Success stories of the business applications of data and the shift to an audience focus through data can are only possible if the culture is right. And that culture centers in the newsroom.

Getting newsroom buy-in and setting up proper internal communication systems are key to news media companies working to create data-positive newsrooms, INMA Researcher-in-Residence Greg Piechota said on Tuesday, following the Audience Analytics Town Hall by INMA and Meta.

“Data strategy is a journey and not a project,” said Piechota, who leads the INMA Smart Data Initiative. “Projects have beginnings and ends. Your data programme might hit milestones, deliver infrastructure and products, but in its essence data is a culture. And work on culture has no end.” 

In the past eight months, Meta, INMA, and the International Center for Journalism teamed up 15 news publishers in Latin America with their peers in Africa, Americas, Asia, and Europe, so they could learn the skills necessary to become customer-centric. 

The Town Hall was the culimination of the programme, led by Piechota, bringing 1,043 attendees from 94 countries together for the three-hour virtual gathering.

During the Town Hall, three executives shared their stories of creating a data-positive culture within their media companies.

Grupo AM, Mexico

Enrique Gomez, chief operating officer at Grupo AM, said the company is moving to a reader revenue model next year and aligned three different projects with five major objectives to get them ready.

“Our goal was to help our newsroom become data lovers so they can better serve our brands, which is why the project was called CONAMOR, which means ‘with love’ in Spanish,” Gomez said.

The five goals of AM’s project are:

1. Change culture.

2. Boost loyalty.

3. Learn to love readers back.

4. Increase engagement.

5. Grow business.

Gomez says when they embarked on this journey, not many people in the newsroom “spoke data.” He often heard complaints that data is hard to find and harder to interpret. He also found not many people in the newsroom knew what their most valuable readers wanted and what needed to change to better serve them.

The solution? AM wanted to make learning about data easy for newsroom staff, so they built three tools to help change the mindset. 

“Let’s build reports that are easy to read and easy to understand,” Gomez said. “We want to make it easy for the newsroom to cater to brand lovers’ needs, and we want to make it easy for them and for the product team to test, learn, and improve.”

The first tool AM built was a weekly report. They send it to the entire newsroom to get them talking about their most loyal readers who they affectionately refer to as “amigos” and “amantes.” “Amigos” visit AM two to 15 times over the course of three days and “amantes” visit 15 times or more in that same timeframe.

They started out sending the reports on Fridays but have since changed that to see if they get better engagement within the newsroom on different days. Perhaps the biggest change to make these reports successful, Gomez said, was hiring a data analyst to gather insights and give them back to the newsroom so reporters and editors don’t have to interpret the data themselves.

“This person also has been running one-on-one sessions with the editors in the newsroom, so that’s also helping make it easy for them to ask questions and get answers instead of them having to go over analytics and look through the data,” Gomez said.

The second tool was creating a playbook that runs A/B testing. Gomez says it’s a simple Google slide that lives on the company’s shared drive. The basic outline of the playbook is, “If _____, then _____ because ______.”  They are testing newsletter subject lines now but Gomez admits it’s difficult to do this type of testing consistently. 

The third tool is live dashboards to understand the behaviour of key user segments. This helps them keep track of what’s happening with their brand lovers in terms of engagement, not just volume.

“Our goal is to move to a more sophisticated dashboard with metrics like active time on page, topics, and data by user segment in real time,” Gomez said. “We learned it’s more difficult to actually execute when you don't have a very sophisticated analytics platform, so it’s still a work in progress but we’re getting there.” 

One of the big lessons learned while creating these tools is that internal communication and branding helps with getting alignment across company teams, Gomez said.

He also realised some limitations: “You can do a lot with free and available resources, but we realised we have to make some investments.” 

He also realised reporters and editors aren’t meant to be tech wizards and encouraged newsrooms to ask for help: “We undervalued technical expertise.”

In the short term, AM wants editors and reporters to fall in love with their readers.

“We’ll be successful if we get more brand lovers. We had a goal of 3% brand lovers as a percentage of our total audience from 2.5% and we actually met that goal.”

AM also wants its editors to fall in love with testing.

In the medium term, once the paywall launches, AM executives hope that will translate into actually getting subscribers. They want the company to meet annual OKRs, and they will look for subscriber conversions and decreased churn, hoping their tools will be good enough to share with other newsrooms.

Grupo Octubre, Argentina

Mariano Blejman, chief digital and product officer at Grupo Octubre in Argentina, told INMA members that his team is working on predicting the success of headlines by using Artificial Intelligence.

“We wanted to create a tool that predicts the success of headlines and we will provide information to news editors so they can make decisions in real time,” Blejman said.

The original hypothesis was if you have two entities that usually show up together, you will have low probability of CTR. But if you have two entities that usually don’t show up together, they will probably go up in CTR. 

“The result of this was bad,” Blejman said. “Basically failure sometimes is an option.”

They had to rethink everything. They trained a model to look through two years of headlines then classified them into two groups. A headline was deemed a success when it would have a CTR better than 6%. It was deemed a failure when it had a CTR lower than 3%.

“With this approach, we achieved 76% precision in predicting classes with just the headlines,” Blejman said. “Then we went to proof of concept and asked the editors to try the headlines in real life.”

The team then looked at whether the headlines they predicted would do well actually did.

Blejman got valuable feedback from editors who said it was difficult to know what to do if they write a headline with low probability of success. 

“We thought about that and added some real time testing of results,” Blejman said. “When we try a title and the title isn’t working very well, we go to historical data and we say these are the last three successful stories that are using the same entity, the main entity that you are trying to solve. The other thing we did was to go to ChartBeat and look at the traffic coming from titles from Google and try to see in real time if we were using titles that were going to work in the next 48 hours.”

Two of the most valuable lessons learned were to fly low and to fall in love with the problem, not the solution.

“Start with something small you can solve in a very short period of time,” Blejman said. “If we were tied to the first idea, maybe we would have ended up not going anywhere. So we shifted the idea and we felt better with this idea.”

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