Keyword-based Ads, Contextual Analysis, and RSS ad-splicing
Qumana has a series of posts that report on the result of a survey they’ve run across their users about advertising on blogs. Two Interesting questions they are asking:
1. Which is more effective: contextual ads or keyword-based ads?
2. Do add in RSS feeds work, or are ads only effective on the blog pages?
Keyword-based ads are ads that are matched against author-supplied tags for the content. Contextual ads are the same, only software analyzers are used to automatically deduce the contextual metadata. All things being equal, a human will provide more authoritative contextual information than a crawler every time, right? Probably, but there are a couple things to keep in mind:
a) In some cases, machines using Bayesian filtering techniques (and other heuristics) may be better at matching your content with the most effective ads than you are, even if you are the author. If you write a blog on say, fly fishing, you might naturally offer the following tags for keyword-based ad matching for your latest post: “fishing”, “Blackfoot River”, “Montana”, and “lodge”. But a sophisticated contextual analyzer might catch something you missed. Using a rich ad matching knowledge base, it may determine that because your post mentions Orvis and Sage Fly Rods, that statistically, an ad for Lexus or Land’s End represents your most effective advertising opportunity. For you, it may have made sense to add the tags “orvis” and “sage”, or even to request ads from the product you mentioned directly. That may improve things, but often the best matches aren’t intuitive or direct at all. In this case, the mention of a sage fly rod might flag content that is statistically compelling for (prospective) Lexus customers.
b) Human added keywords aren’t hard to provide, but only if they are fairly vague. It’s quite useful for an ad-matching engine to know that your post revolves around, say, “Business: Accounting: Tax Negotiation and Representation: VAT Related” to use a DMOZ node in their taxonomy. But that’s a lot to expect from your average blogger. Supplying the keywords “taxes” and “VAT” will help, but the granularity of author tags is likely to be quite coarse for some time, barring some big advances in blogging/authoring tools in the near future.
As for the question of whether ads in RSS feeds themselves are effective, Qumana suggest they’re not currently effective, and I don’t have reliable facts to the contrary on hand. However, I think bloggers, and publishers in general would clearly prefer to use RSS feeds simply as pointers that bring the user back to their content pages, where they control the ads, and everything else about the content presentation. If I recall Dave Winer’s comments on this correctly, RSS is the advertisement; it refers users to the content on your page, which then may be adorned with whatever ads you wish to present there. There’s definitely something to this idea. The other way to look at this is that RSS is sort of transitional in that sense: it was designed to serve simply as interesting pointers back to the content on your site. What’s been happening though is that RSS is becoming somewhat of a victim of its burgeoning success. It’s popularity has gotten the community to start thing of it as the basis for distriubuting content, not just pointing back to it.
If the trend is toward feeds as a distribution channel for your content then, the question of whether ads in feeds work is somewhat academic: they simply must. If they don’t, it’s self-defeating to put your full content into a feed. Both questions raised above aren’t really either/or dichotomies. Over time, a blend will develop between author-supplied keywords and machine-based contextual analysis as the basis for establishing the best match between available advertisements and your content. Similarly, successful bloggers and online publishers will devise a hybrid approach to advertising on their origin sites and in/through feeds. This will be interesting to watch evolve over the next year or so.
