Web publishing would appear to have hit an economic deadlock, one that can -- and should -- be broken quickly.

Readers say animated banner ads on their favorite Web sites are annoying. At the same time, advertisers complain that banner ads no longer draw as strong a response as they once did, and have been working on making online ads more intrusive than they have been in the past in an attempt to get more attention. But will making banner ads more intrusive necessarily make them more effective? According to The Open Source Development Network, static (non-animated) or even all-text ad banners may actually generate better responses than animated ones, at least on text-rich news sites.

OSDN publishes Slashdot, freshmeat, NewsForge, and other sites carrying news and information about open source and free software.

OSDN is almost entirely ad-supported. We have suffered, like every other ad-supported online publisher, from the drop in banner ad sales and rates between mid-2000 and mid-2001. While our easily-defined 'core' audience of leading-edge, high-income programmers and systems administrators has insulated us from the worst of this crisis, our refusal to inflict pop-up and pop-under ads, or banners that contain Java applets or streaming video on our readers has upset more than a few advertisers. Some simply won't do business with us because we have such a restricted ad acceptance policy. Others offer us lower rates than they would if we allowed them to use advertising methods and technologies we feel our readers won't accept.

We need advertising revenue as badly as anyone else in the online news business who doesn't rely on subscriptions or grants as their primary means of support, so we decided to experiment with ways to make banner ads on our sites more effective for our advertisers without annoying our readers any more than absolutely necessary. This article grew out of our first round of studies on how the content of banner ads affects response rates.

In-House Promos as a Test Vehicle

Like most multi-site online networks, OSDN uses a significant percentaqe of its unsold ad inventory for cross-site promotion. We decided to test our initial banner improvement ideas in a promo campaign for NewsForge, a site we launched in August 2000, as the 'online newspaper of record' for the open source community. Our ad test began May 15, 2001, with two 'matched pairs' of banner ads.

The purpose of test pair A was to persuade webmasters to put the NewsForge headline feed on their sites. This was a specialized campaign we expected to draw a low click-through rate, probably less than 0.1%. Ad 1A delivered its pitch as 'pure' copy, in a sans-serif font that showed up as approximately 12 points in our test browser. It read: Put NEWSFORGE headlines on your Web site! We give you a comprehensive Linux and Open Source feed, 100% free, no strings attached, in your choice of four popular formats. Ad 2A was a four-frame animated GIF, with the following copy: Frame 1: NEWSFORGE headlines on your web site Frame 2: Comprehensive Linux and Open Source feed! Frame 3: 100% free, no strings attached Frame 4: 'NewsForge' logo The two ads used the same template and colors. Ad A2, the animated ad, used the same font as the static one, but sized at approximately 24 points in our test browser.

At the end of the first week, the animated ad had drawn a 0.08% clickthrough rate, while the static one showed 0.13%.

Test pair B, another static/animated pair, contained the following 'base' copy: NewsForge has more Linux and Open Source news than Slashdot and Linux Today combined. (Daily color comics, too.) In this pair, our first week clickthrough difference was even more pronounced. Ad 1B drew 0.55%; Animated Ad 2B drew only 0.24%.

Our Readers Expect Text

Several OSDN in-house programmers, who claimed they had learned to mentally block all animated ads while reading our sites, expressed doubt that static banner ads containing extensive text in a single ad frame -- in a tiny (by banner ad standards) font -- would be noticed at all by our readers. They were surprised when the text-heavy static ads outdrew the animated ones so significantly in our test, but they shouldn't have been.

Readers come to Slashdot and freshmeat -- the two sites on which we displayed the test banner pairs -- to read text, and the static, small-font, text-rich banners, while obviously ads and not editorial content, were more in character with these sites than any possible combination of blinking, whirling, clever graphics.

We have not yet set up our sites to run the new IAB standard large rectangular ads, but when we do we hope our advertisers will use these large spaces to display text-rich, informative messages instead of the intrusive, moving/blinking, low-text 'big' ads we have seen on Yahoo!, CNet, and other sites that are already accepting these bigger (though arguably not better) in-line banners.

The editorial department at OSDN, using nothing but static text headlines and teaser paragraphs, routinely produces click-through rates to our stories that range from 25% to nearly 100%, while animated ads on the same sites' main pages, running in the 'premium' position at the top of the page, get an average click-through rate closer to 0.25%.

Is it possible that we, the editorial staff, are better than our advertisers at getting readers' attention? Of course! If we weren't, we'd all get fired. But we have never had a single OSDN advertiser ask us for advice, and online journalism e-mail lists never seem to get posts from advertisers asking why online editorial content attracts attention so easily while online advertising is so roundly ignored.

Perhaps, with all the recent layoffs among online content producers, a few talented writers and graphics people from the editorial side will drift into the advertisng and marketing side of things and start making ads readers like instead of using online advertising methods that are sure to irritate as many readers as they attract. And one message I hope they take with them is that running the same ads over and over is one of the worst things an advertiser can do to an online audience.

Banner Ads Age Rapidly

A by-product of the study described above was a little insight into the phenomenon of ad aging. Take a look at these click-through statistics for Ad 1B, broken down by date: 5/15/01 - 5/21/01: 70,902 impressions, 0.55% click-through 5/22/01 - 5/28/01: 326,884 impressions, 0.43% click-through 5/29/01 - 6/04/01: 354,730 impressions, 0.36% click-through 6/05/01 - 6/11/01: 443,669 impressions, 0.32% click-through Since mid-June, the click-through rate for this ad has remained essentially constant, ranging from 0.30% to 0.33%. But note the dramatic dropoff in the first two weeks. We have seen a similar pattern in all banners ads on OSDN, no matter what products or services they promote.

A new, significantly different ad for the same product or service shoots the response rate right back up, every time. And yet, despite the obviously rapid 'burnout' factor in online banner advertising, we see many advertisers running the same ad, often across many sites, for months at a time.

Gathering the Right Numbers

So far, most numerical data gathered about online advertising has been used for ad targeting. Targeting is obviously not the Holy Grail of Web marketing; if it were, we wouldn't hear advertisers crying about low response rates and demanding lower costs per thousand.

Targeting is only one part of an effective ad campaign. Online advertisers need to start focusing more on the ads themselves. They need to spend some time and money learning what kinds of ads work best on different sites, with different audiences -- and why -- and they need to apply those lessons in a practical way, by making online ads that readers want to see instead of making ads that most Web users consider an annoyance.