Blog Post

Where’s the AdSense of Audio and Video?

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

By giving anyone and everyone the power to automatically attach advertisements to their web sites, Google AdSense has become ubiquitous over the last three years. What Google hasn’t done is syndicate contextual ads for user-generated rich media on such a large scale, leaving that opportunity up for grabs.

Everyone knows that advertising on non-professional rich media is a big opportunity, but nobody’s making a bundle off it yet. But Google will surely be motivated by spending $1.6 billion on YouTube. Yahoo today (prior to announcing lower-than-expected sales for the quarter) announced two related investments: the purchase (for an undisclosed sum) of self-service interactive ad platform AdInterax, and a $45 million investment led by Yahoo in Right Media’s marketplace for non-premium inventory.

The challenges here are 1) determining context for stuff that isn’t text, and 2) selling lots of ads so there’s always something relevant in the inventory. At this point, startups in the space seem to be more focused on challenge #1, while they will have much more trouble than Google and Yahoo on challenge #2. What we are wondering is, how soon will speech recognition (and perhaps image recognition) take off as a way to serve syndicated contextual ads that are matched to what’s really happening in content rather than what it’s titled or tagged?

AdSense’s power comes from analyzing the words on a page. Keywords tie into Google’s massive list of AdWords advertisers. Last quarter AdSense earned $997 million, 41 percent of Google’s total revenue. The company has explored video advertising to an extent, but not on a wide scale (have any of you ever seen a Google video ad?). Google would not provide additional detail for this piece about what it plans to do to beef up AdSense’s relationship with rich media.

The first step would be improving rich media search (something Sergey Brin mentioned on the YouTube conference call). Google bought Neven Vision, an image recognition company, and new family member YouTube has announced that it is developing video fingerprinting technology to aid copyright protection. On the speech recognition front, Google has been much quieter.

Establishing a soundtrack of a piece of audio or video content and analyzing it is something a lot of companies are working on, including search startups Podscope, PodZinger, blinkx, and Pluggd. This seems like the simplest and easiest way to improve understanding of context in podcasts and user-generated video, then place an ad (could be text, audio, or video) next to it (or before it, after it, what have you).

PodZinger, a funded subsidiary of BBN Technologies, recently broke ground by announcing an advertising program for podcasters. The company is splitting revenue evenly from the short audio and video ads, which are being sold through ad fulfillment houses. However, the effort is made much weaker by PodZinger’s attempts to avoid being invasive by only including advertising when 1) a podcaster has opted-in, and 2) a podcast listener is streaming content direct from the PodZinger site. It’s not even clear that podcasts are a massive market opportunity; this limits the company to what’s bound to be a small slice!

Another early effort comes from rival Podscope, which is working to sign up video-sharing sites to use its own speech-recognition-driven ads. Podscope is going the text ads route, displaying new ads every 30 seconds as a video plays. Podscope CEO David Ives set us up with a demo here. Ives contends that using Google’s (or Yahoo’s or Microsoft’s; whatever’s available) “will always be more cost effective than hiring an advertising sales team” for his company, because of the vast inventory of keywords large sites can offer.

Ives railed against pre-roll ads, something the YouTube founders also did quite actively before their acquisition. However, competitor blinkx, which has hammered out deals to provide AI-juiced video search to companies like Lycos and Microsoft, reports it is injecting pre-roll ads into some of its videos with the help of Lightningcast, Eyeblaster, and ITN. Another competitor, Pluggd, which is currently closing a seed round, says to check back in 2007 about an ad platform. ScanScout, which we believe is funded by Ron Conway and First Round Capital, is also apparently entering this space.

We also talked to a few ad companies that are focused on the user-generated space. They’re a bit more tentative, saying they’ll integrate with such a system but not build one themselves. PostROLLER CEO Tod Sacerdoti, for instance, said that even though he’d always prefer to have more data about a video he’s providing an ad for, “The challenge we’ve had is it’s not clear yet that the context actually drives the best economics.” Sacerdoti reports sometimes general categories are plenty to provide a relevant ad. “We’ve seen that knowing something is an automotive video monetizes better than knowing it’s a Camaro video.” (Disclosure: True Ventures, which is an investor in GigaOM, is also an investor in PostROLLER.)

Well, there we go spending more than 800 words wondering what kind of ads we will be served next! Guess that’s life on the internet.

Other interesting takes on related topic from the last few days include a BusinessWeek story on video search, a CNET piece on Google’s video ads, and Niall Kennedy’s roundups on image search and audio search.

13 Responses to “Where’s the AdSense of Audio and Video?”

  1. Liz, I think the main issues for video based Ads will resolve into:

    (i) Getting a much better fix on the user’s profile and real time behaviour as part of building up their “context” – and I suspect this will be partly done by getting user opt-in for some benefit.

    (ii) Being able to repurpose existing video inventory to match to these user profiles.

    (iii) Building tools to allow easy self production of video media for all the SMB’s that are currently the main users of Portal (as opposed to Madison Ave) advertising services.

    Despite all the work on analysing Video, I suspect in the short term the real benefit will come from letting users drive the metadata definition.

    I also do not think that Adsense (ie classified Ad equivalent) is the same thing as what will work in Video Ads – I have no hard data, just an instinct. We have done quite a lot of consulting work on Interactive Video ads and they are consumed very differently.

    I feel a blog coming on…:)

  2. Liz,

    Great piece addressing the issue of next gen ads. I do have to say your premise, “AdSense’s power comes from analyzing the words on a page.” seems a bit off for the traditional ad model.

    Google, with AdSense, took on the daily/weekly classified sections making a market that was global instead of local and relevant to a search for a similar concept.

    Video and audio have different ad markets that less so focus on the direct content in the stream, and more on the target audience that consumes it. I’m not sure what kinds of ads would be populated contextually to match Desperate Housewives, but the audience is so much more attractive than the pure content that’s on the screen.

    Does this mean that technology isn’t going to be important? No. I’d recommend that a model that worked for classified sections is different than the model that is in place for commercials.

    A good first look and I agree that Google will have something to say here. There is a lot of benefits to come from an ad marketplace.

  3. Hi Todd,

    You’re right; video aggregators like Revver are whole ‘nother category of companies that could play this game. Between search companies, ad networks, and content aggregators, it will be interesting to see who can get this done.


  4. I’ve seen an interesting company (which I’m not affiliated with) called Revver. The ad’s are post-roll and pay per click. I’ve tried the system out and it’s a little bit slow but a neat concept!

    Not sure how they determine the best ad to put in a clip but I’m presuming it’s tag based.