What do we know about the user
There are an endless amount of signals that can feed the personalisation of content. There are both implicit and explicit ways of gathering information about a user and how they use the product. It all depends on how much the user is willing to share with us.
Their age, location, sex, race, academic background, reading level, income level, marital status, their content browsing history, how much they know about the story or topic, their mood at the time of reading, political affiliation, the device they are using, the network they are browsing on, religion, referral site, preferred article length, the amount of devices they use during the day, the time of the day, the content their friends are reading, the content they have liked, shared and commented on, their health, shopping habits, fitness level, sleep pattern, if they are bored (scientists can now predict this).
The onslaught of sensors and new devices opens up for hundreds of signals that can potentially be used to improve the users experience of the product.
Personalised front and personalised content
The two most popular ways of personalising content based on these signals are in creating
- Personalised feeds and
- Personalised stories
The most common form of personalisation is through algorithmically generated feeds created by different signals known about the user. This form of personalisation focuses on the problem of information overload and using a sites real estate in the most effective manner possible to create a feed of content with high relevancy for a user.
Social media sites are the ultimate example of personalised feeds.
The last few years has seen a gradual transition from active consumption of news content through traditional media websites to more and more passive news consumption through social media websites. News is finding the reader rather than the reader finding the news. Friends are acting as an extra filter for a lot of the news content people consume. A socially filtered recommendation from trusted friends help people find relevant content with minimal effort.
People are also no longer just passive consumers of content but are also producers, curators, contributors and commentators of content through their social media accounts. The mix of content sources, the type of content and the commentary on shared content creates a socially curated personalised content stream with high relevance to the user. Media companies have been slow to adapt to this trend.
The Nuzzle app is another example of an app offering a social filter, choosing to aggregate and rank popular content from a users social media account.
The Daily beast app use gameified dashboards displaying analytics about stories read and skipped to try and increase user engagement with their product.
The BBC has been working on a personalisation project called myBBC and already have personalised content available through the myNews stream in the BBC news app.
The New York times has created a recommendations page based on the readers browsing history on the New York Times website.
The NPR one audio app promises to connect a user to a stream of public radio news, stories and podcasts curated just for them.
Google news and other aggregators offer personalised streams both based on explicit decisions made by the user and implicit data recommendations based on context and behaviour. (Aggregator apps: friend or foe?)
Flipboard allows users to create their own curated magazines and share these magazines with their friends
Other examples of personalisation are personalised newsletters offered by the Huffington post, personalised push messages like the breaking news app and the ability to follow stories, authors or content tags.
Personalised stories try to adapt both the content and the presentation of a story with what we know about the user. Creating different paths for different people through the narrative of a story.
Many companies have highlighted the demise of the article and the transition to producing smaller reusable structured content chunks that can be combined in different ways to create different narratives based on what we know about the user. Circa had atoms, Vox have cards, and the New York times are working on particles. This approach can not only improve the efficiency of the newsroom by reducing the overhead of using duplicate content in different articles about the same subject but it also allows for the development of personalised stories by combining atomic content units in different ways for different people.
The majority of media houses have developed interactive articles that use signals either implicit or explicit from the readers to adapt the content to the reader. Here are some of the examples.
To highlight the ageing fleet of fire brigade trucks in Norway, VG created a dynamic article that showed the fleet of fire brigade trucks at the location the user choose within the article. Readers could also contribute to the article by uploading private images of the fire brigade trucks.
The New York times article about the best and worst places to grow up published in May 2015 uses location data to personalise the article map and text based on the users location.
As part of a series of special feature articles, VG asked people to contribute their personal stories of family members who died in World War II.
Aimed at highlighting the debut of the youngest ever Norwegian soccer player Martin Ødegaard at the age of 15 VG created a dynamic article where users could enter their name and age to find out who they would have played with if they had their National soccer team debut when they were 15 years old.
This is just the start of the development of personalised stories, the next major breakthrough in personalised content will be through the use of artificial intelligence and machine learning to generate content. Within the media industry the technology is now mainly used for content discovery and curation where machine learning is used to find trending topics or to suggest meta-data tags for content. Once the meta data and number of signals we have improve, machine generated content will become more of a viable option. Where content can be automatically generated and personalised for each individual, opening up opportunities to create 1000s of personalised articles that would never have been possible without the use of AI and natural language generation.
The technology has for a long time been used by trading desks to automatically buy and sell shares, the Associated Press use it to create 3000 financial stories per quarter, Forbes have used it to create market reports such as this earnings preview and it has been used to draft report cards for Yahoos fantasy football league players.
Another often quoted example is the generation of little league baseball content through the use of natural language generation. A machine creates stories outlining the events of little league games in an exciting manner based on data about the games. Games that otherwise would never have been written about if it was to to be covered by humans. Two of the companies working on these solutions are narrative science and automated insights.
Trust is an essential part of any serious news sites value proposition to its readers. You can trust us to give you a factual and balanced view of whats happening in the world. When delivering personalised content this trust now needs to be extended to the personal information shared by the readers. The readers need to trust the news site not to do anything creepy with their information.
There are multiple examples of companies overstepping the creepiness mark, the classic example is the US retailer Target worked out how to predict when a customer was pregnant based on their purchases and sent them promotional material based on this data or Facebooks mood experiment overstepping the mark.
There has also been concerns that personalisation can help reinforce homophily, creates echo chambers, filter bubbles and a Tyranny of the Majority by prioritising what is popular and potentially exposing people only to the news and opinions that they already agree with.
The challenge is to present the user with a mix of what the editors think the user should know with what we know the user is interested in. Newspapers have a responsibility to challenge peoples views, to give people a wide angle view of the world. The advent of more and more algorithmically steered content means media and journalism ethics and responsibilities now need to be extended to the data scientists, designers and software architects designing the algorithms populating personalised feeds and stories. They need to be aware of the potential pitfalls and design systems to safe guard against them.
Another concern with the personalisation of content has been in how it will affect the brand of the company. In many instances the manual curation of content broadcast to everyone represents the brand of the media company. The brand identity will need to adapt to the new world of personalised content.
We are living in an era of information overload. We have moved from a time when a few media companies produced and distributed content to an age where everyone is a content producer and can easily distribute that content to a global audience. The need for a trusted, respected filter for this vast amount of information is needed now more than ever before.
The idea of broadcast news is a legacy feature that is from the era of printed newspapers where the cost of delivering a personal magazine for each user is impossible. Identity driven, digital distribution allows us to tailor the news for each person. Social media sites have made personalised experiences the norm, media companies have dragged print legacies into the digital world and need to pivot quickly to survive.
- How News Consumption is Shifting to the Personalized Social News Stream [August 10 2010]
- How news sites are boosting ‘stickiness’ with personalisation [October 2013]
- Reasons your google search results are different [March 2012]
- Aggregation is deep in journalisms dna [Jan 2012]
- The Future of personalisation at news websites 
- Theres no such thing as an objective filter,Why designing algorithms that tell us the news is hard [June 2012]
- The daily Me 
- The newsonomics of the newly quantified, gamified news reader [Dec 2014]
- Designer or journalist: Who shapes the news you read in your favorite apps? [Sept 2014]
- Tomorrow’s Internet: A World of Hyper-Personalized Tribes? [March 2014]
- The filter bubble is your own damn fault [May 2015]
- How companies learn your secrets [Feb 2012]
- Netflix wants to bring personalisation to mobile devices [Nov 2014]
- Three principles for personalisation [July 2012]
- Washington Post Announces Personalized News Aggregation Site [Feb 2012]
- Personalised news stream [Aug 2010]
- Mark Zuckerberg wants to build a perfect personalised newspaper [Nov 2014]
- Design for personalised article content 
- Serendipity beyond personalisation [July 2014]
- Putting news in context, automatically 
- Review of The New York Times personalisation