‍

Almost every day, I hear testimonials from friends or family members on the amazing video they discovered yesterday, or the new song they just streamed for the first time. In a world where everything is accessible anytime, the feeling of finding exactly what you are looking for is priceless.

Usually, this discovery process of the content we consume has three main sources: 

1. Firstly, the good old “word of mouth”. You will never trust anyone more than your own family and friends, right?
2. Secondly, this content may come from a personality or from a trusted entity. Think about an influencer, a movie critic, or a magazine you feel close to. 
3. Thirdly, it may come from a personalized recommendation of a certain digital platform. Think about these “recommended for you” movies or “similar songs”. 

Nowadays, these three sources of discovery are trustworthy and very well established in the entertainment industry. 

‍

But when it comes to reading news, who do you trust? 

How do you currently navigate through the list of articles being published and advertised every day? 

How do you educate yourself on what is happening in the world, and on what is happening on the topics you care about? 

Similarly to the entertainment industry, news consumption is driven by two sources of discovery: 

1. Word of mouth. Sharing interesting articles through WhatsApp or tagging someone below a certain article. It always works;. 
2. Trusted entity. A newspaper that you like to read or a journalist who shares your opinion.

But at this stage, recommendations made by a digital platform are somewhere between non-existent and not trusted. Why is it so? Is it that difficult to help me read about the stuff that really matters, to me personally? 

Yes, it is. Because news content is unlike movies or music.

News is the Fourth Estate, directly impacting the functioning of our society. News is fundamental, it is rapid and very diverse. 

In this video, I would like to focus on 5 main differences between entertainment recommendations and news recommendations: 

> How our interests change over time;

> How our consumption differs;

> What we expect from both industries;

> Our trust in automation;

> Our explicit feedback.

This brings us to the first reason why personalizing news is more difficult than other content, like music or entertainment. 

‍

With personalizing news, every day is different 

The first challenge with news recommendations is that you’re never sure about what your readers are interested in, today. The world is changing at a tremendous pace, and topics receive more attention depending on factors that you don’t control. In essence, news is driven by model changing events. Model changing events are unpredictable news that have a direct and strong impact on your interest in a certain topic. Think about:

> The surge of a virus in a remote area in Asia;

> The passing of a celebrity; 

> The invention of a groundbreaking technology.

Even though you might not be a big fan of computers and mobile phones, the rise of the iPhone will grab your attention. And for some reason, certain people that are not at all predestined to make it to the front page, eventually do. 

‍

Model changing events are very rare in the entertainment industry. If you don’t like to watch horror movies, well only an Oscar-winning production would eventually make you change your mind. In the news, these events happen every day.

Editors-in-chief are very well trained to handle model changing events. They easily detect news that is out of the ordinary. If you are ambitious to complement the editorial “general” curation with more individual curation, you need to train your systems on detecting model changing events. We usually advise using “trendingness” as a feature. 

Trendingness differs from popularity since it looks at the speed of receiving attention instead of the total attention already received. If popularity over time is a curve, trendingness is the slope (or tangent) of that curve. 

‍

Recommending relevant articles can be easier than you think

The model changing events is crucial to deal with what’s happening in the world right now and how to recommend the right articles. Then it’s when the concept of trendingness comes into play, in other words, what is trendy at the moment. It can be used as a feature to detect those model changing events. 

The model changing event and trendingness are the first challenges that the news industry needs to take into account when personalizing articles. Stay tuned for the second reason why personalizing news is more difficult than other industries.