How do we find out what actually works in mobile advertising?

When Facebook went public two years ago, many criticized it for having almost no presence in mobile advertising. In the last quarter Facebook silenced all its critics by reporting that mobile advertising brought in more than 60% of its $2.9 billion ad revenue.The urgency with which Facebook went after mobile just goes to show how important mobile advertising has become to marketers. And why not?With more consumers glued to their smartphones, it is only natural that marketers reach them there. However, many marketers are not yet sure of what works on mobile. Folk wisdom is that contests, games, and coupons should attract more attention and lead to more efficient mobile advertising. But will it work for every product? In a recent academic study, researchers found that mobile advertising is effective only for utilitarian and high-involvement products. Although these results can’t be generalized easily, they suggest that there might be systematic differences in mobile ad campaign effectiveness based on factors such as the type of product, type of consumer, and temporal aspects such as time, day, month, etc. Marketers need to put on their researcher hats in order to find out what does and doesn’t work in mobile advertising.Micro-ExperimentsMicro-experiments (MEs) are short randomized experiments that can be performed continuously. What kind of MEs should a mobile marketer conduct? There are almost infinite possibilities. Therefore, marketers must identify the most critical factors that they believe affect consumer behavior.For example, if you want to run a contest then a mobile ad that uses the color red is likely to drive more engagement. This is not an established fact; however, academic research shows that red makes people more competitive, which can help companies such as eBay make people bid more aggressively. Therefore, extrapolating this logic to mobile advertising can lead to a simple hypothesis that you can test easily.Similarly, men and women show wide variation in their risk preferences. This knowledge could be used to formulate a hypothesis and MEs can be designed to specifically test campaigns where the risk preference becomes gender neutral.With MEs the main threat is marketers going all over the place rather than narrowly focusing on certain key hypotheses that they want to test. Therefore, I suggest marketers should pay attention to the new knowledge created in academia and not what’s published in pop psychology books by the likes of Malcolm Gladwell.AttributionOne of the key challenges to mobile, if not all of digital advertising, is to extract the attribution. What’s the ROI of your mobile campaign? What if people are watching your ads on mobile but buying the product on the PC? How will you attribute these purchases to mobile advertising? Although there are many solution providers in this space, I suggest you additionally use MEs to get an idea for yourself. For example, you can set up a simple experiment whereby you expose around 10% of your customer base to a mobile advertisement. This has to be a random assignment, meaning you shouldn’t select the customers based on any specific criterion such as geography or age. Next you compare the purchase behavior of these customers with that of the other 90% (or a separate smaller randomly selected sample that wasn’t shown the mobile ad). As the only parameter you changed was the presence or absence of the mobile ad, you will be able to attribute the difference in purchase behavior to the mobile ads with much greater confidence. By conducting this ME regularly, you can get an estimate for the mobile ad attribution, which you can use as a benchmark.AnalyticsMEs are meaningless if you don’t have capabilities to analyze the data. However, for MEs you don’t need 'big data' analytics expertise. Instead you can use simple data analysis tools to determine how your MEs are performing. As you control the sample size of your MEs, you could tailor them to be analyzed in Excel. Multiple regression analysis, analysis of variance, and other simple statistical tests can take you a long way in analyzing your data.The author is Ashwin Malshe, assistant professor of marketing at ESSEC Business School.Malshe will be sharing more on mobile marketing strategies at Marketing magazine’s annual Mobile Marketing Interactive conference, on 4 September.To attend the event, please contact Carlo Reston at carlor@marketing-interactive.com or call +65 6423 0329. Further information can be found by clicking here.