Solving the Search vs. Display Advertising Quandary

image:www.ronedmondson.com
culled from:www.hbswk.hbs.edu

Internet advertising was supposed to make it easier for marketers to measure the impact of their ad buys. But a basic question remains: Do search ads or do display ads create more customers on the web? Research by Professor Sunil Gupta.

by Michael Blanding
The dirty little secret of advertising agencies is that much of their work is pure guesswork. Companies spread out their advertising budgets across channels—a little bit of TV, some print media, a few billboards—and wait for customers to roll in. In very few cases, however, do they know if or how these investments motivated consumers to plunk down their credit cards or fill out an application for a service.
That's why the Internet has been such a godsend to companies, says Sunil Gupta, the Edward W. Carter Professor of Business Administration at Harvard Business School. Finally, by measuring a few clicks of a mouse, companies can tell how successful a particular ad was in making a purchase decision.
"A lot of companies have put more and more money into online marketing because online marketing is more measurable, as opposed to television where you put in millions of dollars and you have no idea what you get for it," says Gupta.
When you really start to examine how online advertising works, however, that return on investment may not be quite so easy to calculate. After all, not all online advertising is the same.
“Once you look at those numbers, I saw of course that the company should be spending more money on search ads.”
The most measurable form of online advertising are search ads—ads linked with results from search queries—which can translate directly into clicks and purchases by placing a product before customers at the exact moment they are making a purchase decision. On the other hand, display ads—such as banner ads on websites—are much more visible but are less likely to translate into direct clicks. By raising brand awareness, however, display ads may contribute to a consumer's eventual decision to click on a search ad.
"When you look at most consumers' behavior, they do a search on Google when they are ready to make a decision," says Gupta. "So search ads get all the credit for the sale. But search ads are at the bottom of the [purchase] funnel; display ads are at a higher level. Everybody in the industry intuitively believes that display ads drive people down the funnel, so they should get some credit—but they don't."
Credit where it's due
In advertising industry parlance, such an issue is called an "attribution problem"—that is, you can't tell which ad to attribute the eventual decision to buy. Faced with this conundrum, most companies allocate their advertising budgets in a very ad hoc manner—throwing money into whatever bucket they perceive to have most influenced past purchase decisions. Some have gone a step further by running experiments with participants to determine which type of ad is more effective. While that might help pinpoint the immediate persuasiveness of ads, it isn't much help to measure their impact over time.
The only way to truly determine the efficacy of display ads versus search ads is to watch the effects over time, and to see how modifications in budget allocations change customers' purchase decisions. Luckily, Gupta had access to just such data.
 Initially Gupta wrote a case study for his MBA second-year course on digital marketing about a major US bank that was deciding where to place its marketing dollars to acquire new customers. Later on, Gupta was able to obtain data weekly from the bank's advertising agency on how much money was spent and how the funds were allocated. In addition, he received data on how many consumers "converted" after clicking on search and display ads. Because a decision to open a bank account is probably a consideration one makes over time—say weeks or months in advance—Gupta could use that historical data to see how spending on different types of ads affected those decisions.
The findings from the research were recently published in the working paper Do Display Ads Influence Search? Attribution and Dynamics in Online Advertising, cowritten with HBS doctoral student Pavel Kireyev and Ozyegin University professor Koen Pauwels. In it, the authors ask two questions: First, do display ads really make a difference in pushing consumers "down the funnel" to making a purchase decision, and second, how big of a difference did they make and for how long?
Of course, implicit in those questions is a third question, says Gupta: "If you can quantify the effects, then the obvious question is how should I shift my budget to maximize my return on investment?"
Display ads effective
While they might seem like simple determinations, the calculations involved in answering those questions are incredibly complex. Gupta and his colleagues solve them using a technique called "persistence modeling." Simply put (and the calculations are anything but simple), you first calculate the expected effect of advertising, and then see how changes in ad budgets change those expectations over time. Employing a series of regressions, the researchers were able to isolate the effects of display advertising versus search advertising.
Gupta says that when he looked at the numbers, he wasn't surprised to find that display ads actually did affect the number of people who eventually clicked on search ads and became customers—with an increase of 5 million display ad impressions resulting in an additional 20 checking account applications through search after about five weeks.
Based on that finding, one might predict that the bank should increase its display ad budget. But not so fast, says Gupta. Two other findings, in fact, would appear to favor search ads instead. First of all, search ads seemed to have much longer-term effects than display ads—that is, money spent several weeks ago on search ads had a significant impact on consumers' decision in the current time period.
Second, because display ads drive consumers down the funnel so that they click more on search ads and eventually become a customer, this process also increases overall ad cost due to additional search clicks. Therefore, the benefit of display ads needs to be adjusted against this additional search cost. In fact, the researchers found that in the long run the cost for each bank account application for search ads was $38. But the cost per application for display ads was actually $57.
"Once you look at those numbers, I saw that the company should be spending more money on search ads," says Gupta, admitting "that's a bit counterintuitive given where we started."
Of course these findings are specific to the particular case: consumers opening accounts within the banking industry. Different industries may see different results based on the particular nature of how people search for products and services online. But in the end, the findings show that the decisions about where to place advertising dollars are anything but simple.
A scientific assessment of exactly how changes in budget allocations affect the eventual purchase decisions of customers, however, can take out the guesswork and help companies determine the best formula for its particular case. 
About the author
Boston-based writer Michael Blanding is a fellow at the Edmond J. Safra Center for Ethics at Harvard University and author of The Coke Machine: The Dirty Truth Behind the World's Favorite Soft Drink.