Skip to main content

Estimating the impact of interacting with sales representatives on customer-specific revenue and churn behavior


This paper uses transaction-level data from a firm in the High Technology sector to compare the revenue garnered from customers who purchase via sales representatives to that garnered from customers who purchase online. The interaction mode may be correlated with the customer’s unobserved valuation for the service, and this potential endogeneity is addressed using an instrumental variable. Specifically, the data indicate whether the customer’s details were recorded in the company’s information systems following an initial contact, such as the download of a free version from the company’s website. The availability of such leads facilitated representatives’ ability to contact customers, generating exogenous variation in the probability of interacting with a representative. Instrumental variable estimates indicate that such interaction lowers the total revenue garnered from a paying customer during the sample period. The primary driver of this result is a churn effect: customers who purchased online displayed a longer overall relationship with the firm, and were more likely to renew an expiring subscription. An alternative explanation, according to which representatives lowered revenues by reassuring customers regarding the adequacy of cheaper versions of the product, is not supported by the data.

This is a preview of subscription content, access via your institution.


  1. Enterprise customers are handled via a specialized sales team, and tend to purchase rather expensive, sometimes tailor-made versions of the firm’s product. Of note, despite the labeling of the online channel as targeting SMB customers, in practice there was no effective mechanism that limited the access of large customers to online purchasing via this channel.

  2. This issue suggests an interesting parallel with the traditional endogeneity problem encountered in the study of the impact of advertising on sales (e.g. Schmalensee 1972; Berndt 1991; Bagwell 2003; Elberse and Anand 2005)

  3. Neslin and Shankar 2009 provide a survey of the empirical literature on this issue. Specifically, Hitt and Frei (2002) and Campbell and Frei (2006) demonstrate stronger client retention in online channels in the banking sector. Danaher et al. (2003) find that brand loyalty in the retail context is strong (weak) when shopping online for products with large (small) market shares. Again in the retail sector, Wallace et al. (2004) find that multichannel shopping increases loyalty. On the other hand, Wright (2002) survey of the impact of online banking cites Beckett et al. (2000) who conclude that it may have pushed customers out of a passive loyalty mode. Ansari et al. (2008) also find that internet usage may reduce loyalty in the financial services sector.

  4. I am grateful to a referee for pointing out this possibility.

  5. They also note that the majority of papers in this literature either do not correct for self-selection, or do so via matching techniques.

  6. A caveat is that, if a customer conversed with a sales team, but did not complete a transaction, and later performed the transaction online creating a new ID, the analysis is likely to miss the fact that interaction with a representative took place. The data do not provide relevant information on this issue, and it is also difficult to gain a qualitative insight into its prevalence.

  7. The data contain very few transactions that cannot be characterized along these two dimensions: six transactions are missing the plan description information, and an additional 263 transactions belong in two other small plans.

  8. Since the transaction time within the day is not recorded, the bounds of (3/31)*99 and (5/31)*99 were used in such cases. The 92 cases alluded to above are those in which the discount fell within the bounds.

  9. These additional regressions are available from the author upon request.

  10. The data also report a “latest marketing” variable which is typically, but not always, identical to the “first marketing” variable that is used in the analysis and discussed above.

  11. A caveat is that some of these customers may have performed their first transaction before the sample period. Notwithstanding this issue, it does not change the point of this analysis: that the technical issue that led to missing channels became less relevant over time.

  12. The need for this manual entry stemmed from lack of synchronization of some of the firm’s information systems.

  13. Such customers may have visited the website, but the visit did not result in a recorded marketing channel that was then associated with their account.

  14. As described above, customers whose first contact with the website took place in January or February of 2012 were much less likely to have their marketing channel recorded than customers whose first transaction occurred in March or the later months. These minor differences in the timing of the first contact are not likely to be correlated with underlying preferences.

  15. In transactions involving multiple months, it is possible that payment is received in installments over time, and some of these payments could, therefore, be received by the firm only after the sample period. For simplicity, I ignore this issue and consider the total amount of transactions signed up for during the sample period.

  16. Some additional regressions below are conducted at the level of the individual transaction. In those regressions, it is understood that the variable W i r e is defined at that level as well.

  17. A referee suggested that using the log of revenues, rather than their level, as the dependent variable will reduce the impact of large transactions. Table 16 in the appendix presents such results, in a manner parallel to Tables 7 and 8 above. As can be seen there, the results using logs are qualitatively very similar to those presented above.

  18. Note that we cannot tell with certainty whether the customer’s first observed transaction is really her first transaction—this would not be the case had the customer performed a previous purchase prior to the observed sample period. Despite this limitation, restricting the sample in this way makes it possible to address the question of interest.

  19. Recall the discussion of this issue in Section 2, where it is noted that, in such cases, customers are reimbursed for the unused portion of the subscription.

  20. When performing regressions at the level of an individual transaction, it is not possible to include a customer fixed effect since the interaction variable does not vary within customer, except in the case of a small number (35) of customers, as discussed in Section 2 above. For this reason, among others, the paper focuses on measures (revenue, duration) that characterize the entire relationship with the customer, and not on transaction-level analyses.

  21. As above, results using combinations of this indicator with indicators for the Direct and Organic marketing channels yield qualitatively similar results.

  22. The same message arises from unreported regressions that perform the same analysis at the single transaction level, available from the author upon request.


  • Albers, S., Murali, K.M., & Sridhar, S. (2010). Personal selling elasticities: A Meta-Analysis. Journal of Marketing Research, 47(5), 840–53.

    Article  Google Scholar 

  • Ansari, A., Mela, C., & Neslin, S. (2008). Customer channel migration. Journal of Marketing Research, 45(1), 6076.

    Article  Google Scholar 

  • Bagwell, K. (2003). The Economic Analysis of Advertising. In M. Armstrong, & R. Porter (Eds.), Handbook of Industrial Organization. Amsterdam: North-Holland.

  • Beckett, A., Hewer, & Howcroft, B. (2000). An Exposition of Consumer Behaviour in the Financial Services Industry. International Journal of Bank Marketing, 18(1), 15–26.

    Article  Google Scholar 

  • Berndt, E.R. (1991). The Practice of econometrics: Classic and contemporary, reading. MA: Addison-Wesley.

    Google Scholar 

  • Berndt, E.R., Pindyck, R., & Azoulay P. (2000). Consumption Externalities and Diffusion in Pharmaceutical Markets: Anti-Ulcer Drugs, NBER Working paper 7772. National Bureau of Economic Research.

    Google Scholar 

  • Campbell, D., & Frei, F. (2006). The cost structure, customer profitability, and retention implications of Self-Service distribution channels: Evidence from customer behavior in an online banking channel. Management Science, 56(1), 4–24.

    Article  Google Scholar 

  • Danaher, P.J., Wilson, I.W., & Davis, R.A. (2003). A comparison of online and offline consumer brand loyalty. Marketing Science, 22(4), 46176.

    Article  Google Scholar 

  • Elberse, A., & Anand, B. (2005). Advertising and Expectations: The Effectiveness of PreRelease Advertising for Motion Pictures. Harvard Business School Working Paper Series, No., 05–060.

  • Gensler, S., Leeflang, P., & Skiera, B. (2012). Impact of online channel use on customer revenues and costs to serve: Considering product portfolios and self-selection. International Journal of Research in Marketing, 29, 192–201.

    Article  Google Scholar 

  • Hitt, L.M., & Frei, F.X. (2002). Do better customers utilize electronic distribution channels? the case of PC banking. Management Science, 48(6), 73248.

    Article  Google Scholar 

  • Horsky, D., & Nelson, P. (1996). Evaluation of salesforce size and productivity through efficient frontier benchmarking. Marketing Science, 15(4), 301–20.

    Article  Google Scholar 

  • Narayanan, S., Desiraju, R., & Chintagunta, P.K. (2004). Return on investment implications for pharmaceutical promotional expenditures: the role of Marketing-Mix interactions. Journal of Marketing, 68, 90–105.

    Article  Google Scholar 

  • Oliver, R.L. (1999). Whence consumer loyalty?. Journal of Marketing, 63, 33–44. Special Issue.

    Article  Google Scholar 

  • Schmalensee, R. (1972). The Economics of Advertising, Amsterdam: North-Holland.

  • Shankar, V., Smith, A.K., & Rangaswamy, A. (2003). Customer satisfaction and loyalty in online and offline environments. International Journal of Research in Marketing, 20, 153–175.

    Article  Google Scholar 

  • Neslin, S.A., & Shankar, V. (2009). Key Issues in Multichannel Customer Management: Current Knowledge and Future Directions. Journal of Interactive Marketing, 23, 70–81.

    Article  Google Scholar 

  • Wallace, D.W., Giese, J.L., & Johnson, J.L. (2004). Customer retailer loyalty in the context of multiple channel strategies. Journal of Retailing, 80(4), 24963.

    Article  Google Scholar 

  • Wright, A. (2002). Technology as an enabler of the global branding of retail financial services. Journal of International Marketing, 10(2), 8398.

    Article  Google Scholar 

Download references


The author is grateful to a company in the High Technology sector for making their data available, and for their helpful feedback. Itai Ater and Saul Lach provided helpful comments. Orry Kaz provided excellent research assistance. This research was supported by the ISRAEL SCIENCE FOUNDATION (grant No. 1338/13).

Author information

Authors and Affiliations


Corresponding author

Correspondence to Alon Eizenberg.

Appendix A:

Appendix A:

Table 16 Regression analysis using log revenues as the dependent variable
Table 17 Additional specifications of the revenue analysis

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Eizenberg, A. Estimating the impact of interacting with sales representatives on customer-specific revenue and churn behavior. Quant Mark Econ 14, 325–351 (2016).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Sales force
  • Online transactions
  • Revenue management
  • Customer churn

JEL Classification

  • M31
  • M54