Automated articles of annual reports don’t quite sound like the path to attracting readers, do they? But they have proven to be a major way for Schibsted regional media house Bergens Tidende to explore the next level of news automation.

Automated journalism started out as a pure experiment in Bergens Tidende almost three years ago. With good help from news automation company United Robots, it has overnight become an appreciated addition of content for subscribers. The real estate robot has paved the way, proving to be a valuable and cherished feature, and has since scaled to several other media houses.

The treasure of data

We already knew that text automation is great for assembly line production of content. If your goal is to publish as much as possible, automation is the only way to do so in the current media landscape. All real estate sales in your region, all football matches in your area, all the traffic reports: If you have good, quality data, you’re all set.

A regional newspaper such as Bergens Tidende is not staffed for niche journalism or hyperlocal content. The newsroom must go with stories that are interesting and important to as many as possible. And it must prioritise time and effort consuming investigative journalism. This potentially leaves society and readers with blind spots.

With this as our premise, last summer another automation assistant saw the light of day in the BT newsroom: The local business robot. In short, the subscriber feature transforms complicated annual reports into easy-to-read articles, enriched with all available editorial tools and context.

In Norway, these annual reports are public material and have turned out to be a treasure chest for automation. The reports tell us which companies are growing and which are struggling. They tell us which local industries are booming and which are in the red zone. They tell us how much money owners and CEOs make. And a lot more on top of that. 

Turning data into news

So we got our hands on high-quality data and it proves to have a lot of editorial value. Our next question, then, became: Is it possible to automate the delivery of general stories with high relevance and newsworthiness? To deliver the actual news?

The answer is yes.

To accomplish this, we imitate as much of the editorial process as possible. A formula-based headline machine scans the texts for the most interesting aspects. The same principle is used for lead text and subtitles. A fin masked system for semantic variations makes sure that articles don’t look too similar.

We have various solutions for illustrations, using both public aerial photos and Google street view. On top of that, we’ve also established an assembly line for graph making, allowing each story to be enriched with sophisticated visualisations.       

This all leaves us with automated content looking like what it is: News stories.

A smash hit

The numbers so far? They are quite astonishing, actually. The service is a massive success with subscribers and has frequently topped the most-read lists since launch. That’s really unheard of when it comes to automated content.

Nearly 100 articles have been read by more than 10,000 subscribers, an outstandingly high number for an automated service in a publication the size of BT. Most of these stories are slightly edited, but still, only a small percent of work is handcrafted. The robot can thus be seen as an energic junior reporter supervised by newsroom staff.

Some more numbers: I

• In total, two million pageviews were generated from July to December.
• In the high season, about 10,000 subscribers visited the service daily.
• In comparison, BT had 65,000 active digital subscribers daily at the turn of the year.

Three for the price of one

The sum of what’s only important to a few, in total, adds up to far more than the news industry has been able to handle so far. Automation is changing this, bringing hyperlocal and niche stories to the readers in a massive volume.

The local business robot builds on this principle. It’s as local as news gets, reporting on neighbourhood business life such as your local diner or your local grocery shop. It is also a niche feature, allowing readers to dive into businesses rarely mentioned in the news. I mean, it’s not often you see headlines like, “This plumber has the highest earnings in town” or “This fast food place has the highest margin in downtown city.”       

On top of that, the feature has proved to be able to deliver well-read and meaningful general articles from all areas of local business life, also the undercovered part.

This is the next level of automation and what I call three for the price of one: General news stories, hyperlocal news stories, and niche news stories.

Value for money, in other words.

But more importantly: Value to readers.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .