Cross-Buying After Product Failure Recovery? Depends on How You Feel About It

Abstract

Cross-selling to customers during product failure recovery (PFR) encounters can be challenging as customers are reluctant to cross-buy after having recently experienced a failure, despite it being recovered. We examine several models of cross-buying and failure/recovery characteristics using a large-scale experiment and secondary transaction data from a Fortune 100 computer systems firm. We find that customers’ integral-affective responses dominate their cognitive responses. Further, customers are more willing to cross-buy when the firm’s recovery effort increases for more severe product failures or those with unstable attributions. Yet, greater recovery effort does little to diminish the negative effect of attributing the failure to the firm. Overall, understanding the relative dominance of a sequence of affective versus cognitive factors and the critical role that contextual factors play in customer cross-buying decisions will help managers design PFR encounters to increase the odds of cross-selling.

Notes

1. We refer to product failure recovery (PFR) encounters as a form of customer service in which customers contact the firm about a product failure and the firm’s customer service agent recovers the failure.

2. Cross-selling refers to the sale of an additional product or service to an existing customer and is, therefore, from the perspective of the firm’s actions. Cross-buying refers to the act of buying an additional product or service from the firm and is, therefore, from the customer’s perspective. While insights from this research could plausibly pertain to upgrading and upselling, we contend that in a PFR encounter, customer service agents sell additional products and services but not primary systems, such as an new computer, which can cost thousands of dollars and is more appropriate for a committed sales team.

3. We do not examine attribution controllability, the third dimension of failure attribution (see Folkes 1984), as nonvolitional causes of failure (e.g., weather) are not pertinent in our empirical context. However, insights from this research can be applied to contexts where controllability is pertinent (e.g., in the airlines industry where weather-related failures are common).

4. We are unable to test H2a/b/c and H5 in Study 2 because of the limitations of the call log and transaction data. Accordingly, we rely on insights from Study 1 to fill in these gaps.

5. Sierra and Hyman (2009) compared a dual-process model of cognitive and emotive factors to two single-process models (cognitive-only and emotive-only) using model fit statistics and found that the dual-process model better explains consumer willingness to incur uncertainty. In a similar vein, and consistent with several other studies (e.g., Fornell and Rust 1989; Rust, Lee, and Valente 1995), we hypothesize differences in model fit (e.g., goodness of fit) between these various models to identify which set of variables provides a superior explanation of PFR cross-buying.

6. Because of data limitations, we were unable to measure attribution stability and rely on the results from Study 1 to inform our findings and recommendations regarding this variable.

7. One might argue that recovery effort is merely a function of how severe the failure is. However, we do not find this to be the case. We see that for the very same type of failure with the same resulting severity, the degree of recovery effort and contact time per customer varies considerably, suggesting that recovery effort differs across encounters and is not a function of how severe the failure is. In fact, the correlation between failure severity and recovery effort is only .027. Thus, recovery effort functions independent of the failure characteristics.