Advertising Effectively Influences Older Users: How Field Experiments Can Improve Measurement and Targeting (original) (raw)

Abstract

Does advertising measurably affect sales? New technologies for tracking individuals’ sales and ad exposure facilitate studying a nationwide retailer’s online brand advertising effectiveness. A controlled experiment on 1,577,256 existing customers measures the causal effect of advertising on purchases, overcoming the attribution problem by exogenously varying ad exposure. The advertising produced a statistically and economically significant effect on in-store sales. The experiment permits a demographic breakdown. Surprisingly, the effects are large for the elderly. Customers aged 65+, comprising only 5 % of customers, increased purchases by 20 % due to the advertising. This represented 40 % of the total effect among all ages.

Access this article

Log in via an institution

Subscribe and save

Buy Now

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

Instant access to the full article PDF.

Similar content being viewed by others

Notes

  1. Ackoff and Emshoff (1975a); Ackoff and Emshoff (1975b) made no attempt to quantify their statistical uncertainty, but Allaire (1975) pointed out that the interesting effects they observed were not statistically significant.
  2. The self-reported age and gender data is not perfectly accurate. We know some users lie about their demographics while filling out online forms, but Fig. 2 does not raise any obvious red flags about deceptive reporting. In any case, inaccurately reported age data would likely make it harder to find differences among differently aged consumers.
  3. We were not able to exclude these potentially invalid observations from our dataset, but they appear to represent a very small part of the sample. To the extent that we include treatment-group members who could not actually have perceived the ads, we will have a slight bias towards measuring zero effect of the advertising.
  4. While a 10 % one-sided test is a much lower standard for statistical significance than is usual in the economics literature, this is a more stringent standard than the threshold used for success in the BehaviorScan experiments of Lodish et al. (1995a), who used a standard of 20 % significance, one-sided. Because of the high variance in sales, we have relatively imprecise estimates of the effects of advertising even with sample sizes in the hundreds of thousands. For more detailed power calculations concerning this experiment, see Lewis and Reiley (2013).

References

Download references

Acknowledgments

We thank Meredith Gordon, Sergei Matusevych, and especially Taylor Schreiner for work on the experiment and data. Yahoo! Incorporated provided financial and data assistance and guaranteed academic independence prior to our analysis so that the results could be published no matter how they turned out. We acknowledge the helpful comments of Avi Goldfarb, Glenn Ellison, Jerry Hausman, Stephen Ryan, Duncan Watts, and two anonymous referees.

Author information

Authors and Affiliations

  1. Google, 1600 Amphitheatre Parkway, Mountain View, CA, 94043 , USA
    Randall A. Lewis & David H. Reiley

Authors

  1. Randall A. Lewis
    You can also search for this author inPubMed Google Scholar
  2. David H. Reiley
    You can also search for this author inPubMed Google Scholar

Corresponding author

Correspondence toDavid H. Reiley.

Electronic supplementary material

Rights and permissions

About this article

Cite this article

Lewis, R.A., Reiley, D.H. Advertising Effectively Influences Older Users: How Field Experiments Can Improve Measurement and Targeting.Rev Ind Organ 44, 147–159 (2014). https://doi.org/10.1007/s11151-013-9403-y

Download citation

Keywords