Abstract

Using a multiyear experiment, we show that advertising to businesses can generate very different responses than has been observed for consumers. First, we estimate larger advertising returns than typically found for consumers: a return on ad spend of 12.0 (95% CI: 4.8-24.5). Second, we estimate longer lags between ad delivery and purchase: 1-5 months for first-time purchases and 5-12 months or longer for repeat purchases. Third, we find that existing business customers are responsible for most of the revenue lift from advertising, though this appears to be driven by existing business customers purchasing new parts. Additionally, our results demonstrate that geography-based switchback experiments that randomize the time between treatments can provide an effective means of estimating lagged effects. For this study we randomized the delivery of digital display ads for electronic components offered by a semiconductor manufacturer; nearly all electronic products require such components.