Jeremy Liew has a worth-reading post over on Lightspeed's blog, with a rather long winded reply from me.
Peer39 is a semantic analysis-enabled ad network. The company's approach to information retrieval, their market focus, and their understanding the limits of their tech are what makes this company the first viable semantic web company. While the company does all the usual "natural language processing" heuristic stuff which has come to be synonymous with Web 3.0 / "The Semantic Web," they also do what appears to be collaborative filtering and machine learning. In other words, they are at least partly making up for the shortcomings of heuristic approaches to information extraction with statistical analysis.
As I've remarked before, it is impossible, given the current (and reasonably foreseeable) state of computer science for IE to work well enough to bring about the vision of the Semantic Web in the oft-cited travel agent example. You can do NLP query parsing, define microformats, come up with better and better ontologies, and so forth all you like, and you will never solve the problem of incompletely, inconsistently, and poorly tagged source data. Machines are too stupid and people are too lazy for all that data to ever get tagged right. These things will not change in our lifetimes.
What makes Peer39 a sensible company is that they understand this and their goal is not to create a domain non-specific, highly accurate, robust information extraction service that enables the Semantic Web. They just want to analyze content somewhat less inaccurately in order to enable ads to be served that will get a somewhat better clickthrough rate. Improving CTRs is highly measurable and gets you paid; online ad serving is one area where having a better mousetrap really will get the world beating a path to your door.
My guess as to why this company is doing it right is because the founders and key technical leaders come out of online advertising and intelligence services. The ad people know where the pain points are and what level of "better" is enough to get market traction; the ex-spies know the limits of semantic tech and information extraction because intelligence services have been using that tech in production longer than anyone - those guys know what level of "better" is truly achievable, and how. This team contrasts with most semantic web startups which are long on "visionaries" and researchers, and short on people who have had to use this tech with money (or lives, or national security) on the line.
This will be an interesting company to watch.
Another eMarketer article that is illustrative of the primitive state of thinking that prevails in web marketing: Behavioral Targeting and Customer Segmentation . It seems like people still are dazzled by the possibilities of having so much fine-grained data, not remembering that having lots of the wrong data is no better than having none, and maybe worse.
It would seem that few marketers are using behaviorally-based segmentation to do their ad buys because they do not believe it is effective. No surprise there. A quote from the article: " 'Despite the inherent logic of placing consumers in various affinity groups based on their actions rather than their demographics, testing which segments work best leads back to behavioral targeting's Achilles' heel: reduced reach,' said David Hallerman, senior analyst at eMarketer. " A more or less true statement, in and of itself, but it completely misses the point.
In the first place, one does segmentation precisely because one wants to reduce reach. It's more important, and generally more efficient, to reach a few of the right people than a lot of the wrong people. That's what segmentation is for, at least as far as media planning goes. However, saying that behavioral targeting is ineffective relative to demographic segmentation is like saying pigs are poor at flying compared to fish. The right way to do segmentation (and yes, there is a right way) is to do it based on needs. Demographics can be correlated with needs, so if all you can get are demos, then you may be better off than with nothing. It is possible that behavioral patterns (meaning clickstreams in this context) correlate well with needs, but that is by no means demonstrated, and there is a penumbra of evidence to the contrary.
What the article gets 100% right is that segmentation is only useful if one picks the right segments. A corollary to that is that to pick the right segments, one must pick the right basis for segmentation. If you want to sell something to someone, it sure helps to know if they need it. It helps to some degree if you know that person can afford it, lives where they can get it, consumes "X" media so you can tell them about it, etc, but those are all secondary. Knowing some people have clicked here, there, and in that spot, in this sequence, within the narrow context of what you've provided them, are of tertiary significance at best.
Pick the right segmentation basis: needs. Do a valuation on the resulting segments - serve the high value ones. Make sure your segmentation solution includes major dimensions that show if/where those high value segments are accessible. If they turn out to not be accessible, tweak and rerun until you get accessible, high value, needs-driven segments. Then place ads.
Mashable has a great post here. 1,000 words worth:
This is bad in so many ways. Adam discusses most of them so I won't bother to here. However, there's one other way in which this ad's placement is bad. It is contextual advertising. Contextual advertising, even when the content context matches the ad and isn't a bad venue for a brand, is a poor way to target. Context is just a presumed indicator of the reader's interests (if you're reading it, you must be interested in it). Wouldn't it be better to know what the ad viewer is actually interested in?
In Tim Zuckert’s article in Ad Age “Become One with the Game” there is good discussion, with numbers, about why games should no longer be ignored as an advertizing medium by mainstream marketers. Sound advice. However, I think some of the prescriptions are not quite on target.
First, why this is sound advice: TV viewing, and all oldline media consumption is down and ever dwindling. Nowhere is this more true than in highly sought after demographics like males 18-34, who are leaving TV in droves, largely for the internet and games (and games on the internet). Forward thinking marketers shouldn’t just be going to games because the audience moving there. They should be going to games because games offer interactivity and a means of achieving engagement that passive media just can’t compete with. The only little thing that gets in the way is that nobody really knows how to do this. Making a banner ad into a clickable game of pong doesn’t count. “Click the Monkey!” really doesn’t count.
Making games is hard. Making games fun is hard, and requires more than a little magic. The most talented game makers never set out to make a game with any top of mind goal other than to make a fun game. Remember “games for girls”? Look at the “serious games” scene. Games with an axe to grind are not fun, and not-fun games are not successful. This is why advergames suck. You can’t create magic while your top concerns are showing your client’s brand in the most flattering light and communicating the product’s unique selling proposition.
OK, so I’ve contradicted myself. I’ve said that ads have to move into games and be interactive and engaging, but I’ve said you’ll never make a fun advergame, so don’t bother. Well, maybe that’s where “becoming one with the game” means something different to me than what it means to Mr. Zuckert. P&G doesn’t produce soap operas anymore – they leave it to people who know the form and want to create compelling content for its own sake. That compelling content creates a venue in which soap companies can place their ads. Aston Martin doesn’t produce James Bond movies. The company is content to have a symbiotic relationship with Broccoli & company, which creates a highly compelling venue in which to showcase AM’s products.
Google and Microsoft understand this dynamic, and have each bought startups (Massive and AdScape) that specialize in in-game ad serving. There are several startups that provide product placement services for games. Most next generation game engines support arbitrarily streamed sound and textures that will facilitate live, personalized ad serving without breaking immersion. Some next generation game engines (and a few current ones) allow for objects with scripted behaviors to be streamed into the game, allowing for real interactivity and unique game mechanics to become part of the marketer’s arsenal.
The tools and infrastructure are all there, pretty much. It’s time for forward thinking marketers to become one with the game, not by trying to make their own soap operas or using banners with annoying clickable monkeys, but by understanding the possibilities and working with people who make great games. Talk with Google. Talk with Microsoft. Talk with people who actually make great games. Most important, if you’re a marketing manager and don’t already play games, play a lot of different games and play them a lot so you know what’s fun, what’s not, and why. That way you’ll know what might work for your brand (and maybe how to go about getting there), why you shouldn’t be making advergames, and you’ll have a lot of fun.
Behavioral targeting is the new frontier in internet advertising. There’s an interesting article on were we are today here in an article on eMarketer. The gist of the article is that it’s not quite there yet – despite all the buzz, there is very little being done among advertisers that takes advantage of behavioral analysis-based ad targeting technology. Partly because it’s new and immature, partly because of privacy concerns, and partly because of the adjacent content –placement problem.
More fundamentally, the real reason is that behavioral targeting is just another proxy. It doesn't solve the real problem. It gets you part of the way there, and maybe part way isn't enough to really motivate advertisers. Sure, behavioral targeting can help you understand what customers have done within the context they have been provided, and thereby helps predict what new customers exhibiting similar patterns in similar context might be inclined to do. This can be quite valuable. However, it doesn’t get to the heart of what strategic ad targeting is all about: knowing who a customer is, what her needs are, and providing something truly, personally relevant.
Behavioral targeting, like nearly all internet ad targeting and behavioral analysis, is essentially forensic. You can only predict based on what you test. You can only test within the narrow context of your existing site design, content, information architecture, etc. Run A/B tests all day long, use really spiffy experimental designs, and use all sorts of fancy math, but at the end of the day you are constrained by your context, and only get viable analysis within a narrow scope. The resulting prescriptions are tactical, not strategic.
To come up with the right creative, to find the right venue to place it, you need to know the customer – who they are, what they need – not their clickstream. Don’t get me wrong: the clickstream is a valuable data set, and is certainly better than nothing. Knowing what your customers do when confronted with a certain situation is certainly better than just guessing and hoping. However, to act strategically, you need to know your customers’ real interests, needs, and preoccupations. Fundamentally, behavioral targeting is destination-centric. Effective advertising needs to be people-centric.