Problem solved: With a local newsroom of only a few journalists, how do you create enough relevant content to attract and keep paying subscribers? That question is at the heart of a project around automated content at local media group Polaris Media in Norway. The starting point was obvious: real estate sales and local business had been top scoring topics or interest among readers for the past few years. Initial results are remarkable: every second to third automated article generates a subscription sale.
“We’ve been very clear internally. We’re using automation in order to be able to publish large volumes of attractive content, but also in order to let journalists focus on our actual journalism.”
Bjørn Rønningen
The automated content project is running at titles in one of Polaris Media’s five regions, Midt-Norge. Of the 14 titles in the region, ten are now (autumn 2023) live with automated content on house sales and short summaries of the annual reports of local companies, with the rest of the titles to come. The newsrooms vary in size from a single journalist to 10–12 journalists. “We used to only be able to cover a fraction of the stories manually that we now publish automatically,” says Bjørn Rønningen, project editor at Polaris Media Midt-Norge. “Take house sales: with robot generated articles, we can now provide a complete picture and believe we thereby contribute to more well informed real estate markets locally.”
All the Midt-Norge titles that are live with automated content, publish all the articles the robots generate. It’s then up to each editor to decide whether to spend time on picking out stories of particular interest and enhance them, by rewriting the headline, phoning up a source to get a quote or adding an image. Says Bjørn Rønningen: “Manually enhancing the articles has a positive effect on how they perform with readers, which is a useful insight for us. But of course, there’s a balance to be struck – in a newsroom of one journalist there may not be enough benefit to warrant the time spent editing the automated texts.”
The primary goal of the project is to drive subscription sales. Initial data shows remarkable results. On average, the Polaris titles achieve a conversion on every second to third robot written article. And for the texts that have been manually enhanced, the figure is 1,25–2 subscription sales per article. “We think these great results are partly due to this content being new, and we don’t expect the conversion level to remain quite as high,” says Rønningen. “Nevertheless, we fully expect to get ROI through subscription sales alone. Any extra ad impressions are bonus on top.” The data also shows that a relatively high proportion of readers are under 40 years old – for the real estate content, it’s at 13%. Bjørn Rønningen adds that the team believes the automated content will have a positive impact on churn: “We think it may contribute to a few percent of readers deciding not to cancel their subscriptions.”
So how do local readers actually find the automated articles? Each title has a dedicated section for the real estate and local business content. There’s also a carousel on each home page showing the 3–4 most recently published texts in addition to which editors sometimes manually publish articles of particular interest on the home page. Bjørn Rønningen says there’s more work to be done to help readers find the stories most relevant to them. “With some 1000 articles coming in every month, readers don’t necessarily see the ones that are closest to home for them. So we’re looking at creating a map or search function that will allow them to sort the real estate content on geography, price, date of sale. And similarly for the business content – making it searchable on size, industry, date. These types of functions are high on our list of priorities right now.”
Too much content, too little time is the classic use case for automated content in local media. In other words, small local newsrooms do not have enough reporters to cover all the journalism and information that local readers need and want. And while journalism is of course produced by journalists, robots can help with the more basic community information content. This is the logic behind long-running automated content cases in local media, like Swedish Mittmedia filling a measured demand–supply gap in their coverage of the real estate market beginning in 2018. Or five reporter newsroom Bärgslagsbladet deploying automated sports coverage in order to focus journalists’ time on stories that engage readers across the board.
Similarly, while driving conversions was the primary objective of the Polaris project, freeing up reporter time was indeed a clear secondary goal. “We’ve been very clear internally,” says Bjørn Rønningen. “We’re using automation in order to be able to publish large volumes of attractive content, but also in order to let journalists focus on our actual journalism.”
Ultimately, it’s a win-win for newsrooms and readers.