It is very common to see the mobile homepage of a news media site mirror the story order and arrangement of its desktop counterpart – and this seems irrespective of whether a native or web app is in place. We have unique audience and front page data available at Visual Revenue that proves this a suboptimal strategy.

With that said, there are many reasons for replicating homepage programming across channels – whether that be technical restrictions or a desire for consistency. Both very valid reasons. However, technical handicaps should rarely be used in defense of an optimal user experience and a consistent editorial tone might not be a warranted concern, given that other channels such as RSS, email and social are already different editorial experiences to that of the desktop homepage. Further to this, consumption data suggests that the audience does not consume content on both channels, desktop and mobile, at the same time, adding additional ammunition to the concept of separate programming.

A great example and very fine news destination in its own right is the local Bergensavisen newspaper [1]. Find below an exciting visual of the above described challenge, or optimization opportunity, if you will.

As you can see, the desktop homepage is replicated on the mobile homepage. The performance, where front page CTR was chosen as a proxy for success, simply does not replicate.

Here lies the opportunity and we encourage all publishers we work with to take advantage of this. Remember, this might be as simple as introducing one feature story on top of the mobile edition that is programmed manually by editors.

Cheers :-)
/ Dennis (@dennismortensen)

[1] Bergensavisen owner, media conglomerate A-pressen, (a Visual Revenue customer) kindly allowed us to show the above screenshots. Without a doubt, they are, one of the more sophisticated publishers we know and their platform integration across 65 newspapers, websites and mobile destinations is simply impressive. Don’t believe that they are not working this angle as well.

Filter bubble is a term coined by Eli Pariser in his book by the same name to describe a phenomenon in which some publishers use algorithms to selectively predict what article a user is more likely to read based on prior behavior or the behavior of similar users – effectively isolating the user in a bubble that tends to exclude contrary information. The extreme version of this is your Facebook news feed.

Sarah used an example of how a given topic (child violence) was reinforced due to a prior child violence article view. This action unhappily moved her into a segment (bubble) which she was uninterested in being part of, and all due to an innocent prior click. The inverse filter bubble, or as she writes:

“what if the filter is wrong? What if it’s not a true reflection, but a false mirror — one that does not respond to fears and prejudices, but creates them?”

That is a super interesting observation!

A fair question on top of the original filter bubble definition and Sarah’s new observation might well be; how do we, the industry at large, help publishers break out of these filter bubbles – or perhaps better put, not participate in creating them to begin with.

To answer this question, or any question of this type, I believe it is important to differ between content creation and content promotion. Whereas the content pool at Yahoo’s (the publisher Sarah used as an example) disposal, by any standard is so large that any editorial (or commercial) willingness to break out of the filter bubble is possible. Which quite clearly and to no surprise leaves this effect as a pure content promotion concept, and one where maximizing for highest possible return at all times is a standard when applying aggressive personalization.

I very much believe publishers can apply a more balanced content promotion strategy while continuing to be financially successful, simply by marrying the Editor with proper decision support - not by eliminating the editor all together. It seems naive to present, as we see so often, two polarized options only:

• Machine learned personalization (which can only be achieved through 100% automation)
• Editors fueled by gut feel and diet coke (hand curated stories selected for the homepage)

I’ve been extremely eager to execute on our goal of becoming the Bloomberg terminal of the newsroom – and not only that, assuring along the way that it was obvious that the platform works for the editor and not the other way around. Decision support! Not editor replacement and personalization with maximum effect in mind at all cost.

Editor Curated Content Promotion (or intelligent Homepage Programming if you will) should be a process where the editor decides what story to go with, but not on experience alone, but a call powered by proper data and full on decision support. Moving us towards the Data Driven Newsroom, not elimination of the news room all together!

That is not to say that personalization does not have its place in front page programming – or news programming in general. I certainly think designated and marked zones, like we’ve see advertisement flow into content spaces, should be used – which is again, exactly how we think about personalization at VR.

Cheers :-)
/ Dennis (@dennismortensen)

Article selection and front page programming, whether that is a completely manual hand curated experience or one of sophisticated data modeling on audience and previous editorial tone, is very much something that impacts the paywall business case.

Let’s use the traditional and well respected metered model of the NY Times as a backdrop for the following three conclusions:

- Better front page programming moves readers towards the subscriber threshold faster
In Visual Revenue’s case, we help increase the number of people that reach the magical 20 stories per month threshold faster. Hence, if you have people consuming only 17 stories per month (and not reaching the limit), with our platform you’d see those people consume 5 or 6 more stories per month, simply because you had the right content in the right positions at the right time. This is NOT the same as decreasing the 20 story threshold to 15, as you would want to keep everything the same – including the positive sentiment when reaching the 21st story. This is a revenue increase opportunity!

- Better front page programming increases the sign-up-as-subscriber conversion rate
Moreover, people that reach the limit faster are not, to put it simply, “tricked” in the process, but rather having discovered stories they want to read on their own. Hence, we see that they are more likely to convert to a subscriber once they reach the monthly limit. The mindset of the reader at the conversion moment, given higher story read frequency (very much like traditional marketing), is simply superior to that of those who consumed stories at a lesser frequency.

- Better front page programming decreases churn
Once you sign up a subscriber, we provide a very traditional and well proven decrease in churn by increased utilization. This is usually a difficult to measure financial tradeoff, but as there is minimal to no cost in increased utilization (more stories read), it becomes slightly easier to conclude a direct positive – and quite profitable – effect to your churn rate (not loosing subscribers as fast). We’ve all seen Netflix and others work this dilemma successfully in their segments for decades.

Cheers :-)
/ Dennis (@dennismortensen)

Nota bene (Icing on top):
I can’t help it, but this VR feature is pretty cool, you can enable Conversion Tracking and Optimization on the Visual Revenue platform and basically track which stories are more likely to lead to a conversion (subscriber) and then we can feed that information back into our model. Heck, you can even use these stories more aggressively elsewhere.

The award ceremony was last week and sadly we were unable to attend (the busy, frugal, startup life we lead!). However, we were extremely excited to find an email from Laura Wall the following morning informing us we’d WON!!!

What the judges said: “Visual Revenue enables publishers to predict the performance of a piece of content, giving online editors real-time recommendations on what to place in which position. Publishers who are not already using the technology will be queuing up to test it!”

We’ve put a lot of blood, sweat and tears into the company thus far and recognition like this is a phenomenal boost for the product, the team and the company! Thanks in particular to our customers for their continued support, without you there is no product.

For a full list of the evenings winners check out the Econsultancy Blog.

Institutions reduce the choices available to their members. (This is Ronald Coase’s famous argument about transaction costs.) This reduction allows better focus on the remaining choices they face.

The editors meet every afternoon to discuss the front page. They have to decide whether to put the Mayor’s gaffe there or in Metro, whether to run the picture of the accused murderer or the kids running in the fountain, whether to put the Biker Grandma story above or below the fold. Here are some choices they don’t have to make at that meeting: Whether to have headlines. Whether to be a tabloid or a broadsheet. Whether to replace the entire front page with a single ad. Whether to drop the whole news-coverage thing and start selling ice cream.

Every such meeting, in other words, involves a thousand choices, but not a billion, because most of the big choices have already been made. These frozen choices are what gives institutions their vitality — they are in fact what make them institutions. Freed of the twin dangers of navel-gazing and random walks, an institution can concentrate its efforts on some persistent, medium-sized, and tractable problem, working at a scale and longevity unavailable to its individual participants.

All fair arguments (emphasis is mine) and very much in line with our thinking at Visual Revenue. Institutions reducing choice materializes in such procedures as Editors agreeing upfront about some articles not being eligible for front page exposure. This could be as simple as the publisher deciding that Articles older than 4 hours cannot be shown in the Hero Spot, that only 2 sports articles can be above the fold at any point in time, or that 60% of the Articles must be from the region of their primary constituents. This type of choice reduction is realized by having us ask Editor’s to outline their Editorial Tone as a set of Persistent Editorial Instructions.

However, the state of affairs is still one of astronomical choice, where the initial limitation of choices, in the online news media industry, still leave you with an inhumanely amount of possible front page layouts. Unfortunately not thousands of choices, as Mr. Shirky so kindly suggest.

In conclusion;
Modern newsrooms either need to reduce choices so dramatically that a human editor can easily and timely decide how to program the front page, or the organization must deploy technology that help reduce choice and calculate best performing outcomes within an editorial framework. This is all we do at Visual Revenue!

Cheers :-)
/ Dennis (@dennismortensen)

• Return of the Editor: Why Human Filters are the Future of the Web
• How Algorithms and Editors Can Work Together to Burst the “Filter Bubble”

I am largely in agreement with folks like Eli Pariser, that we run the risk of creating a filter bubble, as he so politely calls it, in his book and the above Mashable article.

“It’s clear the current [personalization] platforms don’t get us there on their own. But by striking a balance between editors and code, Pariser thinks we can get the best of both worlds.”

So, sorry about the advertising, but welcome to Visual Revenue! I am super happy to see I am not alone in the belief that Algorithms and Editors Working together is a winning formula! The Visual Revenue platform allow Editors to teach and bend our Algorithms as they desire, and I am proud to state that we are, however corny it may sound, the editor’s knowledgable sidekick, certainly not the other way around.

Features like the following speak directly to Editors teaching and bending our Recommendations, in regards to what content to promote on their homepage in real-time;

• No more than 10% of the stories can be Gossip.Showing the engine that our editorial tone is in hard news.  Though entertainment stories are allowed, don’t turn my front page into a Jersey Shore fan page.

• Pinning a story to the Hero spot for 4 hours. Telling the engine to bugger off and NOT recommend other stories to this position, and by the way, don’t suggest this content to move elsewhere.

• Pass on this recommendation for the next 30 min. Telling the engine, that I get it, but you know what, I am comfortable with what I have for now.  However, if it still resonates with my audience a bit later, please bring the recommendation back to my attention.

I think you get the gist of how we operate and who we serve! It is not just about making the biggest revenue opportunity for our publishers; it is about making sure, while doing so, that we help and assist the editor in keeping their integrity and editorial tone intact.

Cheers :-)
/ Dennis (@dennismortensen)

The three of us took to the stage and pitched our respective products to a panel of judges that included John Cantarella, president of digital at Time Inc. News Group; Charlie O’Donnell, principal at First Round Capital; David Tisch, managing director of TechStars New York;  Alexa Von Tobel, founder and CEO of LearnVest.com; and Margaret Tuohy, director of product for CNNMoney.

After the pitches were over, the judges headed off to deliberate and returned to announce that…drum roll please…

Visual Revenue had won!

We’ve spent the last two years interviewing editors, wire framing, coding, tweaking, more interviewing, alpha testing, beta testing, (more interviewing :)) and finally launching the Visual Revenue Recommendation Platform!! We’re extremely fortunate to have such an awesome team (pretty aren’t they!) and I know I speak for all of us when I say we’re truly humbled to have won this event – GREAT job all!

Sadly I was unable to attend the event, but managed to follow it on the train (on my way up to ONA’11) via Twitter. Here are some of the Tweets that were bouncing through the twittersphere:

@ctull13: Key takeaway from #StartUpCNNMoney – if you want that capital, you’d best show some pride in (and passion about) your product.

@lyndabauer:”Really having the value proposition for CNNMoney right up front is most important” Margaret Tuohy #StartupCNNMoney #nyc-tech

@CNNMoneyTech: It’s official. @visualrevenue has won the first-ever #StartupCNNMoney event. Congratulations!

Thanks to the CNNMoney Team for the opportunity to pitch alongside the great guys from Qwanz and Roboinvest and we look forward to partnering with you!

We’ll also be LAUNCHING our new Collaborative Editorial Environment for our Front Page Recommendation Platform at the Conference, so do drop by and visit us at our booth (#15) to find out more:

In order for us to provide our recommendations for what to put where on the front page, we use a plethora of signals as inputs for our model (our secret sauce if you like!). One of these is how well each article you promote on a front page drives further ‘down-stream’ engagement or, put another way, how deep that article drives a user into other site content once they’ve finished reading it. We choose to expose some of the metrics we use for our modeling – article engagement will join others such as individual front page position value and its real-time CTR.

As always, our approach has been a simple one for the UI – the modeling we do to come up with the engagement score itself, is somewhat more sophisticated! Simply put, the more ‘orange dots’ you have displayed the better that article is at driving your readers on to consuming more content.

Two examples:

1) Whilst this article might be a great read, it is not driving your readers deeper into your property. This is reasonably typical of new stories that have no related content or back stories:

2) This article is a compelling piece of content, it possibly has a lot of background material or has great surrounding content (an Editor’s Picks, Must Read or Don’t Miss module) helping drive readers deeper into your property:

The engagement is calculated (as with everything else we do!) in real-time, and the score can therefore change during the day.It should be noted that we separately calculate for different content types (text article, galleries and the like) – this will ensure that an article that drives readers deeper doesn’t get eclipsed by a gallery just because it is perceived to have many page views.

This is a ‘tasty morsel’ to wet your appetites and we’ve plenty more to come over the coming month!

For each content recommendation in each time period, we assign a value.  This value could be in monetary terms, page views, engagement, or any other metric the media property uses to measure performance.  The Recommendation Value takes into account the specific value of that piece of content, content type, the depth into the site it drives the reader and the monetization methods used by the media property.

Based on the action of the Editorial Team, the Recommendation Value is categorized into either Realized Value or Opportunity Cost (See figure below).  Realized Value is the value captured by the media property by acting on the recommendations.  Opportunity Cost is the potential value loss from not taking the recommendations.  Opportunity Cost comes in two flavors:

1. the recommendation was ignored or not taken in time
2. as discussed in an earlier editorial instructions post, editors can put in instructions to shape our recommendations and recommendations prevented by these instructions are counted as Opportunity Cost. For example, if a two-day old Mayor Bloomberg article would generate the most value in the top spot but the editor has decided ahead of time that nothing older than 4 hours can be placed in the top spot, we have traded performance with editorial tone.

We don’t have an opinion on how a media property should set its tone.  We do execute on the publishers tone down to the T and keep track and report on those actions.

Recommendation Value provides full transparency into how the Visual Revenue decision support platform and enables us to clearly communicate the importance of individual recommendation to the editors.

What do you think? What else can we do to support the editorial process while maximizing value?  We would love to hear from you in the comments – or feel free to email us at hello@visualrevenue.com.