Friday, November 28, 2008

Facebook And The Myth Of Contextual Advertising

I didn't like the yellow highlighting telling me what was important (and the pictures were missing!) so I copied this directly from this blog. rather than directly linking to it.

There is a myth floating around that contextual advertising is going to help Facebook justify its $15 billion valuation. The myth goes something like this: because Facebook knows everything about us, it will always be able to serve perfect ads. However, the reality is more like the following:

· Facebook does not know much about us
· The data that Facebook has is not structured
· People are not coming to Facebook to click ads

And even if all of the above were reversed, building a contextual advertising engine is far from simple - anyone who tries faces the same problem as building a personalized recommendation engine. In an earlier post on this blog, we discussed various ways that such systems work. The best example is Amazon, which uses a mix of many techniques to deliver recommendations, and took a decade to build and fine tune it to the point it is at today. So what basis is there to think that Facebook can do the same? Let’s take a more in depth look at Facebook’s advertising play.

How much does Facebook really know about you?
A typical Facebook profile contains basic personal information - name, home town, date of birth and relationship status. Another section shows education and job history, and the rest of the profile is generally applications ranging from photos, to movies, to games, and other random stuff that people find interesting. So how much information is there for Facebook to use?

Very little actually. The site does not really know what I like. It does not know my book tastes, does not know that I am running a startup, does not know that I like Cabs and Pinots. It does not know that I am a Netflix user, that I am increasingly less tolerant of cold weather, or that I have 3 beautiful little daughters.

So Facebook does not really know sophisticated things about me. But even basic information that it ought to know is beyond its grasp. For example, if I add the Flixtster application and start displaying movies that I’ve rated on my profile you’d think that Facebook would learn that I like movies. But it wouldn’t. Facebook’s system has no idea that the Flixster application is about movies and has no idea what kind of movies are being displayed.

The Facebook platform is designed to be flexible and pluggable, but it lacks meta data about the content of the applications. So all the information that is being displayed on our profiles can not become an input into a contextual advertising engine. At least not easily. Of course Facebook can design an algorithm that runs and analyzes text in the profile pages, but such system would not be very good because a lot of guessing would have to happen. What would work, is to let each application deliver ads. Since applications know their own content, app developers know which ads are relevant. But it is unlikely that Facebook would ever go for that, since their whole play is to control the ads.

What ads are we getting today?
The ads that we get served on Facebook today are the direct result of the lack of understanding of its users. Not surprisingly, most ads are about dating. After all, if this works web wide, why not do it on Facebook? But most of the ads that I have seen are either laughable or down right offensive.

Um… Didn’t I say on my profile that I am married? Okay fine, but even if I was looking, I would not want my THE ONE to look like this. Would you? Jokes aside, the advertisement below, which was displayed on my company AdaptiveBlue’s user group, is just completely inappropriate:

This sort of thing can cost us users, some of whom may not even realize that this is an advertisement. All they know is that they are looking at AdaptiveBlue’s Facebook group and could draw misinformed conclusions.

In addition to the ads in the sidebar, Facebook is now showing advertising in the newsfeed. I understand that they want to monetize the site, but this is just really confusing. We have been trained that the news feed shows updates from our friends. This is the place that we are directed to first each time we log in to the site, and having ads there simply creates a bad user experience. Not to mention that I am a happy Netflix customer so the Blockbuster ad that I was shown did not entice me — again, Facebook doesn’t know as much about me as you might believe.

How effective are the ads on social networks?
So at least today the ads are not relevant and we are yet to see how exactly Facebook is going to learn about us. But there is also another problem, a bigger problem which is broader than just Facebook. The question is how effective will the ads be on social networks? On the surface, it’s a no-brainer: every site that has traffic makes money on ads, right? But there’s more to the story than just the surface.

The most effective ad play online is, of course, Google. Its big because its model is pay-per-click (CPC). That is, advertisers only pay if users actually click through to the site. Of all Google’s offerings the most successful one is Google AdWords, in which contextually relevant ads are displayed on top of search results. It is so successful because a search represents intent, so users are more likely to click on ads that advertise things they are clearly actively looking for.

Most portals, media sites and social networks use a different model, based on impressions (CPM). That is they charge advertisers for each time an ad is shown, regardless of whether the user clicks. Naturally, these ads are much cheaper. And herein lies the big issue: If on social networks people do not click on ads, then the only type of ad they can possibly sell are impression based.

Right now, people are using social networks to socialize and browse, they are not actively searching for products and deals and likely, CPC would not be effective. It is still too early to tell, of course, but Fred Wilson, who writes the popular ‘AVC’ blog, created an ad for his firm, Union Square Ventures that illustrates, anecdotally, that users on Facebook are not interested in ads. The ad got practically no clicks, which is curious, because Fred Wilson and his firm are quite famous in tech circles and the ads were targeted at 32,000 Facebook users who declared an interest in technology.

Conclusion
As it stands, Facebook does not know all that much about us, and the ads that we are shown are not relevant as a result. The jury is still out on whether social networks can get big via highly targeted advertising. Early trials have shown that CPC is not a likely route and CPM plays are just not as interesting. Yet, Facebook is certainly very aggressively pushing for monetization, likely in preparation for a future IPO. Will they be able to turn it all around and re-invent contextual advertising? The company is full of brilliant people so it’s certainly possible, but so far it does not look impressive.