8/10/2019 Making Customer Relationship Management Work:The Measurement and Profitable Management of Customer Re

1/10

Lynette Ryals

Making Customer RelationshipManagement Work:TheMeasurement and ProfitableManagement of Customer

RelationshipsCustomer relationship management (CRM) is perceived to be failing, and there is an urgent need for some practi-cal ways to address this issue. The research presented in this article demonstrates that the implementation of

CRM activities delivers greater profits. Using calculations of the lifetime value of customers in two longitudinal casestudies, the research finds that customer management strategies change as more is discovered about the value ofthe customer. These changes lead to better firm performance. The contribution of this article is to show that CRMworks and that a relatively straightforward analysis of the value of the customer can make a real difference.

Lynette Ryals is Senior Lecturer in Marketing, Cranfield School of Man-agement, Cranfield University (e-mail: lynette.ryals@cranfield.ac.uk).Thisarticle is based on the authors dissertation work conducted at CranfieldSchool of Management.The author gratefully acknowledges the data sup-port of a major U.K. bank and of a European financial services company.The author appreciates the comments of the three anonymous JMreviewers and the guidance of the consulting editors on the work. Theauthor also acknowledges the research support of her dissertation super-visor, Simon Knox, and she thanks John Towriss and Sam Dias for theirhelp with the analysis.

Customer relationship management (CRM) is part ofmarketings new dominant logic (Day 2004), but itis more likely to fail than to deliver any business

results (Zablah, Bellenger, and Johnston 2004). Still worse,failed implementation may actually damage customer rela-tionships (Rigby, Reichheld, and Schefter 2002). Thisresearch demonstrates that the implementation of CRMactivities generates better firm performance when managers

focus on maximizing the value of the customer (Gupta andLehmann 2003; Gupta, Lehmann, and Stuart 2004; Reich-held 1996; Verhoef and Langerak 2002). Previousresearchers have suggested that a better understanding ofthe value of the customer should lead to changes in the waythese customers are managed (Mulhern 1999; Niraj, Gupta,and Narasimhan 2001; Reinartz, Thomas, and Kumar2005). Through two case studies that develop detailed dataabout lifetime revenues and customer-specific costs, thesepredictions are borne out. Moreover, information about bothrevenues and costs enables the exploration of whether largercustomers, on average, provide the firms studied with morevalue than smaller customers (Dowling 2002; Kalwani andNarayandas 1995).

Research MethodologyThis section describes the operationalization of the lifetimevalue of the customer. The method I chose was two collab-orative longitudinal case studies, one to explore individualcustomers and the other to explore a customer base. Bothcase studies were with companies that had not previouslycalculated the value of their customers; thus, in both cases,

managers were receiving new information.

Empirical Context

Although relatively few studies on the value of the customerare available (Reinartz and Kumar 2000), published studiesare drawn disproportionately from the financial servicesindustry (e.g., Carroll and Tadikonda 1997; Hartfeil 1996;Storbacka 1997), reflecting good availability of data aboutcustomer revenues and costs to serve. The financial servicesindustry is a context that meets the requirements of the cur-rent study well (Blattberg, Getz, and Thomas 2001; Verhoef2003), and restriction to a specific industry allows forgreater comparability of the method and of the results.

For the purposes of this research, I selected two partici-pating firms from the financial services industry on the

basis of activity, size, availability of data, willingness tocommit the necessary resources to the research, and lack ofprevious information about the value of their customers.The first case study was business-to-business at a Europeaninsurance company (hereinafter the insurer), and itinvolved the insurers key account management (KAM)team, which managed insurance for the companys largestcorporate clients. The second case study was business-to-consumer, and it examined unsecured lending to people bythe personal loans division of a major U.K. bank (here-inafter the bank). At the insurer, data were collected onten key accounts (the firms largest customers). At the bank,

252Journal of Marketing

Vol. 69 (October 2005), 252261

2005, American Marketing Association

ISSN: 0022-2429 (print), 1547-7185 (electronic)

Ryals, L. ( octubre , 2005 ) . Making customer relationship management work : the measurement and

profitable management of customer relationships . Journal of Marketing (69) pp. 252-261. (AR48785)

8/10/2019 Making Customer Relationship Management Work:The Measurement and Profitable Management of Customer Re

2/10

Making CRM Work / 253

the sample consisted of all applications for a personal loanduring a given period.

The research approach was collaborative and longitudi-nal, involving a series of individual interviews and groupworkshops with the two project teams (the KAM team atthe insurer and the marketing department supported by theinformation technology department at the bank). In bothprojects, only customer-specific product and service costswere considered (Niraj, Gupta, and Narasimhan 2001); gen-eral overhead costs were disregarded in the interests of

improving the achievability, costs, and speed of theresearch. Following Berger and Nasrs (1998) and Mul-herns (1999) approach, I confined the definition of lifetimevalue to the direct and potential profits from the relation-ship; other potential benefits, such as attracting other busi-ness (Reinartz and Kumar 2002; Wilson 1996) or learningpotential (Stahl, Matzler, and Hinterhuber 2003; Wilson1996), were excluded from the calculation of lifetime value.I also applied the conventional assumption that after cus-tomers are lost, they do not return (Rust, Lemon, andZeithaml 2004), which is a reasonable assumption given therelatively lengthy contractual relationships underconsideration.

The Lifetime Value of the Customer: Case Study 1

In line with the work of Mulhern (1999) and Jain and Singh(2002), I defined the value of the customer for any customeror customer segment, x, as the future lifetime stream of rev-enues from that customer or segment (CR) minus the futurelifetime costs (CC) for a predicted relationship lifetimefrom time t = 1 to n such that

where i is a discount rate and the function 1/(1 + i)n yields

the net present value calculation.The account managers made forecasts of lifetime dura-tion based on previous experience with that customer andwith other, similar customers. The mean forecast lifetimewas 4 years. This forecast was triangulated by calculatingthe historic mean lifetime of a key account for the insurer,which was 4.75 years (standard deviation = 4.06).

Revenues vary from customer to customer depending onthe number and mix of products purchased (Reinartz andKumar 2002) and, over time, on the share of wallet (Ver-hoef 2003) and the way that share of wallet might migratebetween competing providers (Coyles and Gokey 2002;Knox and Walker 2001). Thus, for time t = 1 to n and forproducts q = 1 to m, the lifetime revenues of customer x areCRx, where

and V and P are the volume and prices of products (bothongoing and new).

In a series of workshops followed by individual calcula-tion using a pro forma supplied by the researcher, account

( )2

11

CRx V Pqt qtq

m

t

n

=

==

( )( )

( ),1

11

CLVxCR CC

in n

nt

n

=

+=

managers forecasted customer revenues based on insuranceproducts purchased and probable price. Multiple iterationsand new data in the form of customers contract renewal ledto revenue forecasts the managers believed were realistic.

To arrive at the value of the customer, it is necessary todeduct costs (product costs and costs to serve) from the cus-tomer revenue stream (Berger and Nasr 1998). Product costsare determined by the volume of sales and the product mix.

The insurers claims costs were treated as analogous toproduct costs and were calculated on the basis of historic

percentage claims for that type of insurance across theentire customer portfolio multiplied by the value of eachproduct purchased by that customer. Thus, the customerclaims cost for customer x, which had bought m insuranceproducts to the value of V over n time periods, was calcu-lated as follows:

where C is the predicted percentage of claims plus a smallprobability of some catastrophic future event if the accountmanagers believed that the customers industry or business

was sufficiently risky to warrant this.Other costs to serve are determined by customer behav-

ior (Ryals 2002a, b). Even the most sophisticatedapproaches to customer lifetime value calculation havetreated these costs as allocatable proportional to sales vol-ume or value (e.g., Berger and Nasr 1998; Mulhern 1999).However, allocating costs in this way makes the assumptionthat all customers buying the same combination of goodsand services cost the same to serve (Ryals 2002b), whereassome customers may be far more costly to serve than others(Bolen and Davis 1997; Cooper and Kaplan 1991; Niraj,Gupta, and Narasimhan 2001). Activity-based costing canbe used to determine the appropriate allocation to cus-tomers of costs to serve (Cooper and Kaplan 1991).Customer-specific costs to serve (CSC) were calculated as afunction of KAM time, administration and processing costs,and costs of risk evaluation. Thus, for time t = 1 to n, thecustomer-specific costs to serve for customer x were

(4) CSCx = (KAM timet costt) + (admin costst

number of claims processedt)

+ (cost of risk evaluationt

number of risk evaluationst).

To estimate the time spent on each customer, the key

account managers began keeping diary-based activity logsthat showed time out of the office with each customer, timein the office on each customers business, and other uses oftheir time. From these activity logs and the use of the meanannual salaries for the three job grades in the team, theapproximate activity-based costs of the client managingeach key account could be projected. Customer-specificadministrative and risk evaluation cost data were collectedfrom the insurers CRM systems.

t

n

=

1

( ) ,3

11

CCCx V Cptp

m

pt

t

n

=

==

8/10/2019 Making Customer Relationship Management Work:The Measurement and Profitable Management of Customer Re

3/10

254 / Journal of Marketing, October 2005

The Lifetime Value of the Customer: Case Study 2

When the total value of a large customer base is to be con-sidered, calculation of many individual customer lifetimevalues may become unwieldy. At the bank, a sample frameof all applications for an unsecured personal loan during aset three-month period (N = 123,442) yielded a sample size(n) of 62,851 customers, after unsuccessful and uncom-pleted applications were excluded. Of the sample, 1513 cus-tomers could not be allocated to a marketing segment, so

the usable sample was 61,338 (49.7% of the sample frameand 97.6% of the completed loans in the period).Blattberg, Getz, and Thomas (2001, p. 23) propose a

detailed definition of the total lifetime value of the customerbase, which includes acquisition cost and probability, reten-tion cost and probability, add-on sales, cost of goods sold,and the number of segments. In practice, this definitionneeded to be modified in operationalization because cus-tomers take out only one personal loan at a time, so add-onsales were not relevant. Future potential sales were disre-garded (Calciu and Salerno 2002; Rust, Lemon, and Zeit-haml 2004). This gave a modified definition of customerequity for each customer segment, CEseg (Blattberg, Getz,and Thomas [2004, p. 17] refer to these as customer

cells):

(5) CE(seg) = (Rseg) [(ACseg) + (RCseg)],

where Rseg is the lifetime revenue of the customer segment,ACseg is the acquisition costs by channel of that segment,and RCseg are the retention costs within that segment foreach customers lifetime.

The project began with an analysis of factors that drivecustomer revenues and costs. Lifetime revenue (loan inter-est) was a function of the size of the loan, additional rev-enues from arrears, and the interest rate. Because the banktypically considered customers in terms of size of the loan,two possible definitions of customer size were considered:

the loan size and the lifetime revenue.In this second project, a combination of simple activity-

based and standard costs was again used. Acquisition costswere relatively straightforward to calculate. Certain cus-

tomers were acquired through certain channels, so the costsof that channel were allocated to the customers who camethrough it.

Retention costs were more complex. Certain administra-

tive costs were standard and applied to all customers. How-ever, other costs to serve in the customer relationships werelargely driven by the customers conduct of the loanin

other words, whether the customer had been in arrears ornot. The lowest-cost customers were the customers who hadnever been in arrears. The highest-cost customers were

those who had been in arrears for more than five months.Considerable complexity was involved in calculating thecosts of arrears, which involved several different activitiesand teams.

Results

Case Study 1: The Value of the Customer at theInsurer

Table 1 shows the ten key customers for which full cus-tomer lifetime value data were developed, ranked by life-

time value. Although only ten customers, these customers

had lifetime revenues of $67.6 million and accounted formore than 10% of the annual revenue of this division of the

insurer. The total value of this key account portfolio was$24.8 billion. Although this is a small sample, two indus-tries (chemicals customers H and C and business servicescustomers K and F) appear twice. These customers have

different lifetime values, suggesting that customer industryis not the chief determinant of the value of the customer.

The relationship among the lifetime revenues, costs, and

value was explored using Pearsons r (Rupinski and Dunlap1996). All results are significantly different from 0 (p