle, high ticket retail, and B2B. because we are able to calculate CE with this metric.
But it is believed that the discounted practice may confuse 19 See Robin Gleaves, Jamie Burton, Jan Kitshoff, Ken Bates,
the analysis in a B2C where the environment is very dynamic, and Mark Whittington, “Accounting Is from Mars, Marketing
and it is better to adopt a simplified approach. Is from Venus: Establishing Common Ground for the Concept
10 When a user searches for a specific keyword, the order of the of Customer Profitability,” Journal of Marketing Management,
results the user obtains is determined by current bids in the September 2008, pp. 825-845, for a distinction between cash
auction. Payment is made by advertisers each time a user flow and profit.
searches for a term and then clicks on their link. 20 When it comes to making informed prospecting decisions,
11 Laffey, 2007. there are at least two ways of considering acquisition spending:
12 Werner J. Reinartz and V. Kumar, “The Impact of Customer Do not include acquisition spending and compare the lifetime
Relationship Characteristics on Profitable Lifetime Duration,” value (LTV) to CoA or include acquisition spending in the
Journal of Marketing, 2003, pp. 77-99. specification of customer value—correctly labeled as CLV—
13 Peter S. Fader and Bruce G.S. Hardie, “How to Project and compare the value of CLV to zero. See Phillip E. Pfeifer,
Customer Retention,” Journal of Interactive Marketing, 2007, Mark E. Haskins, and Robert M. Conroy, “Customer Lifetime
pp. 76-90; Peter S. Fader and Bruce G.S. Hardie, “Customer- Value, Customer Profitability, and the Treatment of Acquisi-
base Valuation in a Contractual Setting: The Perils of Ignoring tion Spending,” Journal of Managerial Issues, March 22, 2005,
Heterogeneity,” Marketing Science, January-February 2010, pp. 11-25.
pp. 85-93; and Sunil Gupta and Donald R. Lehmann, Manag- 21 Kumar and Rajan, 2009, and V. Kumar, Rajkumar Venkatesan,
ing Customers as Investments: The Strategic Value of Customers in the Tim Bohling, and Denise Beckmann, “The Power of CLV:
Long Run, Wharton School Publishing, Upper Saddle River, Managing Customer Lifetime Value at IBM,” Marketing Sci-
N.J., 2005. ence, May 7, 2008, pp. 585-599.
14 Sunil Gupta, Donald R. Lehman, and Jennifer A. Stuart,
“Valuing Customers,” Journal of Marketing Research, 2004,
pp. 7-18.

M A N A G E M E N T A C C O U N T I N G Q U A R T E R LY 39 FALL 2012, VOL. 14, NO. 1

Copyright of Management Accounting Quarterly is the property of Institute of Management Accountants and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.