The topic of a Personalized Newspaper is something I’ve been thinking about for a while now (in the back of my mind for a year or so). The idea that highly targeted content can continuously be delivered to me in a centralized location where I can consume, interact with, discuss with friends and strangers, and share it…something so fantastic that I completely forget what an RSS reader is and can’t remember how I went through life without my newspapers/magazines being dynamic to fit my interests and social graphs. It’s a concept that has been in the minds of many an entrepreneur but, due to the difficultly of it, there have been no real successes so far. These are a few of my thoughts around how I would define a Personalized Newspaper, my experiments with automated content filtering, and some of the functionality and experiences I think a true Personalized Newspaper would require.
Definition of a Personalized Newspaper
• A service that passively delivers and recommends prioritized breaking and timeless content to the user based on their ever changing interests and social graph.
My Experiments with Automated Content Filtering
I have experimented with automated content filtering two times so far. The goal was to take a large amount of content and then automatically filter it down to create a better experience.
ItsTrending (April 2010)
The first experiment was called ItsTrending, a site that kept track of and displayed what content was being shared and liked the most on Facebook. I launched this immediately after Facebook released the Open Graph APIs thinking that it would be the next Digg - but automated - AKA a million times better. Although it attracted users, press, and did actually present some interesting automated results - it wasn’t personalized. Just because a spanish music video has the most shares on YouTube right now does not mean that I am interested in it at all.
Conclusion: What is globally popular is not an accurate representation of what everyone likes. Personalization is important.
FriendShuffle (October 2010)
The second experiment was FriendShuffle, a site where you logged in with your Twitter and/or Facebook account and it let you view a slideshow of content your friends were sharing. The thought here was that your friends would be good curators and the result would be a great personalized experience. Turns out you don’t like your friends because of their interests - you like your friends because of shared experiences (college, work, family, etc.). So right away Facebook’s social graph was producing low quality results (except for when they shared YouTube videos…those were always solid). On the other hand content from Twitter produced much better results - we attributed this to the fact that on Twitter you do actually follow people because you share the same interests.
Conclusion: Personalizing to your social graph does not work. Filtering content based on topic/interest experts/curators does.
Functionality and Experience Requirements
• Serendipity: The ability to discover / be presented with content we don’t regularly consume. This is important for evolving your current interests.
• Editorial: Curators that can filter/prioritize the content in a meaningful way.
• Social: People want to discuss and share topics with friends. It is also important to note that people often hear about breaking news from a friend.
• Personalization: Present topics and content that interests the user. Collect this information (and continue to modify) with no direct input from the user (do it in the background).
• User Interface: It is important to allow simple navigation between the list of topics, a specific topic, a specific publisher, a specific author, and a specific curator.
• Sharing/Broadcasting: It is very important to allow/encourage the user to share or re-share content and/or commentary. This is important for the virality of the product and the value that is added to the content publishers (distribution).
• Content Structure: Topics are important buckets to present to the user.
Concepts That Need To Be Understood
• Breaking News vs. Personalized News: Personalized News needs to match the user’s personal interests and filters while Breaking News typically applies to a much larger audience.
• Real Time vs. Digest: News can be consumed in real time or as a digest with content/data from a set amount of time.
• News Shelf Life: Some stories are relevant tomorrow while some are only relevant now. The newspaper must understand this about the stories and present them accordingly.
• Personalization and Editorial are entirely different but absolutely required. What you want plus what you may be interested in.
• Recommendations from friends are very powerful but these content recommendation must be dispersed throughout the topics the user is subscribed to (as opposed to showing all friend recommendations for all content in one section).
• Content delivered in the Personalized Newspaper must provide the user with a vastly improved experience. The user must commit to consuming content exclusively through the product.
Ultimately the Personalized Newspaper must instantly deliver a meaningfully personalized experience upon sign up and continue to evolve with the user’s interests with no manual input required from the user - behavioral/interest data must be passively collected from multiple sources not just limited to user interactions with the newspaper. The consumption and discovery process includes data from friends, trusted topic curators, the general reaction of the entire audience, and most importantly a small amount of content that falls outside of these filters - allowing the user to discover new content and topics they may not have known about.
The Ustream Broadcaster Grid (AKA the Recommended Ustreamers Section) was placed on the home page of Ustream for two reasons.
1) Quality broadcasters and content can be promoted to the home page no matter if they are live or not. Prior to this featured content was focused on live which can cause issues because there is not always a consistent amount of live content on at any given time.
2) The threshold to join a crowd (similar to following someone on Twitter) is much lower. With just an email address you can join a crowd.
Chatroom like experience where users can login with Facebook, Twitter, Myspace, and/or AIM. All messages entered into the chat are syndicated to the network the user is logged in with (1 to 4) with a call to action for their friends to join them.
Before there was a retweet button on Twitter - PleaseRT.me allowed Twitter users to automatically add a link to the end of their message that when clicked would automatically paste the message in the update status box on Twitter - making it much easier to retweet the message.
I redesigned the Android Broadcaster app in late 2009 to take advantage of the phone’s capabilities, large touch screen, and built in UI, with the goal of creating a dead simple app. The app can broadcast live, record video in high quality to the phone if there is no network connection, initiate a live poll, view chat, and share to Twitter, YouTube, and Facebook - all in real time.