Customer satisfaction, loyalty behaviors, and firm financial performance: what 40 years of research tells us

Extant research has linked CS to four customer-level outcomes (retention, WOM, price outcomes, and spending outcomes) and ten firm-level outcomes (e.g., sales, cash flow, stock returns, and Tobin’s q). Their definition, measures, and respective calculations are shown in Table 2, panel A.

Table 2, panel B reports the nine moderators examined in this meta-analysis. These include (1) contextual factors such as type of exchange and location of study and (2) measurement characteristics including the number of items and the number of scale points in the CS measure, the source of CS measure (e.g., ACSI), the calculation of CS score (e.g., top-box score), and the measurement of outcome (e.g., behavior).¹

We identified studies using computerized searches of Web of Knowledge, ScienceDirect, and EBSCO with the keywords “customer satisfaction” and “consumer satisfaction.” We examined each issue of the major marketing journals in the USA and Europe starting from 1980.² Prior to 1980, CS research focused on its antecedents. We also reviewed and included pertinent articles from the three meta-analyses in Table 1.

A study was excluded if it: (1) measured satisfaction with specific attributes but not overall satisfaction, (2) used a composite measure of multiple outcomes (e.g., latent construct of repurchase and recommendation), and (3) did not report correlations or information that could be converted to correlations.³ When a study provided multiple effect sizes, either for separate samples or relationships, we treated effects as independent. When a study provided multiple effect sizes for the same relationship (e.g., for subsets of the same sample), we calculated the average effect size. The final analyses use 535 correlations from 245 articles (N = 1,160,982).

We calculate inverse-variance-weighted reliability-adjusted correlations between CS and each outcome (Hunter and Schmidt 2004). To adjust for reliability, we use Cronbach’s alpha (Nunally 1978) as a reliability measure and divide the raw correlations by the square root of the product of reliabilities of CS and the outcome. We are unable to correct for reliability for firm-level outcomes because they use a single metric based on archival financial data. We then transform the reliability-adjusted correlations to Fisher’s z coefficients and weight them by the inverse variance (i.e., 1/[N – 3]). Finally, we transform the Fisher’s z coefficients back to correlations to arrive at the weighted reliability-adjusted correlations.⁴ The analyses use a random effects approach for effect size integration.

Table 3, panel A reports that CS has a strong association with retention (r = 0.60, p < 0.01) and WOM (r = 0.68, p < 0.01) and is moderately correlated with spending (r = 0.28, p < 0.01) and price outcomes (r = 0.39, p < 0.01).⁶ The statistically significant Q tests (ps < 0.01) for all four outcomes indicate that effect sizes may vary based on exchange type, market type, location of study, measurement of outcome, scale items, and scale points. Disaggregated results are shown in panel A of Table A2 in Web Appendix A and discussed next.

The CS-outcomes correlation is smaller at the firm level than at the customer level (see Table 3, panel B) potentially because firm-level outcomes are more distal than customer-level outcomes. Different than the association of CS with customer-level outcomes, the magnitude of the association of CS with firm-level outcomes can be classified as small to moderate. ⁸

Specifically, CS has a positive and statistically significant association with sales (r = 0.15, p < 0.01), profit (r = 0.10, p < 0.01), ROA (r = 0.22, p < 0.01), Tobin’s q (r = 0.29, p < 0.01), and stock returns (r = 0.08, p < 0.05); a negative and statistically significant association with cash flow variability (r = –0.10, p < 0.01), risk (r = –0.23, p < 0.01), and cost of debt financing (r = –0.14, p < 0.01). CS has a nonsignificant association with market share (r = 0.05, p > 0.10) and a weak positive association with cash flow (r = 0.09, p < 0.10), which may occur because they likely represent multiple subgroups with large between-group variability in the association (Whitener 1990). ⁹

The Q statistics for all outcomes, except for cost of debt financing, indicate a statistically significant heterogeneity among studies (see Table 3, panel B). Yet, with a small number of exceptions, the association between CS and firm-level outcomes is not statistically different across subgroups based on different levels of moderators (see panel B of Table A2 in Web Appendix A). There are several potential reasons for the statistically nonsignificant results. First, for several moderator levels, each outcome has been investigated by a small number of studies (i.e., k in panel B of Table A2 in Web Appendix A). Second, most of the firm-level studies include samples from multiple industries and preclude us from isolating correlations based on specific industry settings. Finally, published studies typically do not report correlations disaggregated by firm-level moderators such as firm size, advertising and R&D intensity, and industry concentration. Therefore, we report means by subgroups for firm-level outcomes but do not discuss them further.

First, the moderator analysis shows that there is substantial and systematic heterogeneity in the positive association between CS and customer-level outcomes. Yet, we do not understand the different patterns of variability and their implications. As an example, the association of CS with price outcomes is more heterogeneous than its association with spending outcomes across markets, exchange types, and locations of study. Is it because firms have more control on price outcomes but not on spending outcomes? These issues need further research.

Second, studies that simultaneously examine and compare the association of CS with multiple customer-level outcomes under different contexts are needed. Specifically, attention to differences in effect sizes among subgroups as well as their causes and implications is a key research direction.

Third, the association of CS is strongest for WOM, followed by retention, and is the weakest for spending and price outcomes. Future research should develop a conceptual and theoretical framework to understand these relative differences. Thus, is it the case that higher CS is more beneficial for growing new customers than retaining current customers? To the extent that WOM affects the cost of attracting new customers, customer equity research can be expanded by including CS as a contributing factor for retaining current customers and attracting new customers. Third, a wider set of potential moderators including psychological constructs such as trust and commitment as well as structural factors such as company size, industry growth, and competitive intensity should be investigated.

Fourth, these results make a very strong case that consumer behavior scholars should use CS as a consequential dependent variable in their studies. CS has a clear association with actual consumer behaviors and firm-level financial outcomes. Thus, consumer behavior scholars can be reasonably assured that differences in CS are consequential, i.e., predictive of actual consumer behaviors and firm financial outcomes.

Fifth, these results call into question the long-standing insistence on using multi-item scales for measuring CS. The CS-outcomes linkage is impervious to single- vs. multiple-item scales or number of scale points (i.e., 5- vs. 7- vs. 10- vs. 100-point scale). Simple and single item scales suffice; this is an important insight for practitioners who value simplicity to reduce the cost of customer surveys.

Sixth, at the firm level, the mean association of CS with market share (p > 0.10) and cash flow (p < 0.10) is nonsignificant to weak (Table 3, panel B). This may be the case if the association of CS with these outcomes is nonlinear and/or contingent on factors such as firms’ ability to standardize or customize their offerings, the heterogeneity in consumer preferences, and the nature of the offering (e.g., goods vs. services; Anderson, Fornell, and Rust 1997). In the same vein, CS has a stronger association with ROA than with cash flow. While we can speculate on the potential reasons for this, more studies are needed to better estimate the effects and explain the differences. Finally, the small number of studies for subgroups within different levels of moderators precluded specific conclusions; clearly, more studies on CS-firm outcomes are needed.

Figure 1, panel B organizes the outcomes of CS based on their relevance to CMOs, CSOs, CFOs, and CEOs and board members. CMOs who organize their efforts around CS and make CS as their key metric should be able to make a case for their relevance and contribution to customer retention, WOM, spending, and price outcomes. While CMOs are free to focus on other constructs such as net promoter, this research provides clear, strong, and convincing evidence for using CS as a metric to measure marketing and sales performance and relate it to firm performance. Specifically, CS can provide the basis for CMOs and CSOs to collaboratively grow the current customer base organically as well as expand it through additional sales. The positive association of CS with ROA and cash flow and its negative association with cash flow variability speak to CFOs.

In summary, a focus on CS can align C-suite members (CEO, CFO, CMO, and CSO) using a theoretically sound, conceptually consistent, and empirically validated approach. We hope that senior leaders in firms embrace a satisfaction-based approach to strategy planning and execution based on these results.