Craigslist Has No Ads?

In 2003 when social networking was all the rage in Silicon Valley, you could hardly find a hotel meeting room with a capacity over 100 people that wasn’t holding a panel discussion on social networking startups. At one such panel, Mark Pincus, founder of, spoke about a focus group his company had done on users of various social Web sites and their attitudes toward advertising.

Participants said they didn’t like ads. In particular they said craigslist was one of their favorite sites because it had no advertising.

Ponder that one for a moment. Craigslist is 100 percent ads! But due to the site’s nature, users typically only see ads they’re looking for, or at least ads that are in the same category that they don’t mind (and likely even enjoy) hunting through.

I don’t know if craigslist ever ran the experiment, but I’d bet good money that if it replaced some of the screen real estate with free credit report offers or diet offers showing flabby bellies, the focus group sentiment would have been different.

Relevant Ads Perceived as Content, Even by Top CEOs

At a recent annual shareholders meeting, CEO Jeff Bezos made himself popular with the Madison Avenue crowd by opining that advertising “is the price you pay for having an unremarkable product or service.” This is an amazing position to take. Amazon’s recommendation system, arguably a significant driver in its 1990s’ success in eclipsing other online e-tailers, is essentially a sophisticated behavioral and contextual advertising engine, using the Web and e-mail as its primary channels. Bezos just doesn’t think of it as advertising because his users don’t think of it as advertising; it just seems like a useful service that consumes some of the user’s attention to try to connect people with products they might want to buy. Isn’t that exactly what advertising is?

Solving Attention Pollution

When (which later became Overture, then Yahoo Search Marketing) started, it was with the observation that a consumer’s attention is a valuable commodity. If it were openly auctioned off to the highest bidder, the result would likely be efficient, meaning that if advertisers were rational, the highest bidder would in theory be the company with the most relevant offer for that user.

Fast-forward to today. The major search engines no longer just auction attention to the highest bidder. They estimate the probability that a user will respond to an ad, based primarily on the search terms the user just typed in, and multiply the price the advertiser is willing to pay by the probability of a click to determine the expected value of showing each ad.

In this way, a site visitor’s attention is consumed by a tradeoff between the degree to which the person wants to see the ad (estimated using statistics and models) and the degree to which the advertiser wants to show its ad to the person (estimated by the price it’s willing to pay for a click). From a user-centric view, one could argue that results should be ranked entirely by the user’s interest as measured by likelihood to click. But if two ads have the same chance of getting clicked and one advertiser is willing to pay double, the argument goes, that advertiser must have a more valuable service or is rationally betting it can deliver more value to the user. For example, the higher-paying advertiser might have a post-click conversion rate three times higher than the cheaper advertiser’s, supporting the theory that its products were more interesting than its competitors.

Because of these technological and marketplace design decisions, you often find people saying they find search engine ads valuable, as valuable as the search results themselves.

Display Ads in the Dark Ages

In terms of attention consumption, if search ads are a modern Toyota Prius, most display ads are the worst cars produced by Detroit in the ’60s and ’70s.

Search ads are selected based on an attempt to deliver the ads most interesting to the user. In contrast, display ads are generally delivered in an attempt to make as much money as possible for the publisher or ad network. Frankly, the ads you see on a page are more often the result of a media buyer purchasing a swath of online ad inventory because doing so has worked before, which might mean a conversion rate of 1 in 100,000. This is like going fishing with a net and catching 1,000 dolphins for each tuna. There’s just no penalty for catching dolphins (uninterested users) in online advertising…or is there?

In a world where ad blockers are among the most popular Firefox extensions and the Federal Trade Commission is looking sternly at online advertising practices, it seems like the display ad marketplace needs to evolve, and quickly.

The Age of Optimization

The only solution is to serve display ads intelligently. That means analyzing information on the user’s interests, historical response to advertising, and other signals predictive of future ad response; the page the ad is being served on, historically the kinds of ads that have generated response; and any other relevant information. The search guys have it easy; we have to work harder.

George is off today. This column was originally published June 10, 2009 on ClickZ.

Related reading