Brett Gordon: Most people don’t know what they’re going to do tomorrow! So it’s very hard for marketers to make such accurate predictions.

There’s simply a lot of randomness in what you do. Suppose you see an ad today for a new product you’ve never heard of, it sounds like a great product, and then maybe you check it out online. But just before you consider buying it, life intervenes: your baby starts crying, you need to make dinner, you're out of paper towels, and so on. Even when an ad is a great match for someone, it doesn't necessarily do anything, because everything else is just more important. Advertising relies on that one-in-a-thousand chance where the ad lands and where nothing else gets in the way of a purchase.

Johnson: I have work looking at the effectiveness of retail advertising for an apparel retailer. We have two years of sales data from that retailer, we have detailed data about people's ad exposure on the online platform, we know those people’s demographics—and we are only able to predict 10% of the retailer’s sales over the course of two weeks.

Johnson: About a dollar for a thousand ads, or 0.1 cents per ad, is a reasonable price for banner ads. So you shouldn’t expect that ads are going to cause earth-shattering changes in behavior. But, I can tell you from my work examining 432 display ad experiments at Google that the weight of evidence is that overall ads work.

Plus, clicking on the ad isn’t necessary to actually be affected by the ad.

Gordon: That’s right. In one advertising campaign that we observed during our work with Facebook, around 75% of all conversions—instances where people saw the ad and made a purchase within about 30 days—never clicked on the ad.

To put online ads into perspective, a product’s price is a much bigger deal compared to advertising. I like to think of pricing as the hammer of marketing and advertising as the little feather that tickles you into maybe making a purchase or not.

Gordon: This topic recently received a great deal of attention after the scandal surrounding Cambridge Analytica, where the firm was able to obtain a lot of data on Facebook users and link those data to psychographic profiles compiled from personality tests these users had taken.

An important question people should ask is not how well psychographically targeted ads work compared to no targeting, but whether targeting using psychographic information does a better job compared to the existing abilities of advertising platforms to target you.

After all, many advertising platforms have a great deal of information about their users. So, if you already told me that this particular user or cookie visited my site 7 days ago, and that the user is likely male between the ages of 18 to 35, I have a pretty good idea of what kinds of ads to show them. Does telling me something about their psychographic profile change the kind of ad that I would want to show them? It's not clear that the psychographic information really adds much.

Johnson: Cambridge Analytica was trying to upsell what they're doing, so they're basically creating this notion that they can get inside your brain and know exactly what you're thinking and target you based on your psychology. But there really is no evidence to suggest that psychographic targeting is more effective than any other form of targeting. In fact, the most prominent study in support of psychographic targeting used an experiment on Facebook to show effectiveness. But as we've written in a letter published in the Proceedings of the National Academy of the Sciences, the evidence is not quite as strong as it appears. Even Aleksandr Kogan, the research psychologist who sold the psychographic data to Cambridge Analytica, recently called psychographics a “dead end” for marketers.

Johnson: Ad platforms are becoming increasingly sophisticated in matching consumers to advertisers by making inferences about what consumers like. So it's actually in their interest to serve you the most related ads.

And you can see that progression: the early days of the Internet were a bunch of yucky teeth-whitening ads, but now we get ads that are very specific to the life phase that you're in. So if you're pregnant, you're going to be getting ads that are relevant to being pregnant and expecting a child to arrive, and as you're retiring, you're going to get ads that are relevant to that and to the stores where you shop.

Gordon:  I'm sure that some people would be willing to pay to keep their data totally private, but my understanding is that surveys suggest that there are not that many people like that.

Johnson: The reality is that most people are very happy to get Facebook for free, and would not pay for it if it was offered for a price. They may complain about the advertising. But that advertising does, in fact, support a service that they get a lot of value from.

Johnson: In the United States, people are tracked until they say “no.” If you look at a banner ad and you see a little triangle icon in the corner of the ad, you can click on that “AdChoices” icon to learn about online behavioral advertising and opt out of it if you choose. That’s an extraordinary offering. But, I was the first to collect data on actual opt-out rates, and only 0.2% of ad impressions in the United States are from people who have opted out of tracking.

Now, I think most people are unaware that they can opt out. But, I think you have to keep in mind that awareness itself is a choice: if people really cared enough about tracking, they could find out.

With the GDPR (General Data Protection Regulation) in Europe, I expect publishers will eventually be required to get their users’ explicit opt-in consent for online behavioral advertising. This will spell the end of these tactics as almost no one will say “yes” to being tracked online.

To date, most European websites are resisting this strict opt-in approach because, of course, opting out is not a costless trade-off. If you look at people who opt out of online behavioral advertising, they fetch something like 65% lower prices from advertisers.