Peer39 Is Now Semantica
lly Plowing Through 47 Billion Pages A Month
Back in the day, and by that I mean six months ago, a ‘Pivot’ was simply known as ‘a change in strategic directions’. Call it what you may today, Peer39 made one of these eighteen months ago, and hasn’t looked back.
Their team deduced that display advertising was not going to go away, but that at the same time, privacy concerns will raise the levels of scrutiny over traditional cookie-based targeting. They saw it as an opportunity for the company and its semantic URL analysis technology.
Amiad Solomon, Founder & President, was recently in Israel, where the company’s R&D is based. I had a chance to catch up with him and hear about the impact of pivoting eighteen months ago (from ad network to data provider) had on the company.
Let’s begin with a whopper: Peer39 is processing 18,000 URLS per second. This means that their systems get hit 18,000 times per second with requests for semantic targeting information for URLs which Peer39 then has to go out and classify, then determine levels of brand safety for. All this has to happen in under three milliseconds per URL to allow the data to be monetized through ad networks and exchanges which require minimal latency.
Now let’s move on to a double-whopper: the shift from closing media buying deals as an ad network, to data distribution partnerships with the likes of AdMeld and The Rubicon Project, has resulted in a 1300% increase in scale. That’s from 35 million URLs/month in December 2009, to 47 billion in November 2010. Look at that number again, 47 BILLION.
So what does Peer39’s technology do that makes it so compelling?
First, it analyzes page content and classifies it semantically according to the most relevant categories. For example, an article about Disney’s finances might also be classified under ‘Entertainment’ if it includes information about its movies, or under ‘Lifestyle’ if it talks about movie stars.
Secondly, Peer39 assists brand advertisers to be assured that the content that’s adjacent to their ads is in alignment with their strategy. For example, Disney (this time as an Advertiser) will not want to run ads in along content it considers risqué. Budweiser may have other criteria.
In both cases, Peer39 parses the page before the ad is served, and as mentioned above, this happens 18,000 per second.
For the courtesy, Peer39 charges either on a CPM basis (between 5-10 cents for every thousand calls), or based on rev-share (15%-25% of the media buy on the ad network or platform). If you do some basic back-of-the-envelope calculations you’ll come to the simple conclusion that the only way this can become a viable businesses is through massive scale.
Solomon admits that the switch from ad network to data provider was a tough decision at the time, especially since revenues were already coming in at that point. In retrospect, he says, the timing could have not been better as more and more ad networks increased their demand for data, and large exchanges required semantics and brand protection for their real-time bidding platforms.
“We were fortunate and very lucky,” he concluded.