The ad sidebar is coming back

There are ads on this page, in a column forty pixels wide down the left edge. There are none in the article you are reading. Not between the paragraphs, not after the third one, not sliding up from the bottom when you reach the end. That is a decision, and it cost something to make, so it is worth explaining.

The body was the last thing to fall

Advertising did not start inside the text. It started around it -- banners at the top, columns down the side, the newspaper arrangement carried over more or less intact. Then, over about fifteen years, it moved in.

First a unit after the third paragraph, because that tested well. Then one every few paragraphs on long pieces. Then the sticky one that follows you down. Then the video that expands as it enters the viewport. Then the thing that pushes the whole article down eight hundred pixels while it decides what to show you.

Every one of those was justified by a number, and every one of those numbers was real. In-content units genuinely do out-perform a sidebar on click-through. They are genuinely more viewable. If you are optimizing a page against the metrics the industry collects, the arrow points into the body every single time, and it keeps pointing there until the article is a delivery mechanism for the interruptions.

The metrics were measuring the wrong thing

Here is the problem with that arrow. Viewability and click-through both reward interruption, because an ad that gets in your way is by definition viewable and is more likely to be clicked, sometimes on purpose and sometimes because it moved under your thumb.

What none of them measure is the reader who does not come back.

That is not a sentimental point, it is an accounting one. The cost of an in-body ad is paid in a currency that does not appear on the dashboard: a slightly worse experience, distributed across everyone, forever. You can run that trade for a long time before it shows up in a chart, and by the time it does you have taught an entire audience that your site is unpleasant to read.

Nothing in analytics is labelled "closed the tab and did not think about it again."

So people started paying to have the body left alone

Look at what has actually grown over the last few years. Newsletters, where the whole product is text arriving in one piece. Reader mode, which is a button that means "remove the advertising layout from this page." Reading apps whose single feature is stripping a site back to its words. Subscriptions sold explicitly on "no ads."

Every one of those is a person choosing, and often paying, to get the body back. That is an enormous revealed preference, and the industry response has largely been to treat it as leakage to be recovered rather than as a message.

And then there are the blockers, which are the same preference expressed by people who are not going to pay for it. I wrote about that separately: our ad column was blocked outright by Brave, silently, and the reason was that every signal it gave off matched the pattern of the thing people are trying to escape. Blockers are not really at war with advertising. They are at war with a specific set of formats, and those formats are almost all in the body.

What a sidebar is for

The case for a column is not nostalgia, and it is not that a sidebar is better at getting attention. It is not. It is worse at getting attention, and that is the point.

An ad in a column has a fixed place. It does not move the text. It does not appear halfway through a sentence. It does not reflow the page eight hundred milliseconds after you started reading. It sits where it sat when the page loaded, and it is there for as long as you are, which means it accumulates familiarity rather than spending goodwill.

The old sidebar failed for reasons that were real but specific. It was a 300x250 box competing for attention with the article, usually animated, frequently the loudest thing on the screen. That is not a placement problem, it is a behaviour problem, and it followed the ads into the body when they moved.

What is on this page instead is a rail of small marks -- one advertiser per square, forty pixels each, no animation, no expansion. It reads much more like a directory down the edge of the page than like a billboard. You can ignore it completely, and most people will, and that is a fair trade for never being interrupted.

The structural argument, which matters more than the taste one

There is a second reason to keep ads out of the body, and it is the one I did not expect before building this.

An ad in the body has to be injected into the content. That means the ad system reaches into the article: it needs to find paragraph boundaries, insert markup, and re-flow around itself. The article and the advertising become one document. When that document is served, the ad-shaped parts are structurally indistinguishable from the writing, which is exactly why blockers end up removing chunks of the page and breaking layouts.

A rail is structurally separate. The article is a document; the column is a column. It can be fetched, cached, replaced, counted or removed without touching a single word of the piece. Serving it from our own origin -- which is what makes it visible at all to a reader running Shields -- is straightforward precisely because it is not entangled with the text.

Formats that respect the reader turn out to be easier to build honestly. That is not a coincidence, but I only understood it from the implementation side.

What it costs, honestly

I would rather state the bill than pretend there isn't one.

• A column earns less per impression than an in-body unit. Anyone who tells you otherwise is selling something. The bet is that it earns for longer.
• A phone has no sidebar. There is no forty-pixel column to spare on a 390-pixel screen, and this is the genuinely unsolved part. On this site the rail becomes a horizontal strip under the article on a narrow screen, which is honest but is not as good.
• Banner blindness is real. People will learn to skip the column. They learned to skip the in-body units too, which is why those kept getting more aggressive -- and that escalation is the thing I am trying not to start.

The body is the product

The simplest version of all of this: people came for the article. The article is the only reason the page exists and the only reason anyone will ever come back to it.

Selling advertising by damaging the article is spending capital you did not earn -- borrowing attention from the writing to hand to someone else, and booking it as revenue. It works, right up until the moment it doesn't, and then it doesn't work again for a very long time.

So the ads are in a column. If you want one, it is forty pixels and it will never, under any circumstances, appear in the middle of somebody's paragraph.