Customer-directed selling behaviors and performance: a comparison of existing perspectives

A fifth selling perspective was also originally included in the study design—the “Service Orientation in Sales” (SOS) proposed by Keillor et al. (2000). The five item SOS scale was developed to take “into account the seller’s willingness to engage in both selling and non-selling tasks throughout the buyer-seller relationship.” (2000, p. 10). However, based on data collected from salespeople in the first firm we studied (CleanCo), the psychometric properties of this scale were found to be unacceptable (i.e., α = .39; the construct did not appear to be unidimensional; and the resulting loadings were poor). Therefore, this perspective was omitted from subsequent phases of the study and analyses.

While all major construct measures were included in all three survey administrations, slight variations existed. Specifically, for the CleanCo survey, the full scales for both SOCO and ADAPTS were employed. However, our subsequent analyses of that data showed that the newer short forms of both of these scales exhibited superior measurement properties, so only the short forms were used for the rHouse and RentCar studies (which came after the CleanCo study). All of the analyses reported in the manuscript are thus based on these shorter forms. In addition, the measures of trait competitiveness and self-efficacy (which are discussed later in the manuscript) were included only in the RentCar survey. Finally, and as is noted in more detail in the text, the objective performance measures were different across the three companies.

For objective performance, we use a formative epistemic relationship between the measures and its underlying construct. As Diamantopoulos and Siguaw (2006, pp. 266–267) note, a Type 1 error (the underestimation of explained variance on the dependent variable of interest—in this case, performance) may occur if the researcher specifies formative measures in a reflective manner (Law and Wong 1999). Furthermore, the measures of performance reported by the firms we study appear to capture distinct facets of the sales task, and a formative perspective is better suited (than a reflective perspective) to this type of context.

We also considered subjective performance as a formative construct. However, all six subjective performance items are highly correlated with one another, and the loadings that result from a reflective model are all quite large (.8 or higher). Furthermore, running a factor analysis yields a single factor solution, with all loadings .84 or higher. As a result, using a formative epistemic relationship for the subjective performance measures leads to virtually identical results as those obtained when it is reflective. We therefore decided to keep subjective performance as a reflective construct, for three reasons. First, most of the past work that has looked at selling scales has utilized a reflective (rather than formative) presentation of subjective performance, and thus our results are more directly comparable with past work when we also employ a reflective approach (see Franke and Park 2006; Schwepker 2003). Second, the results we obtain using either a reflective or formative approach are essentially the same. Third, we agree with Cohen et al. (Cohen et al. 1990) that when researchers utilize formative indicators they must base their choice of measures on strong theory, and with Bollen and Lennox (1991) that researchers using a formative approach need a “census” of potential measures rather than a “sample” (Hulland 1999). In the present work, we have neither.

This is not the result of high inter-correlations between ADAPTS and the three dimensions of the Selling Skills scale, however. As shown in Table 3, the ADAPTS correlation with interpersonal skills is .39; .40 with salesmanship skills; and .28 with technical skills. Furthermore, multicollinearity diagnostics from the regression models did not indicate any problems.