Globe and Mail uses AI to direct readers to coronavirus coverage

By Sonali Verma

The Globe and Mail

Toronto , Ontario , Canada

Connect      

When the coronavirus started ripping through Canada, The Globe and Mail did what every other newsroom in the world did: Mobilise reporters and editors to focus on coverage. What we did differently was to ignore 99% of the content that had already been posted to our Web site.

That’s because, for about a year now, we have used an Artificial Intelligence system called Sophi to autonomously run practically all our digital properties, including our homepage and key landing pages. While we focus on producing the finest journalism possible, Sophi takes care of placing and promoting it on our Web pages.

Every 10 minutes, Sophi looks at all our content, at every place each story is being promoted, and at all the ways our audience values that content. It finds valuable stories that merit greater promotion and updates every page of our Web site accordingly.

The result has been a dramatic improvement in our digital engagement.
 
Sophi understands how much each story contributes to subscriber retention, subscriber acquisition, registration potential, and advertising dollars. It looks beyond just page views, thus avoiding simply viral content.

Every single article we publish is scored according to how much our readers value it and, therefore, how much value it brings to us:

The Sophi automation score helps identify how much value readers put into different types of content — and how that translates into revenue.
The Sophi automation score helps identify how much value readers put into different types of content — and how that translates into revenue.

Our data scientists also ensure that “promotion bias” is stripped out of the equation so articles can be compared fairly:

By eliminating "promotion bias," data scientists at The Globe and Mail can ensure they are accurately comparing metrics.
By eliminating "promotion bias," data scientists at The Globe and Mail can ensure they are accurately comparing metrics.

Sophi Automation works because it is an AI engine that has been trained by our editors to understand what is relevant to each page (e.g., the home page vs. the business section page) as well as which articles are not appropriate (see Slack integration screenshot below).

Sophi Automation has been trained to understand what stories are relevant to each page.
Sophi Automation has been trained to understand what stories are relevant to each page.

Our editors have also set the parameters for the mix of articles on each Web page (such as how old an article can be, which sections to draw how many articles from, etc.).

The newsroom also works with Sophi, as editors still curate the top three story packages on our home page and business page.

Entering the newsroom of the future

David Walmsley, our editor in chief, believes the newsroom of the future is one where journalists can focus on finding and telling great stories — something the machine can’t do. Therefore he asked our data scientists to automate the Web pages, slowly and carefully testing the results before gradually implementing it across practically the entire site.

Business metrics that matter are up dramatically. In A/B testing against our traditional curation, click-through rates off our home page were up more than 17%; this helped lift our subscriber acquisition rate by more than 10%. We are also generating more advertising revenue due to increased inventory availability.
 
It is worth noting that in the past year, no reader has complained or asked if a computer was behind the site.

Encouraged by this success, we are also turning print laydown over to Sophi — where it does in a few seconds what two people used to take two hours to do — freeing up valuable newsroom talent to focus on other tasks. It will soon also determine what to promote on our social media pages and in our newsletters.

Since Sophi has transformed our own business, we are now offering automation as a service to other media companies along with a suite of related Sophi products, such as:

  • Sophi Next, a predictive analytics tool that flags your most valuable articles for editors at the time of publication.
  • Sophi Now, a real-time decision support tool for newsrooms (seen in the screenshots below).

 

Sophi Automation helps identify which stories are drawing the most interest from readers.
Sophi Automation helps identify which stories are drawing the most interest from readers.

 

By using Sophi Automation, The Globe and Mail let its editors focus on producing valuable journalism rather than running the Web site.
By using Sophi Automation, The Globe and Mail let its editors focus on producing valuable journalism rather than running the Web site.

  • Sophi for Paywalls, which predicts which articles to place behind a hard paywall to maximise subscription revenue and which ones to leave in front to maximise advertising revenue
  • Sophi Dive, a tool that lets newsrooms take a deep dive into which articles/topics are working for them and which they can stop doing.

As other newsrooms discover what Sophi can do for them, they will be able to do the same thing we've done at The Globe and Mail: Improve coverage and tailor it to what readers are looking for while at the same time maximising the way they spend their time.

About Sonali Verma