The ad blocking paradox
As web ads get more and more closely targeted to the user, more and more users are choosing ad blockers, “Do Not Track” and other privacy technology. According to a survey by the Pew Internet and American Life Project, most US users “are not okay with online targeted advertising.” The industry tells us that ads are getting more personalized and relevant, so why isn’t blocking going down instead? Why aren’t users saying,
There’s a magic machine in a data center that will only show me ads for stuff I really want to buy? Better turn off the ad blocker!
For a long time, web ad blocking was a nerdy Internet hobby. It was easy and effective, but few users did it. I once wrote an ad blocker myself, and got just one other user—another Linux freak, who rewrote it. Today, though, mass media attention to tracking, such as the Wall Street Journal’s What They Know series, is finally helping to make ad blocking mainstream. One startup, ClarityRay, reports that 9.26 percent of all ad impressions on 100 popular sites are being blocked.
The conventional wisdom about online advertising is that increasingly better-targeted ads result in more revenue for the medium. However, Mary Meeker’s Internet Trends 2012 presentation tells a different story. What is going on with with “print”? It consumed seven percent of people’s media time in 2011, but 25% of ad budgets! The trend continues in Internet Trends 2013. Print has 6% of the time and 23% of the money. Even as individual online campaigns get higher response rates because of targeting to the user, the value of the web as an advertising medium is staying unnaturally low relative to time spent.
Web advertising’s blocking and pricing problems have a common source: targeting. Conventional advertising takes some of your time but “pays its way” by revealing information about the advertiser’s intentions. Targeted advertising, however, reveals less information, so it’s less valuable and users have an greater incentive to block it.
What does print have that online doesn’t?
Print advertising has the sweet spot of awesome: it’s easy to place an ad based on content, but hard to track individuals.
Hard to track individuals? How is that a good thing?
We’ve got a few steps to work through here. So let’s start with some Nobel-Prize-winning economics research.
How can markets survive information asymmetry?
Here it is: Akerlof, George A. (1970). “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism”. (There’s a summary on Wikipedia.)
Let’s say you have a car that runs fine for now, but that you know won’t last long because of the time with the attempted oil change and the drain plug you found in a pool of oil in the driveway and you cleaned up all the oil with kitty litter but…anyway, you know bad things about the car that a buyer couldn’t. You’re looking to sell. Meanwhile, a seller of a perfectly good, but indistinguishable, car is also trying to sell, but since he’s competing with you, he can’t charge a price any higher than what you’re willing to accept for your “lemon.” Prices tend to get set by what people will pay for the worst possible car, so the market tends to break down.
If sellers have so much more information about the product being sold than buyers do, how does anybody manage to get any business done at all? Dr. Akerlof wrote later, “Indeed, I soon saw that asymmetric information was potentially an issue in any market where the quality of goods would be difficult to see by anything other than casual inspection. Rather than being a handful of markets, the exception rather than the rule, that seemed to me to include most markets.”
If you don’t like the used car example, the job market for computer programmers is another tough one. How do you find someone who doesn’t just interview well and blitz through programming puzzles, but can actually add value to a complex project?
Anybody who buys or sells anything has to spend a lot of time getting around the asymmetric information problem.
And deceptive sellers learn new tricks too. Buyers and honest sellers get one thing figured out, and the deceptive sellers come up with something else. The CIO signing up for a site license faces the same problem as the tourist buying a “genuine Rolex” from a street vendor.
Why does any buyer trust any seller?
The prophet Jeremiah asked,
“Wherefore doth the way of the wicked prosper? Wherefore are all they happy that deal very treacherously?”
But the question we’re still trying to figure out is: wherefore do the honest prosper? Akerlof writes, “Dishonest dealings tend to drive honest dealings out of the market.” How can a market even exist? How can legit sellers earn anything?
Many institutions, norms, laws, and habits have popped up over thousands of years to deal with this problem.
Let’s require licenses for food vendors, and take away the license of anyone who sells tainted food.
Let’s only buy from sellers who offer a guarantee. Or let’s force all sellers to offer some kind of guarantee.
Let’s protect trademark rights for sellers, to give sellers a mechanism for building reputation and repeat business.
Advertising is another one.
Is advertising rational?
How can advertising help make it possible for honest buyers and sellers to work together?
Davis et. al. ask the question, “Is advertising rational?” and come up with:
“It is not so much the claims made by advertisers that are helpful, but the fact that they are willing to spend extravagant amounts of money on a product that is informative.”
That seems wasteful at first, but let’s work through the logic of it.
Buyers are ignorant about the quality of products from which they are choosing.
Buyers become better informed after buying and using a product.
Sellers have a good idea of product quality, and for what kinds of users the product is a good choice.
Sellers have an incentive to misrepresent products.
“What is needed, therefore, is some means of extracting truthful information from producers a means of distinguishing those producers who genuinely believe their product to be of high quality from those who do not.”
So what is a “screening mechanism” that will separate the sellers who believe their products to be of high quality from the deceptive sellers? The idea is to come up with some activity that is costly enough for low quality sellers that they won’t do it, but still affordable for high quality sellers. Advertising shows that a seller has the money to advertise (which they presumably got from customers, or from investors who thought the product was worth investing in), and believes that the product will earn enough repeat sales to justify the ad spending.
Richard E. Kihlstrom and Michael H. Riordan explained the signaling logic behind advertising in a 1984 paper:
When a firm signals by advertising, it demonstrates to consumers that its production costs and the demand for its product are such that advertising costs can be recovered. In order for advertising to be an effective signal, high-quality firms must be able to recover advertising costs while low-quality firms cannot.
Evidence for the signaling model
Davis et al. predict that “products whose main attributes are either easily ascertained or impossible to verify will be less advertised than those whose main features are ascertainable with more difficulty, and verifiable in the medium term.” They divide goods into four categories:
Search goods: easy to check out at the time of purchase, such as fruit or curtains.
Short-term experience goods: hard to evaluate before buying, but easy to try once, such as canned food.
Long-term experience goods: need more experience to judge the quality of an item, such as hair conditioner or batteries.
“Experience is of little value”: Dishwashers and cars.
Products that fall into the category of long-term experience goods have the highest ratio of advertising to sales. Determining product quality is hard enough that the buyer can’t do it at the store or bringing one purchase home, but it is possible, so the buyer can eventually check the seller’s claims.
The conclusions of the paper suggest a positive role for advertising: the fact that a seller is willing to advertise is a useful piece of information for the buyer.
“Advertising enhances the buying opportunities of consumers by alerting them to products about which they know little, and by signalling to them the seriousness of intent of the producer. It is more about informing them than acting as the persuasive door to door salesman. The salesman will persuade them to buy things they don’t need, but he won’t do that for very long.”
Rory Sutherland, vice-chairman of Ogilvy Group, said,
To a good decision scientist, a consumer preference for buying advertised brands is perfectly rational. The manufacturer knows more about his product than you do, almost by definition. Therefore the expensive act of advertising his own product is a reliable sign of his own confidence in it. It is like a racehorse owner betting heavily on his own horse. Why would it be
rational to disregard valuable information of that kind?
A great paper on the signaling model as it applies to advertising is The Waste in Advertising Is the Part That Works by Tim Ambler and E. Ann Hollier. People are surprisingly good at picking out a signal from perceived advertising expense.
How signaling breaks down
Advertising is one support member in the structure of a market. It bears some of the load of holding up the fragile tent under which honest people can do business with each other. Markets tend to fall apart because of information asymmetries, but advertising can help inform would-be buyers by revealing a seller’s beliefs about a product.
What happens, though, if sellers try to reduce the load that advertising carries, by “efficiently” targeting some users and not others? As a member of the audience, the more likely it is that the ad you’re seeing is custom-targeted to you, the less information the advertiser is able to convey. With good enough targeting, you could be the one poor loser who they’re trying to stick with the last obsolete unit in the warehouse.
Advertising can break down as a signaling method when the medium gets noisy enough. Mark N. Hertzendorf explains, in “I’m not a high-quality firm, but I play one on TV” (RAND Journal of Economics, vol. 24, number 2, summer 1991.) On the subject of the signaling model:
This result, however, is sensitive to the assumption that consumers can perfectly observe the firm’s advertising expenditure. This assumption is somewhat unreasonable in light of the fact that much advertising takes place over various electronic media to which not everyone is ‘tuned in.’
Furthermore, the noise complicates the process of customer inference. This enables a low-quality firm to take advantage of consumer ignorance by partially mimicking the strategy of the high-quality firm. That’s in an environment where the presence of many TV channels makes it harder for the audience to figure out who’s really trying to signal. Noise helps deceptive sellers.
Now what happens when we introduce targeting? Let’s give the low-quality seller the ability to split the audience, without the audience members knowing, into marks and bystanders, with marks receiving the ad at higher probability. In that case, marks can end up receiving a signal that they can’t distinguish from that of a high-quality seller.
It’s not about the data
Rebecca Lieb, at iMediaConnection, points out that her BlueKai profile is largely false, and writes,
If ad platforms aren’t delivering the targeting that advertisers are paying for, the emperor has no clothes. And Adam Tanner at Forbes finds that people’s Acxiom profiles are also full of bizarre errors.
But the emperor is fully clothed, just headed in the wrong direction. Targeting serves its intended purpose whether or not it’s accurate, as long as it can split the audience reasonably consistently. It just has to make a user who falls into a targeted group once fall into the same group again on future visits. Even the most basic cookie scheme will do that. An ad network can randomly call some users left-handed and others right-handed, and get it wrong. The targeting system can even mix up two users with the same name, and it still works. Targeting only has to split the audience persistently, so that some have a higher probability of receiving an inaccurate “high-quality” signal.
The result of splitting the audience, though, is that high-quality sellers lose some rewards of advertising, because they become harder to distinguish from low-quality sellers. Where low-quality sellers in Hertzendorf’s scenario must rely on increasing noise in the medium in order to deceive, targeting lets sellers make the first move.
Advertisers who choose targeting are usually not deliberately deceptive. But by choosing to maximize the response rate of a campaign, they end up sending a deceptive signal, reducing the information that the ad provides to the audience.
Awareness of targeting is still growing
If an individual is aware that targeting is possible but doesn’t know if he or she is mark or bystander, the signal is lost. As users learn about targeting, the value of the entire medium is going down, even for advertisers who do not target.
However, some buyers are still unaware of the extent of targeting. One politician saw an ad for a dating site on a political party press release and attributed it to the party, not to the Google ad service used on the site where he read it. From inside the IT business, a lot of tracking and targeting schemes look old and obvious, but some of the audience is still figuring it out.
And the faster they figure it out, the faster the signaling power of targeted ads is going away.
Targeting failure: the email spam problem
All of this isn’t just hypothetical—it has happened before. We can get some hints about the future of targeted web advertising from the history of email spam. Remember CAN-SPAM? This was the US Federal law that overruled state spam laws, some of which were strict, and cleared the way for advertisers to send all the spam they wanted, as long as they followed a few basic rules. It was a huge victory for pro-spam lobbyists at the Direct Marketing Association.
Today, the data, the tools, and even the law are all there for advertisers to take full advantage of email spam. The CAN-SPAM debate is over. The Internet privacy nerds lost, and database marketing won.
So where is all the CAN-SPAM-compliant spam?
Web ads promise targeting, but spam has been able to deliver it for a long time. Why aren’t advertisers using it?
User targeting promises a way to reach users without paying for expensive content. In Ad Age, Adam Lehman, COO and General Manager at Lotame, writes,
With the enormous variety of information available through the Internet, I am able to do research on running shoes across diverse sources. Based on the interests I express through my research, I may be presented with downstream advertising offers, which I can take or leave.
The key word here is “downstream.” Lehman goes to a running site and somehow expresses interest in shoes. Later, while he’s browsing some other, possibly unrelated, site, an advertiser “retargets” him with a shoe ad. The “downstream” site can be running whatever is the cheapest content that Lehman is willing to look at at all. Instead of having to place ads on relevant content, an advertiser can chase the user onto cheaper and cheaper sites.
What if you took everything that web ad targeting promises, and actually delivered it? Deliver ads to the exact right user? Sure. Go by email addresses which are in marketing databases already. Save money on content? Try free. Take every ad targeting concept and max it out, and you get email spam.
But what’s wrong with that? John Battelle writes,
It’s actually a good thing that we as consumers are waking up to the fact that marketers know a lot about us—because we also know a lot about ourselves, and about what we want. Only when we can exchange value for value will advertising move to a new level, and begin to drive commercial experiences that begin to feel right. That will take an informed public that isn’t “creeped out” or dismissive of marketing, but rather engaged and expectant—soon, we will demand that marketers pay for our attention and our data—by providing us better deals, better experiences, and better service. This can only be done via a marketing ecosystem that leverages data, algorithms, and insight at scale.
As they say on the Internet,
dØØd wtf? The first step in me getting a better deal is for the other side to have more information about me, and for me to be
engaged and expectant about that?
A cold call is not advertising
Email spam is the digital version of direct mail, which is the paper version of a cold call. The problem with that whole kind of communication is that it’s based on extremely fine-grained data on the seller’s side, and none on the buyer’s.
An advertisement that’s tied to content, in a clearly expensive way, sends a signal from the advertiser to the buyer. The extreme example here is an ad in a glossy magazine. It’ll still be on that magazine years later, and every subscriber gets the same one. Almost ideal from a signaling point of view. The other extreme is a cold call, which carries no “proof of work” or signaling value. All the information is on the seller’s side, so the cold call is of no value to the recipient.
Which is why users block email spam. It’s worthless. Even spam that complies with CAN-SPAM is worthless. As web ads steadily move further and further away from magazine-style, with signaling value, toward spam-style, with no signaling value, they’re losing value as well.
Targeting failure: legit sites lose, intermediaries win
Ricardo Bilton asks,
While behavioral advertising may be vital to the current makeup of the web, the question worth answering now is this: Is that really the kind of Internet that we want to use in the first place?
Good question. Is it really a good investment for an advertiser to throw money at intermediaries who know lots of math, so that the advertiser doesn’t have to spend money on content to attach the ads to? According to AOL CEO Tim Armstrong, only 25% - 45% of online ad spending makes it as far as the publisher.
Alexis C. Madrigal writes,
The ad market, on which we all depend, started going haywire. Advertisers didn’t have to buy The Atlantic. They could buy ads on networks that had dropped a cookie on people visiting The Atlantic. They could snatch our audience right out from underneath us. And Michael Tiffany, CEO of advertising security firm White Ops, said,
The fundamental value proposition of these ad tech companies who are de-anonymizing the Internet is,
Why spend big CPMs on branded sites when I can get them on no-name sites?
So where do the ads end up? In a lot of cases, on pirate sites. Chris Castle points out that McDonald’s ads are running on sites that carry infringing song lyrics.
Another example is BMW’s Response to Ads for Its Brands on Pirate Sites. Somehow, BMW advertising ended up running on an unlicensed album download page, on a site called mp3crank. Ad networks do respond to pressure from copyright holders and remove ads from known infringing sites, but snatching revenue from original content sites is still a major effect of tracking.
How targeted ads fail brand advertisers
Brand advertisers, who Doc Searls splits out from direct response advertisers, are developing an understanding of the targeting problem and looking for alternatives. John Hegarty, founder of the ad agency Bartle Bogle Hegarty, said,
I’m not sure I want people to know who I am. I find that slightly Orwellian and I object to it. I don’t want people to know what I drink in the morning and what I drink at night. I think there’s a great problem here - throughout history we have fought for our freedom to be an individual, and you’re taking it away from us. I think there’ll be a huge backlash to that and Nike will have to be very careful.
And Richard Stacy writes,
The great thing about advertising is that no-one takes it personally.
Rory Sutherland of Ogilvy Group also comments,
A very intelligent British adman makes the distinction between ads which create sales and ads which create saleability. The kind of adspend-as-signalling Don refers to is very much about creating saleability rather than sales. Conventional media do a better job of this signalling, which may be complementary to—and not replaceable by—money spent online.
Am I right? Testable predictions
Print advertising will keep its premium in spending per user minute. (The harder it is to do targeting in a medium, the more valuable the medium is.)
Higher awareness of the extent to which online ads are customized to the user will be correlated with higher use of privacy tech. (Users want to see “who’s advertising on example.com”, not ads with “relevance”.)
The more privacy tech that a user runs, the more money he or she spends online, even when controlling for demographics, time spent online, and Internet skill. (Avoiding information asymmetry is an important motivation for using privacy tech.)
Some product and service categories (cars, insurance) are still mostly sold offline. The users of privacy tech will be overrepresented in the early adopter group who are buying these categories online, even when controlling for demographics, time spent online, and Internet skill. (Signaling matters more than matching.)
What next? Solutions
Solution: privacy tech
It’s possible for both of these to be true:
This individual ad will have a higher click-through rate if we personalize it to the user.
Online advertising as a whole will be less profitable if we personalize ads to users.
Advertisers and legit content sites would do better if nobody used creepy tracking on users, but if some advertisers track and others don’t, the ones who do can have an advantage. As long as users believe that online advertisers target ads, any non-targeters won’t get the credibility benefit they deserve.
Statistics about performance of individual ads feeds into dangerous self-deception about the industry-wide effects of targeting. However, as the audience begins to understand targeting, the rate of ad blocking increases, the value of web ads decreases, and increasingly crappy ads drive more user demand for ad blocking.
Fortunately, better privacy tech, such as stricter treatment of third-party cookies, makes targeting more difficult and less accurate. The value of advertising across the entire medium rises, it’s harder for ad networks to “snatch” audience from high-reputation sites, and those sites will be able to get more ad revenue and control, at the expense of middlemen.
Just as targeting didn’t have to work with total accuracy to give an advantage to deceptive signalers, privacy tech doesn’t have to be 100% to push things back in the other direction.
Solution: give brand advertising a seat at the table
The current policy debate on web privacy is going much the same way that the debate on email spam did. Responsible users of email for marketing abandoned the debate and let the lobbyists at the DMA got CAN-SPAM passed. That only helped the bottom-feeders (who probably don’t pay for DMA memberships anyway) and made it a never-ending challenge for legit DMA members to get a legit email newsletter through.
There are a lot of details to work out about how the norms and protocols for online ads have to change to support brand advertising, and not just direct response.
But starting with the assumption that the industry will go broke without unacceptable levels of user tracking will get us to the wrong place. Brand advertisers need to send a credible signal, and privacy tech helps reinforce that signal. The less targetable that web advertising is, the more valuable that brand advertisers will find it.
Peter Klein of MediaWhiz, writing in Ad Age, explains Why Do Not Track Will Make Online Advertising Better (Seriously).
Anti-tracking legislation will make online advertising more focused and relevant to consumers. It will set into motion a more innovative and prosperous era of digital marketing, dominated by a healthy respect for consumers’ wishes about how their data are collected and used, and innovative advertising that meets their needs.
Making creepy tracking harder is just what advertising needs. Klein writes,
Do Not Track will force marketers to be more creative in their campaigns, tapping into legally available data—users’ expressed interests. This will foster deeper and more relevant connections between brands and consumers and benefit online advertisers in the long run.
And better privacy tech isn’t just good for advertising, it’s good for the content creators. When advertisers have to target by interest, they have to look for relevant content, instead of falling down the ad network rat hole and chasing the desired user onto the cheapest possible page. All but the bottom-feeder content sites are likely to do better under an improved privacy regime.
Conclusion: don’t panic
I once worked on a web site for a client who showed me another marketing project at the company. It was an animated Christmas card, delivered by email. The file type of the attachment: Microsoft Windows .EXE. Naturally, I pointed out that (1) In those days, when Windows 95 and 98 were the most common client platforms, the .EXE can basically do anything—read the user’s files, trash other programs, whatever. And (2) Internet email is forgeable. The “From:” address can be totally fake. So who’s going to open a .EXE that comes in via email?
Naturally, I was wrong. “Given a choice between dancing pigs and security, users will pick dancing pigs every time.” The customers loved the Christmas card.
Today, though, we have different norms and technologies around security. A .EXE in email will get quarantined, filtered, or buried under layers of warnings.
The same thing is happening with privacy problems. Browser developers are steadily closing the bugs that make creepy tracking possible. And yes, that makes some advertising techniques obsolete, the same way that corporate virus checkers killed off the animated .EXE Christmas card business.
But if you want to send customers a holiday greeting, you still can. And after the web fixes its privacy bugs, you’ll still be able to advertise. It will just work better.
Don Marti <firstname.lastname@example.org>
17 October 2013: Add Michael Tiffany quotation
10 October 2013: Add Slashdot and Doc Searls Weblog links. Remove a duplicated quotation. Add link to Ambler and Hollier paper.
10 September 2013: Break out It’s not about the data section, add a link.
7 August 2013: Chop out a duplicated paragraph.
27 July 2013: Original version.