blog: Don Marti


When can deceptive sellers outbid honest sellers for ad impressions?

14 April 2018

Update 12 Dec 2021: Add link to London Underground example.

Update 8 Jun 2018: simulation, Why digital advertising leaves people underwhelmed

Why does the Peak Advertising effect occur most in the most accurately targeted ad media? Why do people tend to filter out targeted ads, using habit power, technology, and regulation, while paying more attention to less finely targeted ad media?

One explanation is that buying ad space is an example of costly signaling. On this view, advertising is basically an exchange of signal for attention, and ads that don't pay their way with some kind of proof of spend are not worth paying attention to because they don't convey useful information about the seller's beliefs on how valuable the audience would find the product.

Another possible explanation is that targetable ad media are more suitable for deception, and that where advertisers bid for space in a medium, deceptive advertisers will tend to outbid the honest ones.

This seems counterintuitive, since we might suppose that the customer lifetime value of an honest seller's newly acquired customer could in many cases be greater than the profit from a quick score by a deceptive seller. But targeting doesn't just match ad impressions with prospective buyers. When used by a deceptive seller, it can also conceal an ad impression from potentially costly attention.

For honest direct marketers, the expected profit from reaching a buyer is positive, and the expected profit from reaching a non-buyer is zero. But the audience does not just contain buyers and non-buyers. People can also be divided into enforcers and non-enforcers. Enforcers can be anything from professional law enforcement people, to someone who takes apart a bogus product and makes a video about it, to just the writer of a bad online review. What enforcers have in common is that for a dishonest seller, the expected profit from reaching an enforcer is negative.

Some kinds of enforcer can impose costs even without buying. For example, a reader might send the publisher a screenshot containing a scam ad and get the advertiser added to an advertiser exclusion list. Other kinds of enforcer might only take action if they buy the product and find it to be a scam. A deceptive advertiser might incur costs when their ad is shown to either kind of enforcer.

For an honest advertiser, the expected profit from a single impression is:

probability of reaching a buyer × expected profit per sale

For the dishonest advertiser, the expected profit is:

probability of reaching a buyer × expected profit per sale − probability of reaching an enforcer × expected loss per enforcer

The expected loss per enforcer is typically high compared to the profit per sale. For example, a small number of contacts with review writers might require a seller to re-launch under a new name. In an ad impression market with both honest and deceptive sellers, where sellers can choose which impressions to bid on, an ad impression that a deceptive seller believes is unlikely to reach an enforcer has extra value to that deceptive advertiser but not to an honest advertiser. Deceptive sellers will tend to outbid honest ones for certain impressions.

A member of the audience might be able to see targeting criteria, but not the advertiser's internal weighting of targeting criteria. (For example, a targeted ad platform might reveal to you that you are being targeted for an ad because your computer is running the latest release of the OS. What they won't tell you is that the seller is bidding on impressions to your OS version because they're selling a tainted nutritional supplement, and the lead testing department at the Ministry of Health is still on the old OS version.)

So, some ad impressions will tend to be purchased by deceptive sellers, but a low-information member of the audience can't tell which impressions those are. Is this an ad from an honest seller that might be reaching both me and enforcers, or is this an ad from a dishonest seller targeted to reach me but not enforcers? When you read a magazine that reaches a community of practice of which you're a member, you can be confident that product reviewers and editors are seeing the same ads you are. A web ad could be targeted to avoid experienced and better-connected members of the community of practice.

A good example of where lower targeting capabilities can assist norms enforcement is the 378 commmuters who reported London Underground ads that were later banned by the Advertising Standards Authority. Silvia Milano, Brent Mittelstadt, and Sandra Wachter write,

[I]f the ad could have been targeted to just the subset of consumers most receptive to its message, they might not have raised any complaints. As a result, the harmful message would have gone unchallenged, missing an opportunity for the regulator to update their guidelines in keeping with current social norms.

One possible explanation for the Peak Advertising effect is the interaction between deceptive sellers discovering how to use a new ad medium's targeting capabilities to avoid enforcers, and the audience discovering the fraction of deceptive sellers.

Related: Ban Targeted Advertising by David Dayen in The New Republic. (I'm not so much interested in whether or not targeted advertising should be banned as I am in the reasoning behind why people choose to protect themselves from it. The story of matching the exact right buyer to the exact right product is much less compelling for most purchase decisions than the buyer's story of finding an adequate product and avoiding deceptive sellers.)