Making Context King...or...Can Relevance Engines Make Content Pay Off?

Content is all well and good. We love it. We revel in it. We target our editorial with laser precision to meet the needs of our audience. Then we polish that prose to a high gloss, and deliver it on time and on demand for the ever-hungry churn monster that is the Web. Content is King! Yadda, yadda. Of course, as we all now know, King Content does not rule his fiefdom alone, but must share the throne, at the very least with a Queen (ecommerce), but more importantly with a young Prince we will call "context."

Yes, you've heard this before. Web veterans will recall that "contextualized ecommerce" was the hot online business model, circa 1998. According to this pipe dream, smoked by analysts everywhere, all you needed to do was put relevant, well-targeted buying opportunities within a user's eyeshot of your precious content (or wrap enough compelling content around a buy button), then wait for the cash to pour in. Well, wonder of wonders, it turned out that readers were not very anxious to buy their mattresses from The New York Times' marketplace or their L.L. Bean hiking boots from Newsweek. And at the more full-blown content/ecommerce plays like,, and, visitors never seemed to understand the business arrangement—that they were supposed to be buying stuff, not just scanning the pithy prose. Don't you just hate it when consumers don't play along with your business model?

Despite these earlier failures, many publishers are determined to get a bigger slice of the ecommerce pie, and they are hoping that a new generation of relevance engines will get them there. All a relevance engine does, explains Vin Bhat, co-founder, Simile Software (formerly Nano) is "intelligently connect a publisher's content to other relevant content and services." In fact, his is the solution du jour this season, as none other than AOL Time Warner, along with McLatchy and Cox, are testing its effectiveness on select sites. For each new article at early testbed sites like and, Simile automatically generates links to related editorial from the publisher's other properties, to fee-based archival resources, and to very specific offerings from ecommerce partner "We can pretty much take any content stimulus, feed, chat transcript, or op-ed piece and spawn different relevant opportunities off of that," says Bhat.

If the system works as promised, it has the potential to add to publishers' bottom line substantially. In the demo we've seen, Simile Software took an article on cyclist Lance Armstrong and generated highly accurate links to Armstrong articles within the Time archive and even to buy buttons for Armstrong-related book titles (with cover illustrations) at B&N. Since Simile has already indexed much of the existing Time Warner content, as well as the B&N book catalogue, all of these links are generated prior to the article's posting. There is no dynamic page generation going on in the background to bog down site performance. The resulting page is more than a Lance Armstrong article; it is a mini-portal into the topic, a context of other content and ecommerce options for a user who has already declared an interest in this specific topic by virtue of clicking on the article in the first place.

Selling the Goods
But will this sort of hyper-targeting really breathe some life into moribund ecommerce affiliate relationships and slow- moving archive sales? Knowing that ROI is the new name of the game online, Bhat and co. came up with an intriguing formula for how much additional revenue might generate both from increased archive sales and its share of sales via B&N, which currently is producing astonishingly little income for Time Warner despite its massive traffic. Working from conversion statistics it got from Microsoft's MSN portal, Simile estimates that well-contextualized ecommerce offers generate 2% to 4% clickthroughs, about 6% of whom convert to buyers. Extrapolating from this, Simile estimates that off of every 10,000 page views to, it can yield 20 (.2%) buyers.

Small potatoes? Maybe, but stick with the math for a second. is generating at least 30 million page views each month, 40% of which Simile will be processing through its relevancy engine. Finally, fold into the equation the average online purchase from the TW archives ($2.95) and B&N ($20). So far as we can tell, and unlike previous, fanciful calculations for contextualized ecommerce, none of the suppositions in this equation is unfounded. And if all the pieces work, Simile is guesstimating a bottom line for of $600,000 in additional annual revenue.

Can Readers Become Buyers?
The model sounds plausible, to be sure. But even with this much smarter technology driving highly targeted buying opportunities to general content users, some important caveats remain. First, there is still that perennial problem of ecommerce—margins. The profit on popular goods categories like books, let alone the revenue share on affiliate relationships, are so slim that you need to target those buy buttons with laser precision and on a massive traffic stream in order to realize decent incremental revenue. Clearly, relevance engines like Simile Software's will help automate and thus reduce the cost of that targeting process. In fact, Simile lets publishers filter for certain price thresholds, so content will link only to pricier items from an ecommerce affiliate. And in fact, contextualized ecommerce still outperforms banner advertising, says Jace Wieser, chief technology officer of syndication and content management supplier, Yellowbrix. Two years ago, his company tried a similar play by helping publishers wrap well-targeted buy buttons around relevant content. Yellowbrix was seeing up to 13% clickthrough rates on well-targeted buy buttons, and conversion rates of 2% to 4% from there.

But even if the newer technology works to overcome some of the legendary problems with contextual ecommerce, will the users play along? Wieser suggests that it may actually be getting harder to convert online readers into buyers because of the way Web habits have evolved. "People are either in the mode of reading or buying, and more often than not people are not wanting to go off into a tangent," says Wieser. In fact, as Webizens become more experienced online, he finds "they got more focused and less easily distracted by links."

Whether general news and information providers can succeed in flipping visitors from reading mode into buying mode remains an open question. Which is not to say that contextualized ecommerce cannot work because there is some hard evidence that publishers are finding ways of dipping into the etail revenue stream. For instance, boasts great success in its partnership with the comparison shopping engine Integrated seamlessly within the PCWorld. com site, with many links back to the pub's review and buyer's guide content, this engine is now responsible for about 15% of total site traffic, says editorial director Kevin McKean. The lead fees generated whenever a shopper clicks through to a vendor make it the second biggest revenue stream for the site behind advertising. By offering this as a service to users when they are looking to buy, rather than an in-your-face hard sell effort that tries to covert visitors into buyers, McKean feels that he not only maintains the editorial integrity of his site, but gives Web vendors an exceptional conversion rate. "Once a person goes through a site like ours, and because we don't push them, then when that person clicks over to a vendor he is very apt to buy."

The example may serve both as a tale of encouragement and caution. Indeed, publishers can get a decent piece of the ecommerce action, and whether a visitor does choose to click through a vendor's button does have a lot to do with the relevance of an offer to the content. But what is most important to this equation is something outside of the publisher's hands, the context that the visitor brings to the site, whether at heart he is here to read or to shop. That is something that even the most powerful relevance engine cannot affect.