FRONTIERS OF E-BUSINESS RESEARCH 2004

CRM in the Insurance Industry: An Attempt to Use Survival Analysis in Retention and Cross Selling
Maria T. Salazar1; Tina Harrison2; Jake Ansell3 1 doctoral candidate, University of Edinburgh 2 senior lecturer, University of Edinburgh 3 senior lecturer and associate dean, University of Edinburgh Abstract Relationship Marketing emphasises the benefits associated with customer retention over new clients acquisition. However, financial organisations are not still completely enhanced with this idea. In order to improve the retention ratios of organisations, cross-selling has been identified as a very effective strategy with positive benefits for companies. This identification of new business opportunities depends on the detection of product acquisition and timing evaluation. However, there are other variables as those related to the life-cycle whose analysis is crucial to obtain a better picture. Using such survival analysis, this paper illustrates how to promote retention in the insurance industry. Basically, this research helps to understand which products and when those products are more likely to be acquired for a data sample of 16348 customers. Keywords cross-selling, up-selling, retention, survival analysis, relationship marketing, financial services

Introduction
In recent times, managers have started understanding the relevance of retaining active customers with companies as a key component guarantying their survival in the market. Moreover, the benefits associated with companies counting on loyal customers have been highlighted. This loyalty emerges from the relationship customer- organisation, the longer and more intense the relationship is, the higher the benefits. In promoting retention and loyalty, companies have been aware of the relevance of increasing the customers consumption via cross-selling and up-selling. However, the identification of the consumption pattern is not sufficient to promote subsequent purchases. Due to this, predicting the specific time points when purchase decisions are more likely to be made completes the picture and helps companies in designing effective satisfaction and retention strategies. This paper presents a quantitative analysis to discover those undeveloped opportunities, stored in a companys data base, identifying both the products and the time schedule.

68

FRONTIERS OF E-BUSINESS RESEARCH 2004

Theoretical context
Since the early 1990s, there has been a significant amount of research on a new business philosophy called CRM. This seeks to create, develop and enhance relationships with targeted customers in order to maximise the value received by the customer, the profits of the company and the contribution demanded by social forces. CRM is not only a technology application for marketing purposes, it is a cross-functional, customer driven, technologyintegrated business process management strategy that maximises relationships and encompasses the entire organisation (Goldenberd, 2003). Following Greenbergs view (2003), CRM is a philosophy translated into a business strategy, supported by a system and a technology. The origins of CRM are found in Relationship Marketing Theory which is aimed at improving long term profitability by shifting from transaction-based marketing, with its emphasis on winning new customers, to consumer retention through effective management of costumer relationships (Christopher et al., 1991). The benefits associated with retention have a positive impact on revenues, profitability, defection ratio and new business opportunities (Felvey, 1982; Sonnenberg, 1988; Reichheld & Teal, 1996; Best, 1998; Reinartz & Kumar, 2000). Reichheld and Teal (1996) suggest that customers who have been around long enough to get familiar to the companys procedures, will create more valuable business relationships, will acquire more products and will be less price sensitive on individual offers. Amongst all of those benefits, the high level of product acquisition (cross-selling) has been seen as one of the main tools to reinforce the relationship with customers (Kamakura et al., 1991). By increasing the number of products consumed, cross-selling simply reinforces the customer loyalty and raises the cost of switching from the company. As customers remain with the company for a longer period of time, the firm can learn and understand more from the customers buying behaviour and their preference in terms of products and timing. However, loyalty is a very complex phenomenon consisting not only of a repurchasing sequence; it also implies an attitudinal or emotional dimension enhancing the customer relationship to the organisation (Harrison, 2000). The difficulties in addressing loyalty and the little attention that cross-selling has received by academia offers a gap in which to locate this research. Although cross-selling and retention are relevant issues in multi-service/product providers in general, and in financial organisations in particular.

Research context
Until the late 1980s the UK financial service industry was constricted by tight regulations in terms of product scope and geographical limitations. In this scenario the adoption of marketing concepts was not a priority; moreover the industry was quite reluctant to do it. However, with the deregulation of the 1980s, new competitors entered the market, possibilities to diversify were discovered and new opportunities uncovered. In this way, banks, building societies and insurance companies had the opportunity to diversify their offerings by making use of that increased freedom to provide a wider range of products (Mintel, 2000). In addition to this, the advances experienced in management and information technologies promoted the emergence of multi-channel strategies and the definition of more complex offerings. As a result, companies started fighting for customers and realised the relevance of remaining customers with them. However, customers are now more educated, 69

FRONTIERS OF E-BUSINESS RESEARCH 2004

with better understanding of the financial environment and higher quality, service and performance requirements. In this environment financial services have to remain competitive. In order to face the demands from the market-place and the increasing pressure from the competitors, a shift is required from products to customers, from short-term gains to long-term relationships. This means focusing on customer retention, new product development, loyalty and the use of innovative technology-driven strategies.

Aims and objectives of the research

Research methods
To address the research questions, this research analyses customer behaviour information recorded on the data base of a large insurance company in the UK, whose core business is pension plans and investment policies. The data set is formed of a sample of 16348 customers and 48 variables. Given the objectives of this research, the variables used relate to age, gender, ACORN classification1, value, number of policies, products acquired and the purchase dates. The method of analysis is based on statistical approaches including survival analysis, the development of purchase trees and dependency tests. The first step is to asses the impact of retention in relation to the consumption levels and the customers value generated in the company. Following this, the main issue to address is the identification of the consumption patterns to expose the cross-selling and up-selling opportunities. Moreover, this analysis will be completed with the evaluation of customers characteristics and experience with the company. In order to address a better understanding, a purchase sequence is estimated. Using survival analysis, the occurrence probability and timing of specific events are studied. The objective is to predict when the purchase of the next policy will occur. The predictive model used is Coxs model (1) (Cox ,1972) and it is defined as follows: hi(t) = 0(t) exp1xi1++kxik (1)
ACORN (A Classification system Of Residential Neighbourhoods) is a geo-demographic system based on some 40 variables generated from the population census. (A) means Financially Sophisticated, (B) Financially Involved, (C) Financially Active and (D) Financially Inactive.
1

70

FRONTIERS OF E-BUSINESS RESEARCH 2004

The equation states that the hazard for individual i at time t depends on two factors: a baseline hazard function 0(t) and a linear on a set of k variables, called covariates, which may change in value over the period (Allison, 1995).

Results
Descriptive analysis The majority of customers are male (63%). In terms of age, the wider age range varies from 31 to 60 years old which covers 80 % of the companys actual customers. However, the age at which consumers embark on a contractual relationship with the company starts a bit earlier, as 68% of the customers bought their first policy between the ages of 21 to 40 years old. Finally, using the ACORN style of life classification, the groups A and B represent the largest amount of customers with 64% of the total pool. The analysis of the product consumption shows that 48% of the actual consumers are reinforcing their allegiance with the company by buying another product. Thus, this situation provides a wide pool of consumers that could be approached to encourage further purchases. The products with a higher level of acceptance for those who just posses one product are Protection (15%) and SP Investment (54%) which represent almost 70% of all the products acquired. Those buying two policies from the company prefer SP Investment (57%), fewer Protection (12%) and they also show some interest in Regular Savings (7%). This tendency remains for the fourth purchase, where the most important products are SP investment (54%), Protection (11%) and Regular Savings (9%). Finally, those percentages reduce steadily to allow the entrance of new types of products into the baskets of those purchasing more than four policies. In fact, SP Investment and Protection falls to 50% and 10% respectively, allowing products such as PEP, ISA and FSAVC to complete the picture. Assessing retention benefits At the beginning of this article, the relevance of retention for companies was questioned. Using the data base, it has been analysed how enhancing customers in an active and long relationship has a positive impact from the companys perspective (see Figure 1). In the particular case of this company, retention seems to be an issue of particular relevance as just 11% of the initial buyers reach the final stage of five purchases. In addition to that, it can be concluded from the analysis that the longer the customer stays with the company, the more products are acquired and higher the value received by customers. This positive relationship slows down after the twenty-first year of enhancement (value 7 of life@co) with the company and after the fourth product as it colludes with the stage when clients start consuming what they saved.

71

FRONTIERS OF E-BUSINESS RESEARCH 2004

Value by time at the company

No of Products by time at the company

Value by No. of Products

4,00

4,00 4,00 3,00

value

3,75

No. prods

value
3,50

3,50

2,00

3,25

3,00 1,00 2,00 4,00 6,00 8,00 2,00 4,00 6,00 8,00 2,00 3,00 4,00 5,00

life@co

life@co

numprod

Figure 1. Value of Retention An independency analysis has been done to check whether there is any relationship in the value associated to customers with the products they consume and the period of time they maintain an active relationship with the company. At 95% of significance level it has not been possible to accept the independency hypothesis. Thus it can be concluded that the longer the customers remain with the company, the heavier is the consumption pattern and the higher the value they generate for the company. Spotting cross-selling and up-selling opportunities The relevance of identifying new business opportunities has already been explained. On one hand, it increases the revenue of the company (more product acquisition requires less cost), and on the other hand, it reinforces the relationship with customers as the contact with the company is more complete. Using the image of a tree (see Figure 2), the consumption pattern of the most relevant products acquired from the company has been detailed. Following the arrows it is possible to identifiy the policies purchased at any purchase time along with its relative importance over the basket. For example, in the first purchase 15% purchased Protection. Of this group, the cross-selling opportunities (in rectangles) focus on Regular Savings (36.5%) and ISA (12%). Moreover, the presence of Protection during the subsequent purchases (in ellipses) provides a good chance to design up-selling strategies (up-grading or improving the conditions of a product already purchased). As a general pattern it has been discovered how the portfolio components change as long as customers face new needs related to their life-cycle. The main rule establishes that the first round of policies acquisition relies on a house purchase (i.e. Protection). It continues with saving products (i.e. ISA Regular Savings) finishing with the acquisition investment products (i.e. SP Investment, PEP).

72

FRONTIERS OF E-BUSINESS RESEARCH 2004

PURCHASE 1
Protection
(14.9%)

PURCHASE 2
Reg. Savings
(36.5)

Protection
(41.6%)

ISA

(12 %)

PURCHASE 3
SP Investment
(24.1%)

Protection
(63.6%)

ISA
(4.5%)

Regular Savings
(4.5%)

PURCHASE 4
SP Investment
(16.8%)

Protection
(69.3%)

Regular Savings
(4%)

PURCHASE 5
SP Investment
(4.3%)

Protection
(82.9%)

Regular Savings
(4.9 %)

Figure 2. Purchase Tree example for Protection Variables affecting cross-selling Once it has been stated that actually there are cross-selling and up-selling opportunities, the next logical step is establishing how socio-demographic variables affect the likelihood of customers repurchasing (see Table). It can be deduced that the age (agecode) has a positive influence in the model. On the other hand, the age of the customer when he/she started the relationship (agesdtco) with the company influences in an indirect way (negative coefficient 0.544). From those statements it can be inferred that the longer the customer remains with the company2 and the more mature they are, the higher is the probability of repurchasing. In terms of value, the coefficient is positive (0.609) which means that as clients present higher value3, they are more likely to repurchase this product. This is consistent with the significance that ACORN type A has in the model. In fact from the four categories, the highest influence is in A group4. It means that repurchasing could be classified as a usual behaviour from those
2

The younger customers are when starting the contractual relationship measured by the negative value of agesdtco. 3 The total amount of premiums they are actually paying. 4 Exp (B)= 1.309.

73

FRONTIERS OF E-BUSINESS RESEARCH 2004

with high expenditure capacity which are predisposed to be financially involved. Finally, the presence in the model of the variable Marital Status cannot be accepted as its significance is (0.06)5.

B [Link] [Link](D) [Link](M) [Link](S) [Link](U) [Link] [Link](A) [Link](B) [Link](C) [Link](D) VALUE Age at starting with the company (AGESDTCO) AGE ,332 ,651 ,659 ,647 ,269 ,178 ,164 ,046 ,609 -,544

SE ,535 ,568 ,502 ,505 ,054 ,055 ,063 ,084 ,014 ,018

Wald df Sig. Exp(B) 10,588 5 ,060 ,385 1 ,535 1,394 1,317 1 ,251 1,918 1,724 1 ,189 1,933 1,642 1 ,200 1,910 30,523 4 ,000 24,580 1 ,000 1,309 10,289 1 ,001 1,195 6,765 1 ,009 1,178 ,298 1 ,585 1,047 1914,334 1 ,000 1,838 866,393 1 ,000 ,581 106,136 1 ,000 1,192

,175 ,017

Table 1. Coefficients and significance of the Cox Regression Model The final model (2) would be:
Repurchasei(t) = exp0.332 Divorced + 0.651 Married + 0.659 Single + 0.647 Other + 0.269 ACORN A+0.178 ACORN B+0.164 ACORN C+0.046 ACORN D + 0.609 Value - 0.544 Agesdtco + 0.175 AGE (2)

Time analysis Finally, in order to extend the analysis which covers the purchase time sequence, the Time Charts (see Figure 3) describe how repurchasing probability reduces as time goes by. This implies that the company should be enough active in promoting soon the subsequent purchases once the previous ones are arranged. Due to the long-term nature of the products that have been considered, the lag of time between purchases is quite long.

at 95% of significance level. The reason can be because of the high proportion of unclassified data.

74

FRONTIERS OF E-BUSINESS RESEARCH 2004

Survival for purchase (1,2)

Survival for Purchase (2,3)

,8

,8

,6

,6

,4

Purchase Probability
0 5 10 15 20 25 30

Purchase Probability

,4

,2

,2

0,0

0,0 0 5 10 15 20 25 30

Time in Years

Time in Years

Survival for Purchase (3,4)

Survival for purchase (4,5)

,8

,8

,6

,6

Purchase Probability

Purchase Probability

,4

,4

,2

,2

0,0 0 5 10 15 20

0,0 0 5 10 15 20 25 30

Time in Time Figure 3. Years Charts

Time in Years

To make the understanding of the repurchase sequence easier, the following chart (Figure 4) summarises the time schedule that the company should bear in mind to optimise its selling activities.
P1 0
Time in years

P2 2
(2.5)

P3 4
(4.25)

P4
(5. 75)

P5
(7.75 )

Pi = Purchase with i = (1,2,3,4,5)

Figure 4. Purchase Time Schedule at 95% probability

Managerial recommendations
This research has tried to expose the benefits that companies can obtain from carefully managing their retention programs and figures. The suggested method to approach the maintenance of customers has been the identification of cross-selling and up-selling opportunities which could strengthen the relationship customer company). Moreover, this research has addressed which products should be offered next, when the company should approach the customers and the variables affecting the probability of more purchases. 75

FRONTIERS OF E-BUSINESS RESEARCH 2004

Conclusions
This study has addressed from a real example the relevance of retaining customers and its implication for retention and loyalty strategies. This has been analysed from the perspective of discovering new business opportunities. Looking inside the organisation and at its customers data base, the definition of possible selling strategies has been explained in terms of product consumption, customer characteristics and time frame work. The translation of this information into managerial actions could lead to higher ratios of profitability and efficiency of organisations as long with more precise marketing approaches. Finally it supports the underlying idea subscribed by Relationship Marketing Theory which states the relevance of enhancing customers in profitable relationships matching the expectations of both company and customers groups.

Limitations
This study assumes that the observed behaviour is the output of the natural life-cycle flow. However, specifically undertaken marketing campaigns or other external forces have not been considered as possible purchase drivers (Prinzie and Van den Poel, 2004). Moreover, customer behaviour with other competitors has not been considered. Because of using a single data source, the practical recommendations are not robust enough. It means that the opportunities identified could not be so if clients are already consuming products from other companies. The quantitative analysis used seems suitable for the insurance industry where almost all the activities have been electronically recorded for years. However, other industries could not be so up to date to replicate the study or to spread the results. Finally, due to product characteristics and the necessities they try to fulfil (i.e. safety, future income) companys brand, trajectory and reputation might affect positively or negatively the repurchase sequence (Aaker and Keeler, 1991). All these factors should be taken into account when carrying out future research.

76

FRONTIERS OF E-BUSINESS RESEARCH 2004

References
Aaker, D and Keller, K., Consumer evaluations of brand extensions Journal of Marketing (1991) Best, D., The future of information management international Journal of Information Management (1998), vol.8 Christopher, M., Payne, A., and Ballantyne, D., Relationship Management- Butterworth-Heinemann (1991), Oxford Cox, D., Regression Models and life tables Journal of Royal Statistics Society Series B (1972), vol. 74 Felvey, J., Cross-selling by computer Bank Marketing, (1982) Goldenberd, B., What is CRM? What is an e-customer? Why you need them now? in Proceedings of DCI Customer Relationship Management Conference, Boston. MA, 27-29 June 2003 Greenber, P., Creating realistic customers expectations CRM Magazine (2003), vol. 7 Harrison, T., Financial services marketing- Pearson Education Limited (2000), Essex Kamakura, W., Ramaswami, S. and Srivastava, R., Applying latent trait analysis in the evaluation of prospects for cross-selling of financial services International Journal of Research in Marketing (1991), vol. 8 Kamakura, W., Wedel, M., de Rosa, F. And Mazzon, J., Cross-selling through database marketing: a mixed data factor analysed for data augmentation and prediction- International Journal of Research in Marketing (2003), vol.20 Mintel International Group, Customer Retention: Solving the Puzzle [Link] (2002) Prinzie, A. and Van der Peol, D., Investigating purchasing-sequence patterns for financial services using Markov, MTD and MTDg models European Journal of Operational Research (2004) Reichheld, F. and Teal, T., The loyalty effect: the hidden force behind growth, profits and lasting value Harvard Business School Press (1996), Cambridge, M.A Reinartz, W. and Kumar, V., On the profitability of long-life customers in a noncontractual setting: an empirical investigation and implications for marketing- Journal of Marketing (2000) vol.64 Sonnenberg, F., The power of cross-selling The Journal of Business Strategy (1988) January/February

77