Context, Intent and Affinity…the “CIA” of Internet Advertising

This is Part 2 of the US Internet Advertising Landscape. Read Part 1.

Context, Intent and Affinity…the “CIA” of Internet Advertising

From a marketer’s perspective, the efficacy of any messaging medium boils down to its ability to:

• Reach the widest possible consuming population. In Internet advertising, the most common metric for reach is the number of unique visitors delivered by a site or an ad network per month, expressed as a percent of all unique visitors on the Internet for that month. For example, Facebook’s reach in the US is about 21% based on the 42M visitors that use the site divided by the approximately 200M total Internet visitors per month in the US.
• Target actual consumers/buyers in the population. This is often very tricky or impossible for traditional broadcast media like TV, Radio, Newspapers etc to do, and is the most cited reason for not being able to measure the ROI for advertising on traditional media. Internet advertising was supposed to solve both the targeting and the ROI problem. ROI is definitely getting to a point where it is quantifiable for Internet advertising, but targeting, while better than traditional media, has a long way to go.

Targeting for Internet advertising is best measured, in my opinion, via the Click-Through-Rate (CTR), which simply measures how many times an ad has been clicked when it was show to visitors (the latter being the number of impressions for the ad). Quite simply, CTR = 100* Clicks/Impressions.

In the accompanying figure, you should be able to see the reach vs targeting (based on estimated CTR) for selected web sites and ad networks.

Targeting Based on Context, Intent and Affinity

Targeting options provided by current Internet ad systems vary greatly. Before I go on, let me use a simple example.

Suppose you are an advertiser interested in selling high end ski equipment. It is likely that you would be interested in reaching consumers that:

1. Are of skiing age (4-50?)

2. Are actual buyers- so that might refine your age consideration to users with a credit card (18-50?)

3. Have above average income

4. Live within N miles of ski resorts that actually have snow this winter

5. Are interested in skiing as an activity, as expressed by their time spent on skiing sites, skiing groups/discussion boards, outdoors-oriented sites, searching for skiing related destinations, travel packages etc.

6. Prefer to buy than rent….so, most likely those that ski more than M times per year.

7. Prefer branded/premium gear than generics

8. Preferably, already own one of your products

9. Are reviewing buying guides for products in your category or are comparison-shopping for ski gear

10. Intend to ski in the next few weeks

Attributes 1-4 are the consumer’s demographics that all ad systems track and target. Attributes 5-8 make up what is called the consumer’s “affinity” i.e. what he/she likes/dislikes based on past behaviors. Attribute 9 is the “context”- i.e. what content the consumer is currently consuming. Attribute 10 is the consumer’s expressed “intent” to ski, perhaps, because he/she is searching for ski gear, lift tickets, deals at ski resorts etc.

Based on the above example, ad systems use one of the following strategies to target ads:

• Intent-targeting, best exemplified by Google’s Adwords, which looks at keywords consumers enter in Google.com or into search boxes on Google’s search affiliates as an expression of their intent to do/buy something. Sponsored search ads are the output of intent targeting.
• Context-targeting, best exemplified by Google’s Adsense, which looks at the “aboutness” of the content page that a consumer is viewing to display appropriate ads. For example, if he/she is reviewing buying guides for ski gear, it is reasonable to show ads for ski gear.
• Affinity-targeting/Behavioral-targeting, best exemplified by Blue Lithium and Revenue Science, look at recent web behaviors/actions to target ads. For example, if you have spent a bit of time looking at web sites for ski resorts in the last 2 weeks, such a system might show you ads for ski gear, ski resorts, ski travel deals etc.

Relative Effectiveness of Context, Intent and Affinity Targeting

As measured by Click-through-rates (CTR) in the accompanying Figure, Intent-targeting performs best with about 2.5% CTRs, followed by affinity targeting with about 1% CTRs and context-targeting at about 0.2% CTRs. Ad prices (CPM rates in the Figure) follow targeting trends with intent-targeting commanding huge premiums over other types of targeted ads.

This should not come as a surprise. Context-targeting, by definition, has nothing to do with what the consumer is trying to do, since it targets ads based on the content, not the consumer. Behavioral targeting is better because it targets ads based on the consumer’s past patterns. But, neither approach can beat targeting ads to the intent expressed by the consumer. The only way you can beat intent-targeting is to combine it with a consumer’s affinity patterns. For example, if Consumer A and Consumer B both search for “high end ski gear”, but Consumer A prefers certain brands, it is better to show them ads from specific brands compared to Consumer B.

One easy conclusion from all this: it is near impossible to beat Google Adwords, given its ever-increasing reach, its leadership in intent-targeting and its support of text, banner and video formats.

That said, there are opportunities to carve out innovative niches in the broader advertising landscape, as shown in the next section.

Beyond Adwords: Monetization Potential in Social Networks

Searching for innovation beyond Adwords begins with the observation that consumers express intent in more ways than just through their search box. At the moment, I can think of at least three ways:

1) Through their Facebook (and other social network) expressions. Facebook has an added advantage that its users fill out fairly detailed affinity profiles voluntarily. In other words, you have intent and affinity in one place- I am sure someone at Facebook is thinking about how you monetize this combination effectively.

• Facebook Ads are a step in the right direction (they are essentially affinity-targeted ads at this time), but the bigger gains will come when
  
• (1) Facebook uses other ad networks for monetization since that will give it access to a broad range of advertisers and
• (2) sells its intent and affinity information to other ad networks, including perhaps to Adwords itself. Facebook Connect, which allows 3rd party publishers to leverage a consumer’s Facebook identity, is the first step in sharing consumer profiles outside facebook, and with it, the ability to target ad to affinities. I expect Facebook CPMs to be at least 5-10 times what it gets today (which is estimated at a paltry $0.25 CPM. Source: Pubmatic)- in other words, expect Facebook CPMs to be in the $1.25-$2.50 range, if intent and affinity targeting are applied correctly

2) Through their “tweets”- as Twitter puts it, a tweet is “What you are doing right now”. Here again, there has to be a way for ad networks to leverage a consumer’s tweets to deliver targeted ads on the web and mobile phones. This might be the answer that Twitter might be looking for in terms of how they monetize the service. To be clear, I am suggesting that Twitter license the intent data to ad networks and publishers, rather than muddling the tweets with ads.

3) Through their Sharing/”Digging”/Bookmarking behavior. While these features on websites seem to be focused on content pages, it should be possible to gauge intent when users share specific types of content pages. For example, sharing a product review page or a product detail page could foretell intent for either the sharer or the targets of the sharing or both.

On a related note, online classifieds do capture consumer intent since when you look for something on a classifieds site, you are expressing intent. Services like i-list.com allow consumers to post their classifieds to social networks. It should be possible for online classifieds to expand their reach into social networks, and thereby maximize their value.

Read Part 1.

Read Part 3