Skip to main content
Log in

Including the effects of prior and recent contact effort in a customer scoring model for database marketing

  • Original Empirical Research
  • Published:
Journal of the Academy of Marketing Science Aims and scope Submit manuscript

Abstract

Database marketers often use a scoring model to predict the likely value of contacting customers based on their purchase histories and demographics. However, when purchase history has been a partial result of the firm’s own contacting efforts, these contacts should also be accounted for in the scoring model. The current work extends the existing literature to account for the firm’s contacts by focusing on each customer’s most recent purchase. Contacts prior to that purchase are designated “prior contacts” and those after that purchase “recent contacts.” A new latent variables formulation of the customer’s propensity to respond is used to predict the likelihood and time of response as well as the relationship to the independent variables. The methodology also addresses the statistical problems of “selection bias” and “endogeneity,” which have been largely ignored in most customer scoring models. An application to the database of a charitable organization confirms that, in this case: (1) the effect of the firm’s customer contact efforts is associated with a stronger propensity to respond than is the case for the included demographics; (2) the firm’s “recent contact” efforts are associated with larger returns in customers’ propensity to respond than the “prior contact” efforts; and (3) the “recent contact” efforts are associated with an at-first increasing but then diminishing propensity to respond up to a point beyond which actual decreasing returns are observed with further contacts. Clearly, too much contacting can alienate would-be donors. The proposed model is general enough to calibrate such impacts in other database marketing applications where the relative effects might be different.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Similar content being viewed by others

Notes

  1. We could apply the same method but our approach of using a latent construct can simultaneously correct for the selection bias and the endogeneity problems. Our approach is also applicable to database marketing cases in general without the need of creating new variables (e.g., differences and lags).

  2. There is an obvious analogy between our “recent contact” effects and the psychological literature about the constructs of primacy and recency in memory experiments. However, we shy away from mentioning recency in that framework to avoid too many different constructs and, therefore, to keep to one clear definition of “recency” in this discussion.

  3. Donation amount is assumed constant since over 52% of respondents give a single amount (2, 3, 5, 10, 15, or 25 dollars) and over 95% exhibit a coefficient of variation of amounts given over time that is less than 0.5. Note also that this assumption is of little consequence to the approach developed herein. See the managerial section for how results are to be used without explicitly modeling the purchase or donation level.

  4. The test for a significant difference in the MAD value across models is possible because with the Bayesian estimation used, all of the statistics have resulting errors estimated (including the MAD value).

  5. In some contacting situations, the analyst may not be able to isolate whether response is a direct consequence of the contact (whereas in our application donations are tightly linked to the solicitation mailings since they are the major component of the entire campaign). In more complex solicitation campaigns causal attribution may not be so clean.

References

  • Allenby, M. G., Leone, R. P., & Jen, L. (1999). A dynamic model of purchase timing with application to direct marketing. Journal of the American Statistical Association, 94(446), 365–374.

    Article  Google Scholar 

  • Basu, A. K., Basu, A., & Batra, R. (1995). Modeling the response pattern to direct marketing campaigns. Journal of Marketing Research, 32, 204–212.

    Article  Google Scholar 

  • Bitran, G. R., & Mondschein, S. V. (1996). Mailing decisions in the catalog sales industry. Management Science, 42(9), 1364–1381.

    Article  Google Scholar 

  • Blair, M. H., Rabuck, M. (1998). Advertising wearin and wearout: Ten years later-more empirical evidence and successful practice. Journal of Advertising Research, Sep/Oct, 7–18.

    Google Scholar 

  • Davidson, R., & MacKinnon, J. G. (1993). Estimation and inference in econometrics. New York: Oxford University Press.

    Google Scholar 

  • Elsner, R., Krafft, M., & Huchzermeier, A. (2004). Optimizing Rhenania’s direct marketing business through Dynamic Multilevel Modeling (DMLM) in a multicatalog-brand environment. Marketing Science, 23(2), 192–206.

    Article  Google Scholar 

  • Gardyn, R. (2003). Not giving up. American Demographics, 25, 14–15.

    Google Scholar 

  • Gelman, A., Carlin, J., Stern, H., & Rubin, D. (1996). Bayesian data analysis. New York: Chapman and Hall.

    Google Scholar 

  • Giving USA (2007). The annual report on philanthropy for the year 2006. Retrieved Jan 12, 2008 from http://www.givingusa.org/gusa/gusa_order.cfm.

  • Gönül, F., & Ter Hofstede, F. (2006). How to compute optimal catalog mailing decisions. Marketing Science, 25(1), 65–74.

    Article  Google Scholar 

  • Gönül, F., & Shi, M. Z. (1998). Optimal mailing of catalogs: A new methodology using estimable structural dynamic programming models. Management Science, 44(9), 1249–1262.

    Article  Google Scholar 

  • Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47(1), 153–161.

    Article  Google Scholar 

  • Kumar, V., & Petersen, J. A. (2005). Using a customer-level marketing strategy to enhance firm performance: a review of theoretical and empirical evidence. Journal of the Academy of Marketing Science, 33(4), 504–519.

    Article  Google Scholar 

  • Liu, Q., Otter, T., & Allenby, G. M. (2007). Investigating endogeneity bias in marketing. Marketing Science, 26(5), 642–650.

    Article  Google Scholar 

  • Malthouse, E. C. (2002). Performance-based variable selection for scoring models. Journal of Interactive Marketing, 16(4), 37–50.

    Article  Google Scholar 

  • Reinartz, W. J., & Kumar, V. (2003). On the profitability of long-life customers in a noncontractual setting: An empirical investigation and implications for marketing. Journal of Marketing, 64, 17–35.

    Article  Google Scholar 

  • Rhee, S., & Russell, G. (2006). Measuring household response in database marketing: A latent trait approach,” working paper.

  • Rossi, P. E., McCulloch, R. E., & Allenby, G. M. (1996). The value of purchase history data in target marketing. Marketing Science, 15(4), 321–340.

    Article  Google Scholar 

  • Rust, R. T., Lemon, K. N., & Zeithaml, V. A. (2004). Return on marketing: Using customer equity to focus marketing strategy. Journal of Marketing, 68, 109–127.

    Article  Google Scholar 

  • Simon, H. (1982). ADPULS: An advertising model with wearout and pulsation. Journal of Marketing Research, 19, 352–363.

    Article  Google Scholar 

  • The Direct Marketing Association, Inc. (2004). The DMA 2004 response rate report. New York: The Direct Marketing Association, Inc.

    Google Scholar 

  • Venkatesan, R., & Kumar, V. (2004). A customer lifetime value framework for customer selection and resource allocation strategy. Journal of Marketing, 68, 106–125.

    Article  Google Scholar 

  • Venkatesan, R., Kumar, V., & Bohling, T. R. (2007). Optimal customer relationship management using bayesian decision theory: An application for customer selection. Journal of Marketing Research, 44(4), 579–94.

    Article  Google Scholar 

Download references

Acknowledgement

We wish to acknowledge the helpful comments of four anonymous reviewers and the Direct Marketing Association for donating the dataset for our research.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Subom Rhee.

Appendices

Appendix 1

Table 3 Modeling framework as a system of equations

Appendix 2. Variable definition

Demographics (ZIP level data)

  • Gender: 1 if male, 0 if female

  • Income index: median family income of the ZIP indexed to the state median.

  • % White: percentage of HH occupied with white

  • % Black: percentage of HH occupied with black

  • % Hispanic: percentage of HH occupied with Hispanic

  • % HH with children under 18: percentage of HH with 1 of more children under 18

  • # People per HH: median number of people in the HH

  • Median age: householders’ median age

  • Median school years: median years of school for people aged 25 or more

RFM variables

  • Recency: months since the most recent donation

  • Frequency: number of donations/time on database

  • Monetary: amount donated/time on the database

Contact variables

  • Prior contacts: number of contacts up to the most recent purchase

  • Prior intensity: prior contacts/time on database up to the most recent purchase

  • Recent contacts: number of contacts since the most recent purchase, at the targeting date

  • Recent intensity: recent contacts/time since most recent purchase, at the targeting date

Note: In theory, there may be an effect of targeting efforts after the current targeting date but this effect is minimal because:

  1. 1.

    The targeting interval is usually about 3 months.

  2. 2.

    About 92% of responders do so within 3 months after the offer.

Dependent variables

  • Propensity to respond: An (unobserved) latent construct value on an arbitrary scale (it ranges from −10 to +18 in this estimation).

  • Targeting: A unique ‘solicitation code’ attached by the firm to each mailing. Mailing date is also known. If the HH is selected for contact S = 1, and 0 otherwise.

  • Purchase: A ‘contribution code’ is attached to each response so that by matching the codes the purchase can be attributed to a particular targeting. If the HH is selected and makes a purchase R = 1, otherwise 0.

  • Purchase time: Months from mailing to date of purchase.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Rhee, S., McIntyre, S. Including the effects of prior and recent contact effort in a customer scoring model for database marketing. J. of the Acad. Mark. Sci. 36, 538–551 (2008). https://doi.org/10.1007/s11747-008-0086-0

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11747-008-0086-0

Keywords

Navigation