Implications of minimum contract durations on customer retention

Abstract

Customer retention is a major driver of customer lifetime value and is thus a key performance metric in marketing management. Consequently, companies try to retain customers by offering contracts with minimum contract durations (MCD). Using behavioral, psychometric, and advertising data for a large sample of DSL customers, the authors study the impact of minimum contract durations on actual customer churn behavior. The analyses demonstrate that subscriptions with minimum contract durations do indeed help companies to successfully retain customers. The effect is impaired though, as companies typically (must) provide incentives to convince customers to commit to those contracts. We find that incentives attract customers that either cannot or should not be retained and hence require companies to carefully apply both MCD and incentives.

1 Introduction

Customer retention is a major driver of customer lifetime value, especially in the case of continuously provided services with constant usage, such as flat-rate cellular phone and Internet services, cable television contracts, or health club memberships (Bhattacharya 1998). Companies in these industries often use contracts with minimum contract duration (MCD) to increase customer retention by attracting customers with long-term commitment.

It is a common industry practice to reward the customer commitment to MCD contracts with incentives (e.g., rebates or hardware; Kim et al. 2004; Tallberg et al. 2007). These incentives are deemed necessary to compensate customers for the uncertainty associated with their commitment (e.g., decreasing future rates or service quality; Della Vigna and Malmendier 2006; Miravete 2003). However, incentives may also entail undesirable effects for the company (Frey and Oberholzer-Gee 1997), as they attract customers that are primarily interested in the incentives. Such customers may not only terminate the service directly after the end of the minimum contract durations, but they may also show defaulting behavior by taking advantage of the incentives without paying for the service. Consequently, the use of MCDs would not only result in less loyal but also less profitable customers.

Despite its practical relevance, there is little empirical research on the effect of MCDs on customer retention. In this paper, we study the implications of using an MCD in combination with incentives on customer defections and default. We use a unique set of transactional, survey, and advertising data from a European Internet Service Provider (ISP) that offers two contract options: one with a minimum contract duration (i.e., 24 months) and one with no minimum contract duration (i.e., canceled “anytime” on a month-by-month basis) and thus allows for a direct comparison. Our analyses demonstrate that although MCDs are successfully lowering overall customer churn, they also promote undesirable effects that require a careful application of both MCD and incentives.

2 Theoretical background

Companies use minimum contract durations for a variety of products such as telecommunication or Internet services, cable television, or health club memberships; and they often use MCDs in combination with an alternative non-MCD contract. By offering MCDs, companies assume that those customers (self) selecting MCD contracts have a higher likelihood of remaining loyal after the end of the MCD.

2.1 Motives for choice of MCD contracts with incentives

The expected choice of MCD by more loyal customers can be explained by the social exchange theory (Homans 1958). For continuously provided services with constant usage, service providers and customers engage in relational exchanges where both parties share an anticipated future that conveys the desire for a long-term relationship (Arndt 1979; Dwyer et al. 1987). Hence, long-term oriented customers trust that the service provider will be able to deliver the promised service quality and therefore are willing to commit to a long-term contract (Anderson and Weitz 1984; Ganesan 1994). This long-term orientation is higher for those customers that face high costs for searching, setting up, maintaining, or disposing of a service and thus want to avoid these costs (Lam et al. 2004; Lambrecht and Tucker 2012).

It is also plausible though that price-sensitive consumer without an intention for long-term commitment choose MCD contracts primarily to enjoy the incentives provided. This is quite comprehensible as incentives increase the customer’s transaction utility and are coded by customers as gains that alter the reference price for the services (Tversky and Kahneman 1981; Thaler 1985). Furthermore, consumers tend to have time-inconsistent preferences (O’Donoghue and Rabin 1999). Due to such present-biased preferences, individuals give stronger relative weight to gratifications that are immediate or closer in time (O’Donoghue and Rabin 2000).

2.2 Implications of contract choice on customer retention

In the case the choice of MCD contracts is motivated by customers’ long-term orientation, companies may expect MCDs to increase customer retention and tenure. Furthermore, higher customer retention may also result from the unwillingness to switch to alternative choices because of consumer inertia, which lowers their motivation to change (Sheth and Parvatiyar 1995), or because their cumulative satisfaction weighs more heavily with longer relationships (Bolton 1998).

Since monetary incentives may however promote negative effects when customers behave opportunistically (Frey and Oberholzer-Gee 1997), the expectation of higher retention may not hold for price-sensitive customers who choose MCDs predominately because of the incentives. They may well try to break the contract as soon as possible, for example, in order to receive new incentives from a new contract. We denote such defections of customers serving only the minimum contract duration as early defections. In order to study the effects of MCD on different aspects of churn, we analyze the following two research questions:

• RQ1: What is the effect of MCD in combination with incentives (25 % discount) on overall customer churn (including early defections)?

• RQ2: What are influences of early defections?

Although such early defections may not be in the companies’ interest, they mark a legitimate form of churn as the customers fulfill their contractual obligations and stay until the end of the MCD. Another undesirable and highly illegitimate behavior resulting from incentives could be default. Here, customers show no commitment at all and attempt to breach the contract after receiving the incentive. In the case of DSL Internet service, an example of default would be customers consuming the service but refusing to pay their fees. In such cases, the provider proactively terminates the contractual relationships with the defaulting customers. As the incentives may attract such defaulting customers to select MCD contracts with incentives, we analyze customer behavior with regard to default in the third research question:

• RQ3: Does MCD in combination with incentives (25 % discount) lead to more default?

3 Empirical analyses

3.1 Data

For the empirical analysis, we selected the telecommunications industry and assembled a unique dataset from a large sample of DSL customers consisting of survey, usage, and advertising data. The combination of survey and objective data helps avoid a common-method bias for our analyses (Podsakoff et al. 2003). DSL Internet access is a complex service whose quality cannot be assessed a priori and which is offered in a market with considerable competition. Therefore, its adoption is often laden with considerable customer uncertainty and provides thus an adequate research setting. In this case, customers interested in flat-rate Internet and phone access were able to choose between subscription plans with either a minimum contract duration of 24 months (MCD) or no subscription commitment (NoMCD, i.e., service can be canceled with 30 days prior notice) for a technically identical service at the same price. After the 24-month term, the customers have the opportunity to cancel the contract on a monthly basis. Whereas both contract options included free hardware, the company provided financial incentives in the form of a reduced fee for the MCD contract (equal to a discount of 25 % over the subscription period).

The data were collected in cooperation with a major European Internet Service Provider (ISP), allowing us to track the underlying motives and defection behavior of a cohort of approximately 80,000 customers who joined the company during October/November 2008 over a period of 36 months. First, we analyze customer defections based on the behavioral data of the whole cohort to analyze whether the churn rates and customer tenure of MCD customers differed from those of NoMCD customers (RQ1). Likewise, we use the behavioral data to analyze whether MCD customers engage in default (RQ3). After 1 year into the contract, we conducted an online survey and collected the responses from 1,342 customers of the cohort.¹ To test for a non-response bias, we compared the means of key demographic variables for respondents and non-respondents (information provided by ISPs) and found no significant differences between respondents and non-respondents (Armstrong and Overton 1977). We use the survey data in combination with advertising data to investigate the impact of MCD on customer churn for RQ2.

While the overall cohort divided into 40,782 MCD (51.17 %) and 38,912 NoMCD subscribers (48.83 %), the survey sample consisted of 710 MCD (52.91 %) and 632 NoMCD customers (47.09 %). Information on the composition of the sample and the differences between 2-year and monthly subscribers is provided in the Appendix. Overall, NoMCD customers have more experience with previous providers, are slightly younger, and are predominantly male (69.8 versus 58.7 % of MCD customers). With respect to the absolute importance of incentives, customers that chose MCD contracts named the reduced fee as well as the free hardware significantly more often as the reason for choosing the provider than NoMCD customers (72.4 versus 48.7 %).

3.2 Impact on customer churn

We used the behavioral data of 79,694 subscribers to answer whether an MCD helps companies to increase customer retention (RQ1). Because we observed the behavior of a cohort of customers that is right censored, we analyzed the rate of churned customers as well as the customer tenure. The churn rate refers to the share of customers who could not be retained during the 36 months of our observation period. As the churn rate only indicates how many customers defected, we also provide information on the customer tenure by providing the average number of months the relationship of the defecting customers lasted. To test the effect of MCD on customer retention, we compare customer churn and tenure for the two contract options (with MCD versus without MCD) and analyze whether customer behavior significantly differs between them.

The results of the chi-square test indicate that the two contract options differ significantly with regard to customer churn, as well as customer tenure. We find that the overall churn rate is much lower for MCD customers than for NoMCD subscribers. Specifically, during the observation period of 36 months, MCD customers churned 27.1 % less than NoMCD customers. This finding is supported by the significantly longer customer tenure for MCD customers. Whereas the average tenure for MCD customers is 28.5 months, NoMCD customers only averaged 24.9 months.

Clearly, the lower churn rates and the subsequent longer tenure for MCD customers suggests positive results of MCD for service providers, as a substantially higher rate of MCD customers remained loyal compared to NoMCD customers. Whereas NoMCD customers show relatively consistent churn rates across the 36 months, the results also indicate that MCD customers show quite different churn behavior during and after the MCD. As expected, we found a lower churn rate for the time of the MCD, the first 24 months (−30.4 % compared to all 36 months). Please note that under special circumstances, customers may terminate their contracts even before the end of the MCD for legal reasons (e.g., in case of relocation to an area that the service provider does not provide coverage or in case of service/technical failures when the service provider is not able to fulfill the service). However, for the time period after the end of the MCD (months 25–36), we find a significant increase in the churn rate (+60.8 % compared to all 36 months). This development is mainly driven by an extraordinary churn rate after the MCD (the churn rate in month 25 is over seven times higher than the average). This finding indicates that a substantial number of customers exist who wait for the minimum contract duration to end and terminate the contract directly afterwards.

The previous analysis shows that customers exist who demonstrate exactly the early defection behavior addressed in RQ2. For a more detailed analysis, we applied a survival model approach and test whether the end of the MCD marks a motivation to end the customer relationship. Whereas the behavioral data on (voluntary) customer churn (1 = yes; 0 = no) was used as dependent variable, we based the independent variables on the survey and the advertising data (see Appendix). Specifically, the dichotomous variable (END24 _it ) captures the effect that the end of the MCD contract might have. The variable takes the value of 1 in the last period of an MCD contract (i.e., the 24th month) and is 0 otherwise.

To test whether further effects influence this particular churn behavior, we analyze four moderating variables that have been deemed important by previous research and can also be influenced by companies. First, finance, time, or convenience risks may occur for customers when setting up and maintaining a new service. Previous literature has established that a positive relationship exists between perceived risk and customer retention (Hoover et al. 1978). In the case that customers perceive risks (e.g., of financial or technological nature; Eroglu and Machleit 1990) associated with switching providers, we assume that these customers are less likely to defect.

Second, considering the fact that customers also accrue costs (either monetary or non-monetary) by switching from one service provider to another, those switching costs most likely decrease customer defection (Lam et al. 2004). Mooring effects occur when customers perceive the risks to be likely or switching costs to be high (Bansal et al. 2005).

Third, customer satisfaction is generally known to serve as a dominant antecedent of customer retention (e.g., Anderson et al. 1994) and determinant of customer commitment (Farrell and Rusbult 1981), but this variable also works as a push factor for customer switching (Bansal et al. 2005). In the latter case, satisfaction works as an evaluative driver for service switching that motivates customers to leave a provider at the end of the MCD.

Finally, as mentioned above, MCD contracts convey the anticipated relational exchanges between customers and their service provider and disclose the commitment and long-term orientation of customers (Arndt 1979; Dwyer et al. 1987). Hence, we test whether the subscribers’ long-term orientation influences the churn behavior. In the survey of 1,342 customers, we captured the respondent’s satisfaction with the contract terms with a single-item measure (SAT _i ; Bergkvist and Rossiter 2007). We measured the construct perceived risk with three items (RISK _i ; especially capturing risks of financial and technological nature; based on the scales of Eroglu and Machleit 1990) and used three direct items to represent the specific context of the switching costs (SWITCH _i ; adapted from Heide and Weiss 1995). Considering the technically complex nature of the service, the items focused on switching costs that especially occur due to technical incompatibilities (Lambrecht and Tucker 2012). As perceived risk and switching costs evolve as psychological costs to the customer (Klemperer 1995), the related constructs capture the respondent’s perception of their search and switching costs, which do not necessarily reflect actual costs. Furthermore, we used three items to measure the long-term orientation (ORIENT _i ; adapted from Ganesan 1994). A seven-point rating scale was used for each survey item. With regard to measurement reliability and validity, the results indicate acceptable psychometric properties for all constructs. For the formative constructs, no collinearity was observed (Diamantopoulos and Winklhofer 2001; see CI and VIF in the Appendix). In addition, we found no issues regarding discriminant validity (Fornell and Larcker 1981).

Given that it is possible for MCD subscribers to churn within the first 24 months (e.g., due to relocation or service failure), we use a hazard model assuming that all customers will eventually defect (Steenkamp and Gielens 2003; Prins and Verhoef 2007). Hence, the probability of defection λ _i (t) over time (t = 1…36) is expressed as an exponential function as displayed in Eq. 1:

$$ \begin{array}{l}{\lambda}_i(t)={\lambda}_{0i}\cdot \exp (\upsilon +{\mu}_1\cdot \mathrm{P}\_{\mathrm{CONTRACT}}_i+{\mu}_2\cdot \mathrm{END}{24}_{it}+{\mu}_3\cdot {\mathrm{SAT}}_i+{\mu}_4\cdot {\mathrm{RISK}}_i\\ {}\kern7em +{\mu}_5\cdot {\mathrm{SWITCH}}_i+{\mu}_6\cdot {\mathrm{ORIENT}}_i+{\delta}_1\cdot \mathrm{END}24\times {\mathrm{SAT}}_{it}\\ {}\kern7em +{\delta}_2\cdot \mathrm{END}24\times {\mathrm{RISK}}_{it}+{\delta}_3\cdot \mathrm{END}24\times {\mathrm{SWITCH}}_{it}\\ {}\kern7em +{\delta}_4\cdot \mathrm{END}24\times {\mathrm{ORIENT}}_{it}+{\gamma}_1\cdot {\mathrm{AD}}_t+{\gamma}_2\cdot {\mathrm{GENDER}}_i\\ {}\kern7em +{\gamma}_3\cdot {\mathrm{AGE}}_i).\end{array} $$

(1)

In this model, λ _0i represents the baseline hazard with a Weibull distribution,² estimated for each customer (i) by taking into account the different standardized explanatory variables of the main effects (μ) and the interaction effects (δ), as well as the control variables (γ). To control for heterogeneity, we included demographic information (AGE _i and GENDER _i ) and the influence of the competition’s advertising. Here, we analyzed the advertising expenditures (covering TV, radio, print, online, direct mail, outdoor) for the relevant market for the 36 months of our analysis. In this way, we captured customer exposure to competitors’ advertising by considering the share of competitors’ expenditures relative to the total market spending (i.e., share of voice). The competitor advertising (AD _t ) is modeled as a time-varying control effect in our model.

To account for a potential selection bias, we include the predicted values from a first-stage logit model as the instrument for contract choice in the second-stage survival model (P_CONTRACT _i ). In the first stage, we estimated Eq. 2:

$$ \begin{array}{l}\mathrm{Logit}\left({\mathrm{CONTRACT}}_{i,1/0}\right)=\alpha +{\beta}_1\cdot {\mathrm{FLEX}}_i+{\beta}_2\cdot {\mathrm{INCENT}}_i+{\beta}_3\cdot {\mathrm{ORIENT}}_i\\ {}\kern11em +{\eta}_1\cdot {\mathrm{GENDER}}_i+{\eta}_2\cdot {\mathrm{AGE}}_i+{\varepsilon}_i\end{array} $$

(2)

which included the dependent variable contract choice (CONTRACT _i ; provided by the telecommunication company and coded as a dichotomous variable; MCD contract = 1, NoMCD contract = 0), two items to measure the importance of incentives (INCENT _i ), as well as the three items capturing the long-term orientation (ORIENT _i ), four indicators to measure the construct flexibility (FLEX _i ; adapted from Wakefield and Barnes 1996), and the demographic information on gender and age.

The results of the first-stage logit model as well as the Weibull proportional hazard model (Srinivasan et al. 2004) are reported in Table 1. Judging by the likelihood ratio test (χ ²(5) = 147.91; p < 0.01) and a correct classification of 63.6 % of cases (compared to proportional chance criterion (PCC) = 50.2 %, maximum chance criterion (MCC) = 52.9 %; Morrison 1969) for the logit model as well as the McFadden’s R-squared, the fit of the two-stage model is acceptable. The results show that customers select the contract options based on quite distinct preferences. In line with the theoretical considerations, long-term orientation significantly stimulates the adoption of MCD contracts (β ₃  = 0.212; p < 0.01).

Table 1 Results for drivers of churn

In the survival model, we find a significant and negative main effect of both contract choice (μ ₁ = −1.000; p < 0.05) and satisfaction (μ ₃  = −0.076; p < 0.05) on the hazard rate. Unsurprisingly, this implies a significant difference in the contract choices where NoMCD subscribers have a higher churn rate than MCD subscribers. Also conforming to expectations, the result shows that the propensity to defect is lower for satisfied customers than for unsatisfied customers. Furthermore, we find that the main effects of the proposed mooring factors significantly influence the hazard rate, as customers with high perceived risk (μ ₄ = −0.155; p < 0.05) and high switching costs (μ ₅ = −0.127; p < 0.05) have a significantly lower probability to defect. Most importantly, we find a significant influence of END24 _it (μ ₂  = 2.107; p < 0.01). Thus, in line with the previous section’s findings, the analysis indicates a higher churn rate at the end of the MCD.

With regard to the interaction effects, the findings of the survival model are also interesting. Here, only the moderating effect of the end of the MCD on the satisfaction-defection link (δ ₁ = −0.043; p < 0.05) showed a significant influence. The result indicates that the propensity of MCD customers to churn immediately after 24 months does indeed depend on the customers’ satisfaction with the service―the less satisfied, the more prone to churn. This holds especially for price sensitive customers that choose the MCD service primarily for the reduced price. These customers may be less satisfied and, consequently, churn as soon as possible.

Whereas high perceived risk and switching costs usually prevent churn (see main effects), their interaction effects are insignificant. Hence, we find that customers with high perceived risk and high switching costs do not particularly churn more at the end of the MCD. Interestingly, neither the main effect nor the interaction effect of long-term orientation is significant. The service-related and more immediate influences satisfaction, perceived risk, and switching costs apparently supersede the customers’ long-term orientation.

Finally, the results indicate that younger customers show higher defection rates than older customers (γ ₃ = −0.022; p < 0.01). While competitor advertising significantly affects defection (γ ₁ = 0.154; p < 0.01), customer gender does not.

3.3 Impact on default

As indicated above, negative consequences of MCDs may also include default. To analyze the effect of MCD in combination with incentives on default behavior, we use again the behavioral data on the cohort of 79,694 subscribers. For those customers with terminated contracts, we determined whether it was customer-induced (voluntary) churn or initiated by the company because of default (see Table 2). However, only 52 % of the MCD subscribers’ churn can be accredited to such reasons; in 48 % of the cases, the company actively terminated the contract (e.g., when customers fail to make their payments). The rate of defaulting subscribers within the group of NoMCD subscribers (without incentives) is considerably lower (only 13 % of overall NoMCD churn involuntarily). Apparently, the incentives related to MCD contracts attract a much higher proportion of customers with unacceptable payment behavior, indicating that incentives can lead to opportunistic behavior and a much higher (involuntary) churn rate.

Table 2 Distribution of customer default

Overall, the results provide clear answers to the research questions. The analyses show that MCDs do lead to higher customer retention and longer customer tenure. Interestingly, this holds despite the fact that MCDs in combination with incentives also have negative ramifications. Not only do a substantial number of customers churn at the end of the MCD, but they also default significantly more often.

4 Discussion

4.1 Implications

Despite the economic relevance of subscription-based services for telecommunications, media, or utilities, empirical studies are scarce on the impact of minimum contract durations in combination with monetary incentives on customer retention. Using a unique dataset that combines psychometric survey data with behavioral data and advertising expenditures, this study offers interesting insights.

In particular, the results support companies’ rationale to use MCDs for enhancing customer relationships. By addressing long-term oriented customers, we find that MCDs provide a suitable means for companies to increase customer retention, tenure, and ultimately revenues as MCD customers defect significantly less than NoMCD customers.

Notwithstanding, the incentives that companies need to provide for MCD customers to make the commitment have negative effects on the type of customers attracted to the service. Specifically, we find that a substantial number of MCD customers only stay for the MCD. These customers commit to the MCD, receive the incentives in return but defect right after the end of the 24-month MCD. While this behavior may still seem like a fair deal for the company, it neither helps the company with respect to establishing long relationships with the customers nor does it help to increase profits.

A further negative behavior that could be caused by incentives is default. In comparison to non-MCD customers, there is a significantly larger share of MCD customers whose contract was terminated by the company. The provision of such contracts with incentives is obviously a double-edged sword. Offers with MCD and incentives may help acquire customers as providers might tap into different (potentially more price sensitive) customer segments that would otherwise be difficult to reach but may at the same time attract undesirable (i.e., defaulting) consumers. Such customers behave opportunistically and select contracts based on their preference for the incentives. The companies’ rationale of improving customer retention and tenure by providing incentives is evidently not yielding the desired outcome among a considerable number of customers.

The results, however, provide ways for companies to use MCD in combination with incentives profitably. First and most obviously, companies should decrease default among MCD customers. Appropriate screening mechanisms such as evaluating credit scores should be applied to identify prospects likely to default before they become customers. Sophisticated analytical techniques exist to assess the risk levels of applicants (e.g., Yu et al. 2009). Second, high levels of customer satisfaction significantly reduce MCD customers’ likelihood to churn after 24 months. Since MCD customers apparently honor service quality with loyalty, companies can positively influence customer retention by maintaining high service quality levels and customer satisfaction. Third, since incentives appear necessary to attract customers to MCD contracts, companies need to carefully determine the value of the incentive needed to remain profitable. To capture the profitability implications of incentives, it would be necessary for companies to account for the differences in customer tenure and revenue dilution (i.e., the ability of these contracts to target at different customers segments) between MCD and NoMCD contracts.

4.2 Limitations

Some limitations must be acknowledged, which provide avenues for future research. First, our study was conducted in a market with rather stable prices over time. Given the effects of prices on the customer-provider relationship (Bolton and Lemon 1999), binding customers with MCD contracts may have the additional benefit of locking customers into a rather high fee agreement in a market with a strong overall price decrease. Future research could analyze MCD versus NoMCD contracts in markets with declining prices.

Second, the market for Internet access is highly competitive, resulting in above-average levels of customer awareness in this service category. Thus, further research could replicate the results for subscription-based services in different, less competitive industrial settings.

Finally, in light of the finding that companies can determine the range for the incentives, we suggest further research on the absolute size of the incentive and its effect on contract choice. For example, it might be interesting to experimentally analyze the incentive thresholds for changing contract options.

Appendix

Table 3 Measurement and descriptives of variables