Say Hello to Highlighter

highlighter-sm-promoMost people have encountered some sort of personalization in their use of the Internet.  If you’ve bought anything on Amazon, you have definitely seen the “People who bought this also bought….”  It’s an amazing feature that makes Amazon a lot more useful, but, sadly, it only works on Amazon.  Why can’t Amazon purchases give me great recommendations on ebay?  Why don’t the things I read on the LA Times help me find great stuff on USA Today?  Wouldn’t it be great if the whole web could understand me and show me great stuff everywhere I go?
Today, we are excited to launch Highlighter, a little Chrome extension that is designed to “highlight” a personalized newsfeed of what you might want to read on any content site you visit.  No biggie, it just personalizes the whole freakin’ Internet.
Here’s how it works:
  1. Install Highlighter for your Chrome browser from the Google store here.
  2. Highlighter uses your recent browsing history to get to know you and build your Interest Graph based on the content you’ve been reading
  3. As you surf the web, Highlighter shows this handy dandy tab when it thinks there are stories on the current site you’re visiting that you’ll likeScreenshot 2013-11-04 14.09.06
  4. Click the tab to see the whole newsfeed of content Highlighter has picked just for you.Screenshot 2013-11-04 14.09.22
  5. The more you use it, the smarter Highlighter gets.
We are always looking for ways to make tools like Highlighter better. Feel free to inundate us with any and all feedback at
We hope you enjoy it!

Personalization Will Drive Publisher Empowerment

There’s a seismic shift in the way publishers are relating content to users and its going to realign traffic acquisition across the web. I spoke about it at a recent Digiday Summit that you can view here:

Publishers are re-architecting their sites to include more streamy, information-dense-feeds and relying on personalization to drive the organization of information in those feeds.

Much like how search’s information-dense-feed, powered by link graphing, tailors itself to your inquiries or Facebook’s feed, powered by the social graph, arranges stories for you based on your friend’s interests, publishers are seeking ways to intelligently organize information for each user.

Gravity builds Interest Graphs for every user on the webbernet. Its a complex, nuanced and massive undertaking. It requires our little start up to rack our own servers (no Amazon hosting, folks) to confidently operate at the speed and scale necessary to take on this task. This Interest Graphing, in real-time, and the ability to deliver relevant stories per user is transforming the web. Instead of users hitting the back button after they gobble up a piece of content, a publisher’s information dense, personalized feed will introduce the next relevant piece of information for the user. This dynamic will path that user forward into more articles on that site or send them to other partners who will then owe that traffic back to the traffic origination point.

Ultimately, as more publishers adopt personalization, they will re-establish themselves as valuable destinations for users and valuable partners for each other, shifting the balance of traffic acquisition across the web.



Broadcast Interactive Media & Gravity Announce Partnership!

We’re excited to announce the BIMLocal Ad Optimization Platform is joining Gravity as a new partner adopting our industry-leading personalization technology throughout its network of 3,000 news and information sites. This partnership will allow BIMLocal, part of Broadcast Interactive Media, to provide personalization for its partners as a means to tailor site experiences to individual user interests.

Local, general news is a particularly competitive content vertical and BIMLocal’s new personalization capabilities create differentiation and improved reading experiences for users across their network. Gravity will enable each user to build an Interest Graph in order for BIM properties to determine which articles are most relevant to the user. Each time users visit a BIM site, they have personalized stories that interest them the most. BIMLocal’s network of sites includes local television and news sites across the U.S., as well as national business-to-business, entertainment, sports, lifestyle/travel, home/garden and mobile sites.

Welcome to the Gravity team, BIMLocal!



Everyone Redesigns

Hoarding WindowWhen surfing the web I often get frustrated with how much clutter is on most websites these days.  When I worked on The Wall Street Journal site we struggled with what things we could remove. Somebody clicks on everything and as a news site it’s an industry standard to have a “read this next”, or a “Most Popular on Facebook”, or a “Trending on Twitter” so you just keep adding more and more. It’s starting to get a little silly though. You just can’t keep everything, it’s like living in your own filth.  So a new trend is emerging: clean and simple. Less is more. Great websites like Quartz and Mashable have already drastically simplified the notion of how to build an awesome news site.  Here are a few of the modules that are cluttering up websites these days:

The Newsreel Newsreel

This guy has the best placement of any recommendations on a website.  It sits above everything – even the content itself and when I worked at it was always the best performing module mainly because of it’s prime location.  Some websites will put it below the navigation, others place it right in the header next to the logo but in general it’s usually a horizontal list of a few headlines. It’s purpose is to target the people visiting from search engines or social media who don’t want to actually read that article they landed on.  Instead of leaving the site, the hope is they will see something better and stay within your site.  That’s why it’s at the top of an article – it’s trying to get people to click before they have even read what they came there for.

End of Post

Economist End of Post ModuleThe ying to the newsreel’s yang.  End of post modules are typically identical to the newsreel in design but sit immediately after the content before the comments section.  This is one of the best performing units in terms of click through rates (CTR) and product-wise it just makes sense.  Someone has just read your amazing piece of content and they like it so much they are starving for more so you hit them right where their eyeballs are with your best content and they keep eating pages.

Rail Most Popular Rail ModuleAlmost all websites have some sort of left or right hand rail which is the narrower column of ads and other modules that sits alongside the content.  This usually turns in to a dumping ground for every module you can imagine – Like us on Facebook! Look at all our Tweets! Most Popular! Most Social! Ads! Related Headlines! More Ads! Rails are one of the worst cluttered areas on a website and putting headlines in it is usually an attempt to get people to click to another article if they abandon the current article they are reading.  Basically it’s the place good ideas go to die.

The Toaster

Forbes ToasterThese are nifty little animated pop-ups that only come up when you’ve scrolled towards the bottom of the article.  They are neat in that they don’t take up any additional space on the page until they are needed and since they are small they typically only have 1 story in it so it’s a very clear signal on what to read next.  The problem is they are usually used in conjunction with other modules.  My favorite is when it covers up another another recommendation module in the rail – like there isn’t enough places for me to click.


All of these modules have to be powered by something and just like there are lots of places to make recommendations, there are even more ways on how to make those recommendations:

  1. Latest – Just a raw feed of news sorted newest to oldest.  Blogs are notorious for organizing their whole site around this (Gravity Blog included).
  2. Most Popular – Whatever the metric is: most page views, most emailed this is meant to be the stories that have been the most popular across all the readers in the last few days.
  3. Most Social – Kind of like most popular but measured by social interactions like Facebook Likes or Twitter mentions.
  4. Editor’s Picks – Yes, editor’s are algorithms too.  And sometimes they can be a very good algorithm identifying stuff before it has become popular or finding the hidden gems in a site.  Editor’s can also give a stronger editorial voice to a site which may include showing stories that are important and not just popular.

Personalization – Both a Design and an Algorithm

Looking at the algorithms above you’d think that personalization is just another way to sort a list of headlines and power a newsreel, end of post, or any other module. And most companies doing recommendations have created easy copy-and-paste widgets that do just that (Gravity included). The problem is that no matter how great the personalization is, if you put up yet-another-module on a page that has 40 other headlines for someone to click it’s not going to do that much better.  You need to simplify.

Simple is Scary

It can be really hard to figure out what to remove.  Every module, if measured (you are measuring every module right?), has a direct contribution to page views and taking it away might mean less pages overall.  But in reality it’s not that black and white.  People won’t just stop clicking if a module is taken away – they will look somewhere else.  It just means the stakes are higher if there are less places for them to look.  And that’s where personalization shines.

Recommendations, both from a design and algorithm have always been one size fits all.  But now for the first time with personalization, you can have millions of versions of that same module.  The occasional visitor can see most popular, the social networkers can see what does well on Facebook or Twitter, the sideways Google users can see things related to their search term and there long-time engaged readers can see things based on all of the content they’ve already consumed (their interest graph).  Personalization is the best of all signals: editors, social, popularity and ultimately, interests.  The stakes might be higher when showing less recommendations but if the quality of those recommendations are better there is nothing to fear.  Redesign away.

Too Many Stop Signs

Stop SignSurfing the web today, we frequently encounter stop signs, those little indicators that we’ve finished with a page (long comment streams, generic recommendation units, the giant stack of ads, even the footer that’s usually ignored). They leave us little choice or incentive but to hit the back button and return to our point of discovery. Take a moment to think about your own consumption patterns. If you’re like me: you arrive, you consume…and you bounce.

A key reason for this behavior stems from how most websites are architected: as discrete article pages linked from a central hub/home page. Such article pages were originally envisioned as digital representations of their physical predecessor in magazines and newspapers. This kind of non-flowing, discontinuous pagination is also a remnant of past technological limitations (low bandwidth, coupled with memory limitations, encouraged publishers to design lightweight pages). And let us not forget, the page view driven ad model incentivizes publishers to maximize page volume, often at the expense of a streamlined user experience.  It is an interesting state of affairs that the digital publishers today are often as entrenched in their own outdated modus operandi as the “old media” they were once so eager to supplant.

As we’ve overcome these technical and informational architectural limitations, new forms of content consumption and discovery have emerged. They emphasize content streams that enable the user to flow from one piece of content to the next. The Facebook News Feed, Tumblr Dashboard, and Pinterest boards are excellent examples. These information dense feeds, tailored to the individual reader by leveraging implicit and explicit signals, are extremely effective for continuously engaging consumers and providing a discovery experience that encourages them to go forward not back.

It’s time for content sites to get out of the past and rearchitect their article pages to flow from one story to the next. Take a cue from sites like Facebook (users average 6.8 hours per month), Tumblr (1.5 hours per month), Pinterest (1.5 hours per month) and Quartz. Remove the end of post stop signs that tell consumers the show’s over. Program a continuous news feed for each user. Lead them down a path that helps them discover the amazing content you create and watch your engagement levels multiply.