Aggregate Knowledge raises $5M from Kleiner, on a roll

You’re familiar with Amazon.com’s recommendation feature: “People who bought this book, also bought these books.”

Aggregate Knowledge is a Menlo Park start-up offering such a recommendation service on a mass scale — to any Web site. But it does Amazon one-better by watching consumer reading patterns online, and giving recommendation feedback immediately. (Amazon updates its recommendations once a month)

By all accounts, AK is doing very well. It started in April, and is already making $2 million in annualized revenue, according to chief exec Paul Martino. Tomorrow, it will announce it has won $5 million from Silicon Valley venture capital firm Kleiner Perkins Caufield & Byers. First Round Capital and others invested $500,000 in an earlier seed round. It employs 21 people, up from three in April. Recommendations are hotter than many people realize. Amazon says 35 percent of product sales result from recommendations. Martino, formerly at Tribe, said he noticed the power of recommendations while working at his previous company, Tribe — and thus his decision to launch AK.

We last wrote about AK here

Overstock.com is one of several sites that have implemented it. Shoppers of a gift basket (see image below), will see items that previous readers have gone on to view after viewing that item — saving users time, and helping them get to their likely destination quicker — since AK knows what previous readers ended up viewing.

aggregrateknowledgescreenshot.bmp

In his earnings call last month, Overstock’s chief executive Patrick Byrne says integration with AK was easy, and that it’s providing a “nice, measurable lift” despite being up only a few weeks.

AK offers the service for news sites, too: It links to articles that previous readers of the same article went on to read. It also helps find more relevant ads, tracking which ads are popular based on the behavior of past viewers. This is where AK hopes to beat Google. Take, for example, a reader of Fox Sports, who learns their team going to the Super Bowl. Google might offer an ad for ticket merchant RazorGator. However, AK would skip RazorGator altogether and offer a way to buy Super Bowl tickets directly. In other words, it will offer an ad, a product, or a service – depending on what the reader is most likely to want, based on previous behavior. AK tracks click streams during sessions on a Web site; it does so anonymously, aggregating data so it knows what readers are most likely to do.

AK gets paid based on performance. If the customer is a news site, AK gets paid for increasing page views. If the customer is a product site, AK gets paid if it sells more products.

AK takes several days to customize its product for sites. By first quarter next year, Martino tells us, he wants to make it plug and play. VentureBeat, for example, could get a widget that allows its readers to see what other readers have also read. Sphere does something similar now. See the “Sphere it” button at the top of this article. If you click it, you’ll see mostly other blog related material. However, Sphere’s recommendations are based on related sites and content, not necessary on where people have actually gone.

AK’s competitors include Boston’s ChoiceStream. Its software reportedly takes longer to deploy. Loomia, of San Francisco, is another player.

Trackbacks

  1. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  2. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  3. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  4. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  5. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  6. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  7. [...] 世界电子商务巨擘,Amazon 亚马逊的销售,35%来自于各种推荐服务,详见 35% of Amazon’s sales。 [...]

  8. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  9. [...] video recommendations account for about 30% of all views.  Some have speculated that up to 35% of Amazon’s sales are driven by recommendations.  And at Netflix, a full 60% of all video rentals are driven by [...]

  10. [...] 9月21号Greg Linden又发表了一篇blog提到了这个问题,里面给出的数字是35%。然后他引用的是这篇文章。这篇文章中有下面一段话: [...]

  11. [...] of Netflix views are a result of Netflix’s personalized recommendations” and “35 percent of Amazon product sales result from [...]

  12. [...] 9月21号Greg Linden又发表了一篇blog提到了这个问题,里面给出的数字是35%。然后他引用的是这篇文章。这篇文章中有下面一段话: [...]

  13. [...] recommendations" and "35 percent of [Amazon] product sales result from recommendations" ([1] [2])When doing personalization and recommendations, implicit ratings (like clicks or purchases) are [...]

  14. [...] a San Mateo, CA-based start-up offering recommendation software for content and other services, has received $5 million in funding from Kleiner Perkins Caufield & Byers. First Round Capital and others [...]

  15. [...] 从搜索引擎得到以下几个数据: 1. 推荐系统带来的销售额占整个Amazon的20%-30% 来源: blog.kiwitobes.com/… 2. 推荐系统占35% 来源:venturebeat.com/2006… [...]

  16. […] mix, with its “What Other Items Do Customers Buy after Viewing this Item?” According to Venture Beat, these automated recommendations generate 35 percent of the site’s […]