At its Worldwide Developers Conference on June 5, Apple announced that one of the tentpole features of macOS High Sierra will be anti-ad tracking technology:

Intelligent Tracking Prevention in Safari uses machine learning to identify and remove the tracking data that advertisers employ to follow users’ web activity.

At first glance, this may seem to be bad for Google and other online advertisers. However, that perception is mistaken.

Before I explain why, let me talk a little bit about how non-internet advertising works. Suppose I want to put an ad on a billboard on the side of the highway. First I would work out approximately how many people drive by the billboard in an average month and their demographics, then I would purchase the space for a predetermined period of time based on that guess. Suppose I want to show a commercial during a television broadcast on air or on cable. First I would guess how many viewers will tune into a show based on past ratings, then buy an ad for a certain commercial break segment of a show guessing approximately how many people will see it. Suppose I want to buy an ad in a newspaper or a magazine: again, I can only guess what the circulation of the edition or issue will be and rates will be loosely keyed off of that. In each of these cases, the advertiser might get lucky and have the ad seen more than expected due to an unexpected surge of ’traffic’ or the advertiser might be unlucky and the ad might be seen less often.

In general, we can think of there as being two broad categories of ads. Some ads are about introducing a new product category to consumers, and others are about elevating one product over another in an established category. The first kind of ad is non-zero; that is, by spending money to increase awareness of a category, both I and my competitor might have increased sales. The second kind of ad is zero-sum, because if I spend more money on such an ad, it takes customers away from my competitors without growing the market. As markets mature, they generally shift from advertising of the first kind to the second. In the first kind, the return on investment for an ad is positive, and there is no reason not to spend as much as possible given current financial obligations increasing that return. In the second kind, ROI begins as positive but quickly levels off and can even become negative. Worse, your need to advertise is controlled by your competitors. The more your competitor advertises, the worse for your business, and vice versa. It’s a prisoner’s dilemma in which all competitors would be better off if no one advertised, but if Tide advertises, it forces Purex to also advertise to keep from being wiped out.

Because companies only advertise if they believe the ROI will be positive, the total size of advertising market is always going to be a fraction of the size of the rest of the economy. In addition, advertising is a highly procyclical industry. When there is an economic boom, companies will invest more in advertising to try to get an advantage against rivals, and when their is an economic bust, companies will reduce their advertising to the degree possible given that their rivals are also reducing advertising.

From all of this, it should be clear that the idea that internet tracking is good for the advertising industry is ludicrous. It is in the interest of companies that buy ads to target their ads to only the highest ROI individuals. It is in the interest of companies that sell ads to prevent this from happening and force companies to pay for more ad views than they “need”. In most advertising segments this occurs naturally because it is impossible to tailor ads to individual views. A billboard cannot change based on who is driving by it. Internet banner ads, however, can be tailored to the individual. The results of this are well known: ubiquitous ads for things that you already bought last week.

For example, I enjoy writing with a fountain pen, so I am often stalked by ads for a pen that I already bought. In a world without tracking, pen advertisers would spending their money on sponsoring blog posts and YouTube reviews rather than just showing me more of what I’ve already decided to buy, and this would benefit writers and consumers much more than the current arrangement, which only benefits ad buyers by increasing their ROI and lowering their total spending.

Let me talk about another kind of ubiquitous ad. If you’ve spent any time reading on the internet, you have been grossed out by an ad from Taboola (“Content You May Like”) or OutBrain. The reason that publishers allowed such gross ads on their sites is that Taboola would pay in advance for a reserved space instead of paying by the impression.

The shocking thing to me is not that there are scam ads on the internet about punching the monkey and one weird trick. In comic books and in the back of Popular Science, there were scam ads for seamonkeys and X-Ray glasses. What is shocking is the scale at which these ads are allowed to proliferate. Every major website allows these garbage ads to take up premium space on their sites, instead of just being hidden off on the equivalent of the backpages.

Advertising on the internet does not have to be terrible. There is no reason that it can’t be like advertising in newspapers: sold in advance and vetted by editors to ensure that it meets the the tone of the publication. To the degree that anti-tracking technology takes away the temptation to last second auctions of impressions, it will be good for the advertising industry.

(Sidebar on The Deck, which sold ads in a non-invasive way: When they shut down, the post-mortem explained that “In 2014, display advertisers started concentrating on large, walled, social networks.” The Deck model of showing ads without tracking was a good one, but as more cost-effective social network advertising became available, advertisers switched away from banner ads, dooming them. Adding more tracking didn’t grow the size of the advertising market: it shrank the market. The solution is a marketwide ban on tracking, not a larger race to the bottom.)

One last medium of advertising is worth mentioning: the podcast advertisement. So far, podcast ads have been like broadcast ads in that they are purchased in advance based on estimated numbers. One difference is that unlike broadcast ads (and like magazines), “reruns” of podcasts don’t get new ads inserted, instead the original ads are included as is. Some people in the podcast industry believe that they could negotiate higher rates if only they had better analytics. This may be why Apple is reportedly adding analytics to the next version of the Podcasts app. I believe this is a mistake. Learning exactly how many people skip over ads in podcasts will not cause rates to go up enough to offset decreases in purchases. As mentioned, industry rates are a function of ROI, and ROI is tightly bound to the nature of the product and its competitors. Adding analytics can only chase out money from the system that is being “wasted” on inefficient purchases, it cannot add new money to the system. It is for the good of the podcasting industry that they had better hope that Apple does not actually ship this product and thereby make podcasting as a whole worse off.