Context is King
It’s no surprise that the context in which an ad runs has a significant impact on its effectiveness. And that’s especially true for brand advertisers who are going online with their campaigns. I’ve looked at this phenomenon from plenty of different angles around here with posts like Advertisers Say Contextual Offers Best ROI and Contextual Targeting Yields Highest Return for Brand Advertisers. And now to finally put the question to bed, a new study by OTX Research has been written up by Jonathan Lemonnier in AdAge that seems to prove the point conclusively. They found brand recognition could be increased 19% just by running ads in context. By in context, or contextual targeting, they mean running ads on non-endemic content that happens to be relevant to the products or services being advertised. That is not the same as site targeting which is running ads on sites that are inherently relevant due to the overall genre they serve. This is significant to note because it opens doors for advertisers to dramatically increase their reach. No longer must advertisers find vertical endemic sites specific to their industry, they can now reach out to a much larger general audience to find specific content pages talking about related topics and actually drive greater brand recognition. This can even mean dipping back into the remnant pool for high performing inventory if you have the means of categorizing pages to determine true meaning. And that is sort of the rub here; it is not easy to discern meaning from the chaotic web and especially the unmanageable long tail. And when you step outside the well-lit confines of the premium world you immediately face the challenge of brand safety. How do you guarantee your brand will run alongside appropriate content when you are out there in the wild-wild-west of the web’s seedy underside? Real meaning and real brand safety take granularity. I am always surprised when I look at the contextual targeting solutions on the market today and realize how shallowly they categorize. Everyone seems to offer 20 or 30 top level channels and that makes good sense on the surface. That accurately reflects how advertisers generally see the world. But a high level channel only puts you in the ball park and does not significantly drive up targeting-based effectiveness. Knowing that a page is about Automotive is a good start but that will not significantly impact clicks considering this level of broad channel-ization has already become a commodity within our industry. It is when you can discern manufactures of cars from types of minivans that you really begin to be relevant to the end user. And when you can determine a page is not only Automotive in nature, and not only related to subcompacts, but also that it is about Toyotas, Hybrids and Fuel Efficient Alternatives, and Hydrogen technologies, then you have deep targeting that has a significant impact on relevance and performance. But sadly there are very few systems that can go beyond a rudimentary understanding to determining the page was really about hydrogen fuel cells on environmentally friendly vehicles. Without this level of granularity you are stuck running an Automotive ad or if you are lucky maybe a subcompact ad and settling for average click rates. The real performance comes in when you know the true meaning is environmental in nature and you can target eco-friendly ads that will have a significantly higher recall rate. I’ve seen semantic engines with only a few hundred total categories across only two superficial levels into which they must classify all the eligible content from the more than 20 billion web pages out there. To really understand meaning in the sea of available ad space you need a solution that is far more granular. You need at least thousands, if not tens of thousands of categories if you really want to certify that a page is brand safe with the goal of maximizing effectiveness and subsequently increasing ad revenues, ROI and page yield. This is why LucidMedia has more than 13,000 fine-grained subcategories behind our contextual engine. It allows us to determine the true meaning of a page before an ad is served and makes sure the most relevant ad can be shown to the user. In some of our earliest tests this deep categorization yielded a 76% jump in clicks in the average direct response campaign versus the typical run-of-network buy. When you want relevance on the web for advertising you need depth and breadth. Having just breadth only gets you half way there.
Tags: branding, contextual, effectiveness, otx, reach, targeting
May 23rd, 2008 at 6:16 pm
As contextual targeting becomes more mainstream, inclined, and baked into ad networks and exchanges, new patterns of contextual targeting will emerge. One imminent pattern is micro-segmenting, wherein brand and performance campaign managers try to optimize their campaign’s goals by slicing-and-dicing the inventory very finely. When that happens, a few hundred categories are not going to be enough; it’s likely that a few tens of thousands will be required. Publishers too will demand much more nuanced categorization of their own pages in order to increase revenue. A precedent for this pattern is the segmentation within the inventory of Yellow Pages.
A second pattern is an increasing sophistication in targeting. Clearly, placing in-context ads increases click-through, and hence revenue. However, in-context placement is a mere way-station in the path to a bigger and better strategy for ad placement, namely, exploiting the structure of the taxonomy itself in order to improve placement, hence clickthroughs, hence revenue. For example, placing an ad about photography on a page about photography is relevant for sure, but may not maximize revenue. However, placing an ad about cameras on the same page will likely increase revenue. In fact, that ad about photography may do better on a page that lists cameras, than yet another ad about cameras!
The bottom line is that the market is trending towards fine-grained categorization, because doing so increase revenue in two ways: one, increasing the worth of the inventory, and two, increasing user engagement.