Who clicks on online ads?

 Who clicks on online ads?

And do they really work

At the outset, let me clarify that this break from a COVID-themed blog does not mean pandemic fatigue on my part. I am religiously (for want of a better word) adhering to mask wearing, physical distancing and the occasional self-quarantine. We still have some way to go before we can afford to lower our guard.

 

I decided to explore this topic because it has been bothering me for a really long time. Maybe it’s just me, but I have rarely ever clicked through an online ad. I am not even referring to the generic banner ads, but the “targeted” ones displayed for our consumption by the devious folks at Google and Facebook. I am thinking, ads for cheap flights that keep popping up, after you’ve made a flight booking. When I see such ads, my train of thought goes something like this: Dude, I just booked one! I won’t be needing another for a while. Besides, if I needed to book flights every week, you can be sure that I am not the type who is looking for “cheap fares”! So I am going to save you some money by not clicking on your ad, because I have no intention of booking another flight. So much for “targeting”!

 

Companies spend billions on these ads, and I naturally assumed that they must be getting their money’s worth. In 2018, the global spending on online ads was USD 278 billion, which, to put it in perspective, is more than the combined GDPs of Bulgaria, Costa Rica, Uruguay, Croatia and Belarus (don’t ask me why I chose these countries. But if you prefer, another metric, this is the equivalent of Chile’s GDP). But then again, there was always that nagging feeling: maybe companies just believe that their ads are working! Can we really tell if online ads work?

 

Mad Men and the illusion of persuasion

Though advertising has been around for well over a century, few people have tried to see if it actually works! As for online ads, most of the influential work is surprisingly quite recent. But stepping back just a bit, the traditional approach to seeing if advertising works has been by relating the increase in sales after a campaign to the cost of the campaign, and calculating the return on investment. But even in conventional advertising (television, print) there is an obvious problem here. As an example, ads for air conditioners go on in the beginning of summer. People buy air conditioners in summer. But this does not automatically mean that the advertising was responsible for the sales. People may have anyway gone ahead and bought them. Economists call this the endogeneity problem. Online ads, specifically the ones triggered by searches (search engine marketing) may appear targeted, but are affected even more by the endogeneity problem. Consider that you wanted to buy an LG microwave oven. Your search results page on google would look something like this.

Note that the top result is a paid ad. Meaning that if you were to click through this, the company would pay google for your click. But since the page ranking algorithm in google has also brought up the company website as the second result, you would probably reach the website even if the company had chosen not to advertise. But if you clicked through the ad and made a purchase, conventional calculations would consider the ad as having contributed to the sale. You can imagine that this (at least in your case) would overstate the importance of the ad. It is easy to see that in order to estimate the true effect of the ad one would need to compare sales in the presence and absence of ads (while maintaining all other factors constant). This is best achieved by doing randomised experiments.

 

A group of economists did just that. They convinced eBay to randomly stop paid ads on different search platforms, and different marketing regions, and estimated the real effect of running online ads. Their results were discomfiting. They showed that for brand-keyword search advertising (such as “LG microwave”; eBay in their case), the return on investment was a negative 63%! The conventional estimate was plus 4000%! Even for non-brand name search ads, traffic reduced by less than a quarter (when they were stopped) but did not lead to a loss in revenue. If you are interested in details, you could look at this working paper (w20171.pdf (nber.org)) which is a surprisingly easy read for something written by economists! Other studies too have shown broadly similar results. So, at least for an established brand^*, it appears that purchasing online ads is poor investment.

 

Homo economicus and other imaginary creatures

But why then do we continue to see these pesky ads? One explanation proposed by economists is that this has something to do with distorted incentives within firms. In plain English: How else will advertising managers and departments keep their jobs? But to me this seems an unlikely explanation. I find it hard to believe that companies whose sole motive is to make fat profits will allow a bunch of mid-level (advertising) managers to spend wastefully, contrary to all economic wisdom. I offer the more charitable (and my favourite) explanation, that this is all because of the human predilection to draw spurious causal inferences from correlated data. Even CEOs and CFOs, despite their years of training and experience, apparently still remain vulnerable to these foibles. (For more, see Home remedies, hydroxychloroquine, and other illusions (randomramblings2018.blogspot.com)). So the next time you encounter an ad online for a popular company or product, you know that it’s likely just a wasted piece of code. But on the brighter side, know that your decision to click through or not, has the power to make someone else pay for the annoyance. You can choose which one, depending on your mood!

 

 

^*Of course, relatively unknown firms need advertising to publicise their brands, consistent with the informative view of advertising. And firms facing intense competition may choose to continue advertising in a bid not to cede space to rivals. But the value of such “defensive” advertising is unproven.