Data offers tremendous benefits to news media companies, but one area that’s too often overlooked is its importance in creating a more customer-centric business.

“In journalism, we use the five W’s,” Riske Betten, digital director at Mediahuis Nederland, told attendees of the INMA/Meta Audience Analytics Town Hall on Tuesday. “If you leave one of those out of your article, it might be incomplete. But a colleague shared we also have five W’s in data.”

Those five W’s are: why, why, why, why, and why.

By asking that question more often, media companies can get to the bottom of their problems, such as a technology hiccup, and find ways to please the customer.

Data is also useful in looking behind goals and finding the appropriate drivers and influencers for each one. Then, there are indicators that can be identified to confirm that all these elements have fallen in line.

“These are the targets you will hit if you have everything in place. When it comes to using data, you need to get a grip on the chain of elements.”

Building a customer-centric business depends on listening to what the customer says they need. It can be done in many different ways, Betten said: digitally, through testing, or through face-to-face conversations. How that listening is done is less important than just doing it: “Thinking you know your customer is the biggest downfall because you are not your customer.”

 One way to become more customer-centric through data is to make sure the data you’re collecting is telling a story to the teams that are analysing it, because “that’s what journalists like, and if you’re just gathering figures, there’s no guarantee they’ll use it.”

Make the data compelling to them, and they will use it to help transform the business.

“It comes down to getting more information about your customer,” Betten said. “Whether that’s using algorithms, A/B tests, user research, or through conversations. Start getting closer to your customer and keep in mind the customer is why we are doing these things. We want them to be satisfied.”  

As publishers shift from content to audience analytics, they learn new data skills such as segmenting readers to tailor products and marketing. They learn how to predict their behaviours, such as subscriber churn. Increasingly, publishers run-data informed experiments to test their ideas. 

As publishers shift from content to audience analytics, they learn new data skills such as segmenting readers to tailor products and marketing. They learn how to predict their behaviours, such as subscriber churn. Increasingly, publishers run-data informed experiments to test their ideas. 

Three news media executives shared their company’s experiences with this.

Editora Globo, Brazil

Alexandre Cordeiro, digital strategy manager at Editora Globo in Brazil, opened with the goal of his team: to define their main business KPIs and share them to all stakeholders, through a unique and easy-to-use dashboard that will ultimately facilitate the decision-making process.

“At the end of the day, we’re trying to transform that into actionable insights,” he told INMA members.

He acknowledged there were several challenges that must be addressed, such as how to evolve the way the team generates intelligence through data and how to transform that into actionable insights. The chances of achieving those goals will be increased by delivering:

• A standardised set of KPIs that align with corporate goals.

• A method or template to follow up on tests and their key findings.

• An easy-to-understand dashboard.

• A focus on execution.

• Automated dashboard that’s as close to real-time as possible.

• No human interaction dependencies, if possible.

“The focus on execution should be in bold,” Cordeiro said, stressing its importance. “100% automated and machine based is a difficult thing to achieve, but that’s our goal.”

Cordeiro shared Editora Globo’s user journey. “It represents a sense of movement, from an unknown user to experimentation.”

It then moves into the log-in and free user phase, converting into subscribers, and then churn and dormant users. The team must then learn and interact with users, and transform the data into information and actionable insights. 

“We’ve tried to come up with a dashboard that’s not only easy to understand and beautiful, but something that can activate and transform that into action,” Cordeiro said. “This dashboard represents what we call actionable KPIs.”

The team looks at the dashboard on a weekly basis and takes action on it, informing their decision-making process.

Cordeiro shared the lessons his team learned through this process so far:

• Ideation vs. execution: There’s the planning mode, but execution is more important.

• Top-down approval: Make sure the C-level is on board.

• Validation with stakeholders: The data must be validated and resonate with colleagues.

• Simplification is key: If you get too complex, the process and ideas will suffer.

• Prioritisation with other areas: Have an overview of the priorities of other departments, as you’ll count on them to accomplish these goals.

“We’re still in the works on making that happen,” Cordeiro said. “There’s the ideal world versus the real world when it comes to timelines, so make sure you understand that.”

Grupo Reforma, Mexico

David Hinojosa Aguirre, deputy director of IT research and development at Grupo Reforma, told INMA members that company went through the Facebook Accelerator programme to understand why users churn.

“With that information, we think we can make more informed decisions to achieve increasing subscribers by 15% in the next year, and 60% by 2025.”

To do that, Grupo Reforma developed a hypothesis: Developing quantitative and qualitative tools will lead them to know the reasons for subscriber churn. These tools would generate ideas and lead the team to make decisions that would transform their content and the way they present it to users. They would also help them find out how relevant their offers are to a user when deciding whether to buy or cancel.

“Based on that hypothesis, we developed the project, which was to build two major tools,” Aguirre said. These were a quantitative tool and a qualitative tool.

• The quantitative tool consisted of building a data model based on the navigation and conversion from subscribers who churned. Using this tool, the team could discover the behaviour patterns before churning.

• A qualitative tool would help them find out directly from former subscribers why they had churned. They could organise, categorise, and count the factors, and define strategies and actions to reduce the churn.

For the quantitative tool, Aguirre explained that the team defined four different milestones:

• Prepare, process, and analyse. This meant gathering browser variables such as frequency, content consumed, time spent, etc.

• Define an engagement index formula that would reveal the relationship between engagement and churn.

• Share this information throughout the team via engagement index dashboards.

• Take actions on the data, performing continuous A/B testing and measuring those results. Also, constantly improving the data model.

“We tried different formulas and variables, but we finally found a formula that described this engagement index,” Aguirre said. “The higher the engagement index, the lower the propensity to churn.”

The engagement index formula created a red zone, and if any user was in that zone, the team needed to take action to prevent the user from churning.

The qualitative tool also had four milestones:

• Find out why former subscribers churned, using a survey. Former subscribers were given 14 days of free access if they completed the survey.

• Use this information for product and UX improvements.

• Analyse browsing behaviour before churn for interesting insights.

• Implement the tools into the team’s daily process for insights that could define strategies and actions.

“We want to know why they churn,” Aguirre said. “We had a very good open rate [on the survey] and a good acceptance rate of the offer, approximately 70%.”

Former subscribers gave Grupo Reforma a lot of product improvement and UX suggestions through the surveys, along with the reasons why they had churned. “They gave us a lot of insight and we correlated that with the browser behaviour we had from the quantitative tool.”

This led to a lot of strategic tactics the team implemented to reduce churn and retain subscribers.

“It’s a continuous loop,” Aguirre said. “With that, we realised that our hypothesis was right. We can make more informed decisions that can lead us to our goals.”

The democratisation of data is important for the team to make those informed decisions. Aguirre shared the learnings his team acquired from the accelerator:

• The relevance of a data-driven decision workflow.

• Focus decisions on users and not exclusively on content.

• Knowing what users do, and how and why they do it, is essential.

“We have to take a look at the data, but get insights from them and apply them to strategies and decisions,” Aguirre said. “We have to focus on the user and know the users.”

He shared his team’s measures of success and next steps:

• Identify the most important reasons for churn and user behaviour, and take actions.

• Achieve a change in Grupo Reforma subscription and churn rates that reflect more user engagement.

• Use the new dashboard as a basis for decision-making that will lead to the goal of a 15% increase in subscriptions by October 30, 2022.

Gazeta do Povo, Brazil 

As a smaller publisher with fewer resources, Gazeta do Povo provided another interesting case study. Guilherme Vieira, director of business, technology, and operations at the company, shared his team’s journey with INMA members.

“In the last couple of years, we learned how to sell a lot of subscriptions,” Vieira said. “Our main strategy is selling subscriptions, but we are struggling with our churn rate. It’s about 3.8% and our goal is to reach 2.5.”

Like Grupo Reforma, the Gazeta do Povo team came up with a hypothesis: to identify subscribers with a propensity to churn through data.

“To do that, we should develop a propensity model through data collecting about the subscribers,” Vieira explained. “We decided to build a predictive model.”

That model was the first target for the team. The second target was to use that score to drive marketing efforts to prevent or reduce churn rates by 10%.

They created a special team and a timeline to implement and achieve this. There were several steps involved: 

• Step 1: profile definition, data samples, and features map.

• Step 2: predictive model development.

• Step 3: data collection.

• Step 4: model implementation and dashboard.

The model considered several data sources: subscriber data, subscription data, marketing, and content access.

“It’s dozens of data [points] to determine which one contributes more,” Vieira said.

To test the hypothesis, the team separated a training group from a testing group, and ran the algorithm with the training group that consisted of 70% of the total. This allowed them to test if the variables were working well.

They discovered that more recent subscribers and those who visited the Web site less frequently had the highest propensity to churn. 

“The big learn is: You have to identify each subscriber and establish the right communication with him,” Vieira said.

Using all this data, the team then assigned a churn propensity score for each subscriber, and this proved to have a 67% accuracy.

Vieira shared the next steps his team plans to take:

• Model automation.

• Marketing and subscriber touch point tools integration (Salesforce, etc.).

• Build a health subscriber portfolio dashboard.

• Model enrichment and improvement by testing new variables.

• Increasing engagement and churn prevention action plans.

“We want to integrate the tools and build a healthy subscriber portfolio dashboard,” Vieira said. “We want to perhaps reach a higher propensity score. We haven’t used the score to prevent churn yet, but for the next year, that’s one of our main goals.”

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