This paper describes in meticulous detail a study on a specific marketing strategy pursued by American Express and many other companies. American Express makes a promotional offer using Twitter. A consumer accepts and later confirms the offer also using Twitter. The consumer must have a public Twitter account and have linked the account with their American Express account, with what is termed an “easy opt-in process.” The decision to accept the promotional offer, that is, to patronize a brand, is broadcast to the consumer’s own set of followers.

Data mining techniques are used to define a model of normal additions to the fan base for each brand. Using econometric techniques, the numbers of added followers beyond normal expectations were found to be substantial and statistically significant (0.0020, p < 0.001). The study also supported the hypotheses that this effect was less for brands with a smaller existing fan base, and that promotional messages broadcast at times of peak usage produce better results. The study is positioned in terms of previous research in social marketing and other related areas such as event studies. For example, the authors apply a methodology of causal inference similar to what is done in finance. The authors performed what they term robustness checks and falsification tests to confirm that the effects discovered were real and not accidental. Though the writing is often non-idiomatic English and some captions for figures are confusing, the paper is worthy of study by researchers and marketing professionals.