• Save
The Changing Character of Customization: Content Personalisation in the News
Upcoming SlideShare
Loading in...5

The Changing Character of Customization: Content Personalisation in the News



A talk given at Google (UK), September 2011.

A talk given at Google (UK), September 2011.



Total Views
Views on SlideShare
Embed Views



3 Embeds 10

http://www.linkedin.com 5
http://www.test.city.ac.uk 3
http://paper.li 2


Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment
  • Hello and welcome and thanks for coming. And thanks to Meg for inviting me. When Meg asked me to talk about my research into personalization, I scratched my head and thought “what on earth” am I going to tell GOOGLE about personalization. After all, most of the time I talk about the subject I start by showing Google products such as
  • Fast flip, ‘News for You’ on Google News, and iGoogle to convince audiences of sceptical academics or publishers that, in the words of Eli Pariser,
  • “the era of personalization has begun”. As an aside, Parisercredits you guys with ushering in the ‘era personalization’, when you started
  • to personalize search results in December 2009! Anyway back to the head scratching! In the end I thought you might be interested in what the study of a particular domain – news and online journalism -- tells us about personalization. Why?
  • Well, over the last 7 years news and information sites, particularly those from ‘established’ media companies have been getting more and more popular. As you can see from this slide, they now make up a greater number of the ‘top 50’ UK online brands. None of the TV broadcasters or national newspapers listed here appeared in the top 50 list in 2004. They must be doing something right then, including, perhaps, their deployment of personalization. Established media brands are worth watching not only because of their increasing popularity, but also because of how they drive discussion and debate on other sites.
  • For example in one American City, Baltimore, 95% of the stories containing new information that appeared in a one week period came from traditional media. Blogs, Twitter and local websites—at least in Baltimore—played only a very limited role. They were mainly a way to disseminate stories from other the traditional media.
  • The research I’m going to talk about today has involved interviews with the editors of these sites, and three surveys, over a 4 year period, of the different types of personalization they use. Thanks to Meg for helping with the last survey.
  • This slide is a gives an overview of the trends I found over the last four years. It shows how personalization features are increasingly common on news websites. On average sites offer eleven different forms of personalization. There are, of course, considerable variations, showing how personalization is more readily accepted by certain audience segments, for example subscribers to specialist news websites like the FT and the Wall Street Journal.
  • This slide also illustrates my distinction between two different forms of personalization, active and passive. With active personalization usersregister their own content preferences.
  • Passive personalization infers preferences from data collected, for example, via a registration process or via the use of software that monitors user activity. As we’ll see later there has been a dramatic decline in the form of personalization that demands the MOST input from users:
  • what I call ‘MyPages’. Here’s an example, the MyTimes page on the New York Times website. By contrast, there’s been significant growth in some of the passive forms of personalization, in particular what I call ‘Social Collaborative Filtering’, such as the:
  • Facebook ‘Activity Feed’ plug-in, seen here on CNN.com.
  • Indeed, although active forms of personalization are more common, passive forms have actually been growing faster in percentage terms over the last 4 years.
  • I’ve already mentioned a couple of the categories of personalization I defined in my research. There are 18 in all, they’re listed on this slide. I don’t plan to bore you by going through each one in turn, but rather I’ll highlight some of the important trends and developments. As you can see there are a vast variety of approaches, particularly in the active forms of personalization.
  • This is indicative of the ongoing search among news providers to find the most effective types of personalization—balancing the need for precise matching of content to users’ interests with the need to make the process of actively setting up personalization as easy as possible. The size of table 2 and the changes that I’ve observed suggest this search is still going on. Within table 2 there are two opposing strategies:
  • one of partially reshaping the existing news website, through, for example ‘MyPage’ functionality; and the other
  • of selectively sending material out to another device such as an RSS reader or an SMS inbox.
  • What’s interesting, at least to me, is that neither fully reflect some of the early predictions for personalization. For example Nicholas Negroponte’s influential concept of the ‘Daily Me’ where a single “interface agent” would construct a personalized newspaper for users.
  • One of the most notable changes that took place last year was the sharp increase in personalization using recommendations from social networking sites. What I call ‘Social Collaborative Filtering’. In every instance this was via the Facebook ‘Activity Feed’ plug-in, through which users receive recommendations from their Facebook ‘Friends’.
  • The problem with this form of social filtering in the news domain is the infrequency with which these plug-ins update. For example, although I’ve got a larger than average Facebook network, if I log in to see my Friends’ activity on CNN.com, 60% of the recommendations are from months ago, ancient history in news terms. This is typical for these Facebookplugins on news sites.
  • The reasons are, firstly because they only show links to stories that appear on the site hosting the plugin, and secondly the average Facebook user doesn’t post many links to stories on news sites – an average of between just 2 and 3 a year! It is clear that the increasing use of social media is prompting news websites to adopt new forms of presentation – including personalization – at a rapid rate, but that such developments are in their very early stages.
  • Another challenge is to find a balance between usability and relevance. Despite considerable evidence pointing to the fact that users are passive creatures, news providers have spent a lots of effort producing elaborate ‘MyPages’, to allow users to assemble whole pages of customised news. Here’s the New York Times’ version
  • here’s an another example from the Independent. They are much like iGoogle in concept, allowing users to select from a set of modules to add to the page.
  • These ‘MyPages’ are in decline, dropped by the websites of the Washington Post, the Sun, the Telegraph and being phased out by the New York Times and the Wall Street Journal. It seems like the uptake of these relatively demanding services hasn’t been enough to justify their existence.
  • Theeditors I interviewed put this down to audience passivity and the difficulty users have accurately predicting their content preferences in the dynamic news domain.
  • In contrast to ‘MyPages’, personalizable mobile editions have been growing rapidly, now offered by every news site. On average news sites provide personalizable ‘apps’ for at least 2 devices and over half had a personalizable mobile version of their site. This is not surprising given the growing numbers of smartphone users.
  • Indeed there are reasons they might be particularly good platforms for personalized information delivery. Firstly, due to their smaller screens and input devices their browsing capabilities are limited. Secondly
  • being location aware, as more and more mobile devices are, lends itself to personalization.
  • It’s surprising then that the personalizable mobile editions and‘apps’wereremarkably static in nature, with a minimum of personalization.
  • On average, they offered barely one and a half different forms of personalization, compared with an average of over20 for the full web editions. This thirteen-fold difference may be explained by the fact that most of the ‘apps’ were first generation, but the notion that mobile devices such as the iPad are better suited to passive consumption may be a factor.
  • I’ll stay with the idea of the passive user now and talk about the rise in passive forms of personalization. In particular what I call ‘contextual recommendations’ and ‘aggregated collaborative filtering’. ‘Aggregated collaborative filtering’
  • is where selections of news stories or other content (such as readers’ comments) are automatically filtered by popularity. Variables include ‘most read’, ‘most emailed’, and ‘most commented’. This example is from the BBC news website. I found this form of personalization was universal. It’s popular with readers and editors alike.
  • The reasons? Firstly it’s passive, requiring no effort from readers. And secondly editors like it because it increases page views, but in doing so usually reinforces their editorial judgement, as many of the stories recommended have already been selected on the front and section pages.
  • You can see this in these examples from the BBC News website where all of the ‘Most Shared’ stories were chosen by editors to be featured on the Home page.
  • We shouldn’t underestimate how much editors value their own editorial judgement! Both because it’s tied up with their sense of professional identity, but also because they believe readers want the world filtered.
  • Another form of passive personalization has also been growing, what I call ‘Contextual recommendations’. This is where
  • Links or indeed whole pages of content are created algorithmically. This is an example from the BBC News website and uses technology from Autonomy and Moreover to creat contextually related links from the story being viewed, in this case a story about the Chinese economy.
  • This is another example. This is one of many so-called ‘Topic Pages’ found on the Sky News website. These are created automatically by software from Daylife, and are aggregations of stories and pictures previously published on Sky News, mixed in with Wikipedia entries, Reuters quotes etc.
  • Daylife are by no means the only company involved in providing personalization technology to news sites. Here are some of the others. For me, as a journalism scholar, these developments are interesting because they signify a move away from the traditional concept of
  • the journalist as gatekeeper, deciding what to publish and what to ‘spike’. This gatekeeping role is increasingly being replaced by algorithms. This raises questions about accountability and transparency when
  • companies such as Taboola promise publishers that their recommendations can be skewed for commercial reasons. Is that being “evil”?! If so then I’m sure it’s something Google will never do! There’s more to be said about how algorithms AND users are displacing professional editors, but I want to finish by putting the mechanisms of passive personalization in their commercial context.
  • Traditional news publishers have been losing advertising sales, in the US since 2006 as this graph shows.
  • Although their online advertising revenues have been going up (shown in red on this graph), since 2007 they’ve not been going up as fast as their traffic (shown in blue). This is, in part, due to the effect of the recession on advertising sales, but it also a reflection of the declining proportion of online advertising expenditure finding its way to publishers. Pariser reports an 80% drop between 2003-10.
  • One of the reasons is rise of targeting advertising. In the local online media, for instance, targeted display ads which “match ... ads to the interests of individual consumers” grew by 22% in 2010 (shown in brown on this graph), whilst spending on static display ads declined by 12% (shown in blue).
  • Publisher’s margins are getting squeezed between the Ad Delivery players and the Ad Targeting Players. I would argue that personalization has emerged as an increasingly popular strategy for news publishers because it allows them to capture data about users and reduce their dependence on external suppliers of such information.
  • Of course you guys want news publishers not to worry about technology. Let us do personalization for you, you say! Well I say that news publishers need to think again before getting into bed with external providers of technology and user data. Such a strategy may be dangerous because utilizing such services come at a price: squeezed advertising margins. If news providers want to slow, if not reverse, declining advertising revenues AND challenge the reluctance of news consumers to pay for online content, personalization may help.
  • Firstly because there may be a correlation between sites ability to charge a subscription and above average deployment of personalization. You can see from this slide that I found that the three sites that offered the most personalization features are all now charging a subscription.
  • Secondly because personalization can provides a means for traditional news providers to acquire, and crucially control, a range of data about their audience. Being in control of such data will be vital as online advertising becomes more dynamic, targeted and data-driven.

The Changing Character of Customization: Content Personalisation in the News The Changing Character of Customization: Content Personalisation in the News Presentation Transcript

  • “The Changing Character of Customisation: Content Personalisation in the News”
    Neil Thurman
    Graduate School of Journalism
  • “the era of
    has begun”
    - Eli Pariser.
  • Media Companies in UK’s 50 most-popular brands online
    Growth principally coming
    from TV broadcasters:
    • ITV
    • Channel 4
    • Sky
    and national newspapers:
    • Associated Newspapers
    • News International
    • The Telegraph Source: Nielsen, Feb 2011
  • Source: Pew, 2010
  • Qualitative Interviews
  • Growth of adaptive news at eleven national US and UK websites, 2007-10
  • Growth of adaptive news at eleven national US and UK websites, 2007-10
  • Growth of adaptive news at eleven national US and UK websites, 2007-10
  • Growth of adaptive news at eleven national US and UK websites, 2007-10
  • Table 2: Active Personalization
    Table 1: Passive Personalization
  • Table 2: Active Personalization
    Table 1: Passive Personalization
  • Table 2: Active Personalization
    Table 1: Passive Personalization
  • Table 2: Active Personalization
    Table 1: Passive Personalization
  • “Imagine a future in which your interface agent can [read / view / listen to everything], and construct a personalized summary. This kind of newspaper is printed in an edition of one. . . . Call it The Daily Me.”
    Nicholas Negroponte, Being Digital, 1995
  • Changes in deployment of passive personalization features at eleven national US and UK websites, 2007–10
  • Changes in deployment of active personalization features at eleven national US and UK websites, 2007–10
  • What the editors said:
    “the time and effort to personalize something” would put off all but a “relatively small number of people”—Steve Herrmann, editor, BBC News website
    “If you determine in advance who ‘The Daily Me’ is . . . Then you may miss some of the important things that you didn’t know you were”—Rich Meislin, NYTimes.com
  • Changes in deployment of active personalization features at eleven national US and UK websites, 2007–10
  • “fat finger problem”
  • Table 3: Adaptive interactivity on different platforms compared.
    Data from eleven news providers, Dec 2010
  • Changes in deployment of passive personalization features at eleven national US and UK websites, 2007–10
  • Passive Personalization:
    ‘Aggregated collaborative filtering’
  • What the editors said:
    “[the most read and most commended features have] gone down well . . . we don’t do enough”—Neil McKintosh, then Head of Editorial Development, Guardian.co.uk
    “I think there is something very useful and informative about what people are email each other””—Pete Clifton, then Head of BBC News Interactive
  • Editorial judgement
    “if there’s anything we have it is our judgement about what people are interested in”
    —Anna Spackman, then editor Timesonline.co.uk
    Readers “want someone to do some of the filtering work for them”
    —Neil McKintosh, then Head of Editorial Development, Guardian.co.uk
  • Changes in deployment of passive personalization features at eleven national US and UK websites, 2007–10
  • Passive Personalization:
    Contextual recommendations
  • Implicit Personalization: Contextual recommendations
    …full ‘topic pages’
  • Implicit Personalization: Contextual recommendations
  • “Automatically direct users towardhigher CPM ... content while maintaining recommendation quality and user engagement”
    —Taboola website
  • Source: Newspaper Association of America
  • Source: Newspaper Association of America
  • Targeted display
    Non-targeted display ads
  • Source: Newspaper Association of America
  • “Journalists should worry about creating the content and other people in technology should worry about bringing the content to the right group...by personalization”
    “Once we get personalization working for news we can take that technology and make it available to publishers, so they can transform their website appropriately [to suit the interests of each visitor]”
    —Krishna Bharat, Google (quoted in Pariser, 2011: 62-63)
  • Provision of personalized news features at eleven news websites, Oct–Dec 2010
  • Provision of personalized news features at eleven news websites, Oct–Dec 2010
  • Neil Thurman