Danny Yee >> Web Design >> Publicity

An Experiment in Micro-Advertising

I'm not a real fan of advertising - and I browse the web with image-loading and Javascript disabled, so I don't see that much of it. But I was curious about how (if) online advertising works.

So in May 2001 I carried out an experiment in micro-advertising, on two different sites - Robot Wisdom, my favourite web log (and an example of a quality hobby site), and Google, my favourite search engine. Both schemes involved simple text-only ads, clearly marked as paid ads (the kind of advertising that doesn't annoy me). The sum involved in both cases was small - US$20 for a week on Robot Wisdom (about 30 000 impressions), and US$10 for 667 impressions on Google - hence the term "micro-advertising".


Dealing with Google is simple - the ease of use of their search engine is matched by the straighforwardness of their AdWords system. [Even if you aren't interested in advertising you might like to play with this - you can see rough figures for how many Google searches there are for particular sets of words.]

Dealing with Jorn Barger, who runs Robot Wisdom, was rather more complex. Rather than a straightforward "pay for ad" setup, he had a scheme where he offers you advice on your web site, since he doesn't want to link to anything difficult to use or frustrating. Some will find that annoying, but it is (or was) a damn good deal - for $20 ($100 for commercial sites) you get really useful comments on the usability and effectiveness of your site. (Of course it helped that my approach to web design is similar to Jorn's - I don't know how useful his comments on an all-Flash bells-and-whistles site would be.)

The ads

Sponsored Links
Disposable People
A passionate but scholarly study of
modern slavery (by Kevin Bales).

The ad I ran on Google is to the right.

With AdWords, I could choose the keywords. I opted to have the ad appear to people who carried out searches containing both the word "modern" and the word "slavery". (I chose those search terms because they fit the book/review near-perfectly, and that review because it helps to raise awareness of an important issue.) The ad appeared to the right of the search results.

The Robot Wisdom ad (with wording chosen by Jorn) looked like this:

Paid link: 563 lively book reviews on all subjects (Danny Yee)

This appeared above the "Headlines" and was untargetted (it appeared to all visitors to the page).


Not counting my own tests, there were six clickthroughs from the Google ad - three within the first forty impressions, nothing for around 500 impressions, and then another three in the last hundred. (I have no explanation for the unusual distribution - the whole campaign lasted around two and a half weeks, so it's not a simple "day of the week" effect.)

The Robot Wisdom ad generated around 75 clickthroughs in a little over two days. Unfortunately at that point the provider hosting the site changed to a "no ads" policy, so Jorn moved the link into the main web log where it ran just like any other item (without the "Paid link" tag) - and that generated over 400 clickthroughs in the following week.

None of the six Google clickthroughs resulted in more than the single page view, while users referred from Robot Wisdom averaged more than six page views. (That's not unexpected given the specific anchor used in the first case and the general one in the second.)


The Google AdWords setup doesn't seem to work that well. Yes, my ad could certainly be improved on, but surely not enough to make that much of a difference. For comparison, during the period the ad was running Google referred more than fifty people to that same review through normal search results (for searches on "Kevin Bales" and "Disposable People"). But I still see AdWords ads while using Google, so some people must have more success with them.

One response would be to make the ads more prominent, obviously - and the way too small font should definitely be changed. The more prominent "sponsored links" that appear immediately above Google search results are presumably more effective.

The Robot Wisdom ad was clearly much more effective - it would probably have generated more than 250 clickthroughs if it had run as an ad for the whole week. Though there may be some novelty effect, with Robot Wisdom regulars clicking through the ad just to see what it was all about, this should be more than enough to be interesting to real advertisers. (Jorn plans to resume running paid links at some point.) The most obvious candidates here are sites with similar content to those Robot Wisdom links to regularly... but possibly not the exact same sites.

A real problem in both cases is preserving independence and integrity. Advertising on both sites would be hugely successful if it wasn't marked as such - just inserted into the Google search results or the Robot Wisdom log - but that would compromise them as resources and alienate their users.

Another problem Google and Robot Wisdom both face is that they are too well designed and consistent. Google users know where to find their search results, and that they are pretty reliable, so there's little incentive to look around and notice the ads. Similarly, regular Robot Wisdom users head straight for the latest headlines, bypassing the "guff" at the top of the page. But I'd hate to see either Robot Wisdom or Google damage their functionality in order to improve the effectiveness of advertising.

I dearly hope that Google finds a revenue model that allows it to keep running the way it has been: it plays a key role in levelling the playing field and helping smaller independent sites attract visitors. And I'd like to think Robot Wisdom (and other quality hobby sites) could produce at least enough income to pay for their operation costs. But I know of no easy solutions here.


Slashdotted - check out the comments there.

One reader pointed me at:

Rusty pointed out that the effective clickthrough rate is almost exactly the same for the two ads - 0.9%. Robot Wisdom is just a lot cheaper. And of course Google is simpler and scales better (dealing with dozens of small hobby sites could be really messy), so they can justify a premium.

A couple of people mentioned that they'd been using Google regularly but hadn't even noticed the AdWords ads...

Update - Kuro5hin

Book Reviews
history, literature, computing, biology, sf/fantasy, etc.
In April 2002 I ran the adjacent ad on Kuro5hin (4000 impressions for $12).

The clickthrough rate was again 0.9%. Kuro5hin advertising is untargeted, but if you're advertising something likely to attract its readers (or of fairly general interest) then it's probably better value for money than Google. And the facility for users to add comments to the ad is fascinating - some ads spark quite lively discussions!

Kuro5hin is one of the "cutting edge" online communities, so it will be interesting to see how its novel approach to advertising works out.

Last modified: April 2002

Publicity << Web Design << Danny Yee