The site uses cookies that you may not want. Continued use means acceptance. For more information see our privacy policy.

Advertising and Blocking

Subscribing versus micropayments versus advertising.

Journalists online often make their money from advertising. And there have been a number of stories over the years about the damage of ad blocking. Seldom (never?) do you see the exact amount they make per view or per impression or however you want to measure how much money they make from me looking at their articles.

I read a bit of stuff from Ars Technica, and a rough count for the past month is at least looking at (which is admittedly different than reading) about 70 different URLs (some might be little more than photo galleries, so they aren’t all necessarily articles). They have two subscription options that would apparently supplant any money they would otherwise make off my visits if I saw their ads:

  1. $50 per year, which is about $4.17 per month.
  2. $5 per month.

Assuming the past month is my average viewing of their site, it would work out to between $0.0595 and $0.071 per URL, if I subscribed. The problem is that I don’t believe they make anything like ¢5 or ¢7 per view from advertising. If that were the model, that I actually paid them a nickel per URL, I would probably read less, at least until I got a handle on how much it was costing.

Loyalty is also a hard argument to make on the web, when I find my reading habits shift over time. If I have to go to each content site and manage subscriptions, figure out whether I want to give $5 to Ars Technica or $5 to some other site, that’s time I’d rather spend doing something else. Anything else.

But if they told me they make $0.007 per view (my completely blind guess), I might gladly pay $0.01 per URL (on average), especially if it were a cross-site system, and I doubt I read more than 100 articles per day. That is the micropayment model.

In other words, if sites like Ars Technica or Business Insider want to shift the model, they should get together and try to build the system that lets that happen. Because right now there’s a lot of difference to me between $0.49/month (the $0.007 ad rate), $0.71/month (if I paid a cent per article), and $5.00/month (if I paid their subscription rate).

Now, I’m sure their argument is that not everybody who reads will subscribe, so the $5 is designed to subsidize the freeloaders and still be low enough that people with modest incomes can afford it. The counter to that is that there are many articles I do not read from them. I would prefer to pay for the ones I actually read, to encourage the journalism I care about.

It’s the difference between paying for access to all the music from some record label and paying for specific albums by specific artists.

On the other hand, I do have my Ars whitelisted in my ad blocker. Edit 20150910: I have removed the whitelisting, as Ars Technica uses moving ads and Adobe Flash ads.

When Ads will Target Computers

Once computers begin buying and selling by themselves, they will need advertising.

Mug of robots asking, "Can I borrow a cup of robots?"
Lightly modified; original by hobvias sudoneighm (Flickr: striatic)

With the rise of the cloud, and the expected future of autonomous systems, we will start to need advertising to computers. That is, a system may want to buy itself increased capability, or buy its humans goods or services based on calculated needs.

There is an important question of what computer-targeted advertising would look like. It seems entirely plausible that the advertising industry is not equipped to deliver compelling advertisements for computers. Traditional advertisements rely heavily on appeals to emotion and cultural triggers.

Computer Ads for Computer Needs

So the first type of advertising for computers is selling them things they need. More storage space, software upgrades, that sort of thing. Computers will likely only want to know the specifications of their potential purposes, eliminating the need for stylized advertising and flowery language.

They may want more data than many vendors currently deliver. Computers will want to more heavily study production quality and vendor reputation, at least for components or upgrades that are critical to their continued operation (e.g., for a cloud-based backup the computer may want to know about dependability, but also lock-in costs more than a human would).

The other factor here is that as computers do begin to engage in commerce, it is expected that services themselves will begin to cater to computers. Instead of building products that appeal to IT managers or individual humans, the products will be designed to fit the computer’s needs. The marketing schtick will fall by the wayside, as will value-added items like training services.

Finally, computers will also want to advertise themselves to other computers. If they have extra resources, they will want to use them for offsetting their own costs, for example. Or they might do volunteer work, like testing software they use for bugs, in their spare time.

Computer Ads for Fulfillment of Human Needs

The second type of advertisement for computers involves telling computers about products and services that they will buy on behalf of humans. This sort of advertising will be much harder to nail down. Will a human tell their computer they want edgy or fashionable items when available? Will the computer recognize when the products offered to it are being skewed based on data mining?

It seems plausible that computers will gain some ability to seek out occasional alternatives to consumable products, to give humans the opportunity to try to switch between alternatives. Marketing materials will likely shift to try to trigger that mechanism more frequently.

If you are buying Foo™ Food and only trying alternatives every six months or when the cost of an alternative saves you 10% or more, Bar™ Food might try to figure out your computer’s alternatives cadence, and then hit it with a 10% discount the next month, to see if a double-period of the alternative convinces you to switch brands. Of course, other factors besides an alternative schedule and price discount may be used to determine computed purchasing decisions.

The other side of that is that computers may find novel ways to avoid price discrimination. This might take the form of pre-shipment secondary markets where one computer buys an item and resells it to another at a slight markup but relative discount to the advertised price.

Ad-blocking Computers?

Would computers see any value in blocking advertisements? Given they do not have the same attention deficits as humans, it is unlikely. If other constraints compel it, they will, but otherwise computers seem like great candidates for honest advertising.

Against Nonpersistant Web Advertising

Long story short, some advertisers don’t understand the web. Even big ones.

I have a lot of complaints against advertisements in general, which I will not be discussing in this post. Instead, this post is dedicated to one of the biggest problems with web advertising: not being able to see the advertisement again.

Take YouTube as an example; if you visit a YouTube video, they apparently invoke some Pseudo-Random Number Generator to determine both whether to show an advertisement, whether it should be video versus popover, and whether it should bear any relation to the video.

Worse, though, is that you can’t replicate the PRNG; they don’t throw you the seed they use or anything. So if you start to watch a video, and decide later you actually want to click that advertisement, you can’t. If you stop watching a video advertisement a few seconds in, closing the window, but then process the snippet you saw and decide you want to watch it, you can’t.

This isn’t unique to YouTube. All over the web there’s different advertisements with each pageload, and you can’t get back to them. Some sites like Reddit are a little better in that they do have persistent addresses for at least some of the advertisements. But you still can’t see the advertisements you’ve seen.

You can’t share these advertisements. They’re like ghosts. They’re black ops. Appearing for a mere moment, just long enough to draw the knife blade across your throat. Okay, maybe that’s too dramatic. They’re more like insects that buzz up by your ear just long enough to make you wiggle a little.

“But!,” I hear the marketadvertbrandologists say, “Advertising is not about selling products, it’s about…” and there they go trying to sell advertising as being the equivalent to your grandmother’s cake or grandfather’s pie. That they want you to feel like a kid in a candystore when you think about your vitamin supplements, carrying around a brown paper sack and scooping the goodies into it.

The main thing is, even that excuse doesn’t work if people want to spread the advertising message. If Smith says, “Jones, did you see the really cool video advertisement for genetically modified hermit crabs?” Jones says, “No, I did not.” Smith says, “Well, that sucks, because I can’t give you a link to it.”

Just one more reason to dislike advertising.