Family Learning Center

Posts Tagged ‘Algorithm’

Netflix settles privacy lawsuit, ditches $1 million contest

Netflix has canceled its $1 million contest aimed at finding a better recommendation engine in the wake of a privacy lawsuit settlement. The company informed its users today via the company blog, noting that it had “reached an understanding” with the Federal Trade Commission, leading it to ditch the Netflix Prize contest.

Netflix first announced the contest—actually the sequel to its original contest—in August of 2009. The goal was to crowdsource its active user base to write a more intelligent recommendation engine based on users’ past rentals. This is something Netflix already does, of course, but there’s always room for improvement; the company wanted to find the Next Big Thing™ by offering $1 million to the person with the best algorithm.

Part of the contest involved Netflix disclosing what it considered to be anonymized user data to those trying to come up with solutions. This, however, led to a lawsuit by a closeted lesbian mother who argued that Netflix had not sufficiently anonymized the information and that she (among others) could be easily outed due to her own rental history. Indeed, within weeks of the data being released, researchers had found a way to use an external data source to decode an individual’s viewing history with surprising accuracy, but Netflix did not immediately withdraw the contest.

The FTC eventually got involved and spent the following months discussing the issue with Netflix. That brings us to today: Netflix has officially settled the lawsuit and, as part of the settlement, is giving up the contest. “The resolution to both matters involves certain parameters for how we use Netflix data in any future research programs,” wrote Netflix’s Chief Product Officer Neil Hunt.

This doesn’t mean Netflix is giving up on improving its recommendation engine—Hunt ended the blog post by saying the company would continue to explore better options—but the company will have to come up with new ways to “collaborate with the research community.”

Read the comments on this post

Read the whole story…

Reader Play: Google Reader’s New Fast Flip Style Interface

Google just launched a new Google Labs product for Google Reader: Google Reader Play. Reader Play is a new, highly visual way to browse your Google Reader subscriptions that is somewhat reminiscent of Google’s Fast Flip. It replaces the busy Google Reader interface with an interface that focuses on a single story. Whenever a post includes videos or images, Play with highlight these and give you the option to read more of the text as well. This new interface allows you to browse through the feeds you already subscribe to, but Google Reader Play also emphasizes Google Reader’s ability to recommend items from around the web for you based on your preferences.

Sponsor

Big on Recommendations

As Google notes, Play will learn from your preferences, based on the articles you read and “like.” You can also choose from a set of categories (tech, entertainment, arts, business, etc.) and Google Reader will create a personalized stream of items just for you. According to Google, Play uses the same algorithm as the Recommended Items feed in Google Reader.

Play will even work if you don’t have a Google account. While you can’t star, like or share items, you can still browse interesting posts based on the categories you choose. This should make it a good tool for those users who don’t want to go through the effort of setting up a feed reader and subscribing to hundreds of different feeds.

Get Started

To use Google Reader Play, just head over here or look for “View in Reader Play” in the folder settings in Google Reader. You can switch stories by using your arrow keys or choose the slideshow mode that will automatically forward to the next story after a few second.

Discuss

Read the whole story…

Mediagazer: Techmeme Launches Memetracker for Media News

Techmeme founder Gabe Rivera just launched Mediagazer, a new memetracker for topics related to media news. This new site will be based on the same technology as Techmeme, memeorandum, the gossip site WeSmich and the baseball memetracker Ballbug. The content on Mediagazer will be edited by Megan McCarthy.

Sponsor

As McCarthy notes in her announcement, “media business is in tumult” and this is a news vertical that lends itself to memetracking. Not only are there lots of interesting news stories from a large variety of sources, but these sources all tend to link to each other a lot, which makes it easier for the algorithm to find related stories.

Mediagazer is the first new service that Rivera’s team has launched in four years. As both Rivera and McCarthy note, the team has spent the last four years learning about what works (and what doesn’t). Based on this experience, the team has “outfitted the site with the latest iteration of our automation engine, and have launched it from the outset with a dedicated human editor.”

It will be interesting to see how Rivera’s team will manage the overlap between the tech news and media news sites. Currently, for example, this VentureBeat story – which is about both the tech and the media business – is featured on both sites.

Unlike Techmeme, Mediagazer doesn’t feature a leaderboard, but there are mobile sites for smartphones and feature phones.

Judging from what we have seen so far, Mediagazer will surely become another must-read site for anybody interested in the media business, be it blogging, e-book or the state of the newspaper industry.

Discuss

Read the whole story…

Hacking the iPhone App Store’s Ranking Algorithm

If you’re a mobile app developer, one of your biggest concerns is getting noticed. Companies like Smule and Tapulous already have recognized brands; however, for the independent app designer, the promotion process requires a fair amount of strategy. One proven method of increasing downloads is becoming listed as one of the App Store’s top selling services. We spoke to faberNovel’s Baptiste Benezet to find out how indie developers can hack the App Store ranking algorithm.

Sponsor

Benezet explains that although Apple offers a variety of featured content within the web-based App Store interface, 80% of sales are made via the mobile ecosystem. While sales are driven by Apple editorial content, category browsing and rankings, because there are no opportunities to advertise in the mobile app store, it’s more important than ever to promote your app using the right marketing mix. The company collected data from the Regie Autonome Transports Parisiens (RATP) French commuter app to create a good formula for product, pricing, placement and promotion.

Product & Pricing: faberNovel found that a differentiated pricing strategy was a good way to earn revenue. While the company hit a few hiccups with its initial pricing of 1,59€, after cutting the premium service cost by 50% and building a phenomenal free experience, the new pricing strategy generated 3 times more sales. Says Benezet, “A good free version is crucial in pushing users towards the premium service, especially when the cost of the paid app is high.” In addition to monetizing via app sales, the company earned a good chunk of revenue through in-app advertising in the free version.

Promotion: Buying impressions on major websites was abysmal in drawing new users. From 500,000 impressions purchased on major French networks, the site was only able to generate 4 app sales. Benezet suggests that cross-application advertising is a better solution as relevant users are targeted via the devices they’ll use to download the new apps.

Place: Benezet argues that in the world of iPhone applications, size matters. Once the RATP application download was reduced below 10Mb, the company saw its largest sales peak ever as users were able to access the service via their 3G networks (rather than via the web-based App store). In addition to this consideration, Benezet’s most groundbreaking finding was in the App Store’s ranking algorithm. Explains Benezet, “The formula for App rankings only accounts for your last 4 days of sales.” The formula is 8 times the sales of the current day + 5 times the sales on the 2 proceeding days + 2 times the sales on initial date.

Benezet suggests that a promotional campaign of 4 consecutive days is more likely to pay off with a top ranking rather than a staggered date campaign. Furthermore, because app sales tend to increase by about 20% on Saturday and Sunday (even with a commuter app), it is most advantageous to be one of the top ranking category apps at the start of the weekend. As your sales increase over the weekend, your app’s daily ranking increases and the next four days revenue generation becomes a self-fulfilling cycle.

While the rankings remain relative to other application sales, faberNovel believes this model will help you increase the success of app downloads. After testing the theory, the company welcomes feedback to faberNovel.com.

Discuss

Read the whole story…

Google Declares "Living Stories" Experiment Success, Offers as Open Source

It’s been just over two months since Google, the New York Times and the Washington Post joined together to experiment with a new way to provide news with Google’s Living Stories. Today, Google has declared the experiment a success and has said that it will offer the project’s functionality to the general public.

Sponsor

According to Google, 75% of people who sent feedback regarding the Living Stories project have said that they preferred the format to traditional online news. But maybe that’s the thing – “traditional” online news is just the knee jerk reaction of a previously print-based industry trying to jump on the bandwagon. Trying to keep up with the health care debate by typing those words into Google or even Google News or any number of other search engines is often more a waste of time than anything else.

Google’s Living Stories allows media to provide news in a way that lets readers fully explore a story from a central location. They don’t have to scroll down through a page of search results or unrelated articles. Instead, a single page break a story down into four main components: a general summary at the top, a list of filters along the left side, a time-line of important events along the right side and a stream of updates and articles in the center. Living Stories is like a personalized RSS feed reader, but customized to pay attention to just that one story. The story is customized to the user, keeping track of what they have already seen so that it can alert them when new content is available.

When we first saw the news delivery system two months ago, we said that we could “imagine other publications employing this kind of system of organization” and now we’re glad to see that this will become a reality. Something to note, however, is that this is not an algorithm-based service provided by Google – the content of a Living Story is completely determined by an editorial team. In a way, it’s just another content management system, but one that is tailored to telling a single story and, from what we’ve seen, telling it well and from a number of angles.

Disclosure: ReadWriteWeb is a syndication parter of the NYTimes.

Discuss

Read the whole story…

Google Real-Time Search Is Now Live

Google turned on real-time search on its main site this morning with little fanfare, after a huge announcement and demo earlier this week. Go to google.com and search for a news query (Obama Nobel will work) or for something like “marshallk on twitter” and you’ll see a live streaming section of the page displaying messages from Twitter, newly updated web pages and other real-time information.

Yahoo! announced its integration of Twitter alone in its search results today, but it’s buried in obscure parts of the site and isn’t based on access to the full Twitter fire hose. Google’s implementation is much better than what Yahoo! or Bing are doing and is going to introduce millions of people to the concept of real-time search for the first time.

Sponsor

As we wrote on Monday: The new type of results are well-integrated, unobtrusive, diverse in contents and formatted simply. It appears to be a job very well done.

The new Google Real-Time Search includes far more than just Twitter results. In addition to newly updated web pages it will also include updates from users of MySpace, Facebook, open-source Twitter alternative Identi.ca and more.

As we wrote in our recent research report on the real-time web and its future, Google was already quite strong on real-time search with the addition of Google News results and timely Google Suggest items to the search experience.

Google has been working on real-time search for years. In the spring of 2006, the story goes, Google launched Google Finance onto the Web and was promptly dismayed to find that the service didn’t appear in a Google search for its own name later that day. It was after that, and a few other similar experiences, that Google engineers created an algorithm called QDF, or Query Deserves Freshness. QDF determines when results for a query need to be augmented with the newest content available, in addition to the content with the highest PageRank.

The inclusion of this type of real-time search result into the Google results page could be of comparable significance to the addition of news, videos and images via the OneBox program in 2006.

Thanks to Bob Ngu on Twitter for the heads up that this was live.

Discuss

Read the whole story…

Digg For Bargains: Deals.Woot Is Now Open To The Public

Woot, the popular bargain site that offers one good (sometimes great) deal a day, has just launched a new portal at deals.Woot. The new site is a fairly major departure for Woot, which up until now has been driven by product selections from a team of Woot employees (aside from the main Woot.com site, which is often tech/geek focused, there are special subsites for shirts, wine, and a handful of others). Unlike these sites, Deals.Woot is run by its users — it’s essentially a Digg for bargains.

The new site features a list of top deals, as voted on by the community and chosen by the Deals.Woot algorithm. This will be going head to head against other deal sites like SlickDeals and FatWallet, which have well established communities. Woot already has plenty of fans, but it may take some time to build out a base of deal hunters.

But the very top of the site actually isn’t dictated by users. Instead, it’s dedicated to “Sponsored Deals”. Woot explains that these deals are paid for by advertisers, but that they’re still bargains:

OK, yes, companies pay a little something to be Sponsored Deals. But we don’t allow just any old crap in this section. Sponsored Deals are proposed to us by other retailers, manufacturers, and even other daily deal sites. If we find the deal compelling enough that our members will appreciate us bringing it to their attention, we’ll feature it here. Believe it or not, we have a reputation to uphold.

The site has been available for weeks before now, but was only available until members up until a few hours ago.

Crunch Network: CrunchGear drool over the sexiest new gadgets and hardware.

Read the whole story…

Slowpokes to get penalized in Google search rankings?

WebProNews dug up a pretty interesting nugget in an interview with Google’s Matt Cutts over the weekend at PubCon — there’s a faction of Googlers who are lobbying for site speed to be a factor in search rankings.

That’s not taken into account at the moment, although Google tries to boost the speed of its own services with a religious ferocity. (Consider that even a 100 to 400 millisecond delay can reduce daily searches per user by more than half a percent.)

Google pulls in a few hundred different factors in search rankings and pushes changes to the underlying algorithm at least once a day on average.

Cutts said:

We’re starting to think more and more about — should speed be a factor in Google’s rankings? I mean, in AdWords, if your site is slow, that can be a factor in how much you have to pay in AdWords.

Historically we haven’t used it in our search rankings, but a lot of people within Google think the web should be fast.

It should be a good experience, so it’s sort of fair to say if you’re a fast site, maybe you should get a bit of a bonus. Or if you’re a really awfully slow site, maybe users don’t want that as much.

(Via SearchEngineLand)

Read the whole story…