How The Atlantic, Weather Channel use data to design audience experiences

By Shelley Seale




There is more to premium news experience than quality content. And, with the right audience research and data, publishers can redefine subscriber and member experiences.

At the INMA Media Subscriptions Summit in New York today, three speakers from very different organisations shared key insights into doing this.

Emily Goligoski, head of audience research at The Atlantic, opened the session with a question: What do newsrooms have in common among vastly different types of news publishers?

Her answer? “They are all very rigorous in understanding what their audience knows, and this changes what they produce on a day-by-day basis.” 

That includes understanding all of the following: 

  • How your journalism can affect the lives of readers, viewers, and listeners.
  • How you can use your products to regularly deliver on their motivations.
  • How to orient them and be a trusted part of their week.
Emily Goligoski of The Atlantic says good audience research can help publishers understand a variety of questions about their readers.
Emily Goligoski of The Atlantic says good audience research can help publishers understand a variety of questions about their readers.

Goligoski  said publishers can take a lesson from the tech industry, which hires teams of researchers for continuous product feedback as they develop their product. “We can benefit from user experience research such as this.”

What will it take for more people around the world to find your content indispensible? These are the kinds of insights you might gain about people who benefit from your journalism.

“We’re thinking about their attitudes, what they think and need, and their behaviours. There’s a lot that we can be learning.”

When audience research is done during the planning phases of editorial, product, and marketing, it can provide insights during four stages:

  • Discovery: identify audience needs.
  • Early exploration: provide concept validation.
  • Ideation: co-design.
  • Testing and iteration: A/B testing, data science collaborations, and usability testing.

Improving the signal-to-noise ratio

“I do not underestimate the amount of inbound tweets, emails, and conversations that our reporters are contacted with about how wonderful or terrible their work is,” Goligoski said. “It’s really important for us to balance what it is we’re hearing. That’s why we undertake rigorous and disciplined research.”

She added that if publishers only paid attention to social media, they might think everything was wonderful. “It’s only when we go and are thoughtful and go deep in soliciting the opinions of people who aren’t regularly asked that we hear more valuable critique.”

How The Atlantic calibrates what it learns

“We are rigorously human-centred,” Goligoski said of the audience research team. “We exist to advocate for our readers and help the publication grow. We’re a complement to data science and market research. Sometimes a question can be answered by one of those alone, but very often we all need to be at the table to come up with the most informed responses.”

How The Atlantic seeks answers

Goligoski said they try maintaining flexibility by creating small, interdisciplinary teams. Key collaborators include editorial, product, design, data science, marketing, and advertising.

She shared the research methods her team uses, with the first three being the ones used most frequently.

  • Interviews.
  • Surveys.
  • Co-design exercises.
  • Concept testing.
  • Analogous research.
  • Usability testing.
  • Diary studies.
  • Ethnography and observation.
  • Conversation facilitation and design thinking.

“Interviews are good for more detailed information, while surveys are good if you have limited time or resources,” Goligoski said. “Co-design is thinking about news product prototypes we have that are incomplete, and they can help us complete what they should look like.”

How to talk to people

Language is important, Goligoski said. “I do think that the way we talk about the people we serve matters.” She suggested trying not to talk about people as users. “If I can talk about individuals, about people, we’re getting to human-to-human relationships faster.”

She also pointed out that publishers don’t just have one audience — they have many, made up of current, prospective, and cancelled audiences that they want to serve. Readers, listeners, and viewers are examples of the specific different types of audience members they have.

How (and who) The Atlantic recruits

Talking to people who used to pay for your work and stopped doing so can provide important insight and answers. If someone says they’re too busy, Goligoski said, never take that as an answer. Dig deeper for the reason. Is it payment? Something else?

The Atlantic uses a combination of a reader panel, friends and family, and third-party recruiting to study, made up of current active readers, former readers and subscribers who cancelled, and prospective audience members (growth audiences).

“In undertaking any kind of audience research, the results you gather will be a direct result of the care you put into how you recruit the people,” Goligoski said.

Why publishers should talk to people

Everyone has opinions, but listening to this feedback doesn’t mean the readers’ opinions will run the newsroom. “We need to consider that and balance it with editorial instincts.”

Goligoski shared four specific reasons why The Atlantic talks to people:

  • To gather frequent qualitative insights as they ask the question: Who are our readers and how can we better serve them?
  • To identify unmet reader, viewer, and listener needs.
  • To balance unsolicited audience feedback and who they are listening to.
  • Membership insights.

“We thought a lot about the difference between subscription and membership,” Goligoski said. “We see that there is a distinction between members and subscribers. Members give of themselves and it’s less of a transactional relationship.”

  • Paid subscribers get access to a product.
  • Participating members give time, energy, ideas, and money to join a cause because they believe in it.

At the heart of the membership puzzle is the contract between the site and its members. What do you give? What do you get?

One thing that is quite universal no matter a company’s business model, is that certain traits should stand out about the brand:

  • A sense of uniqueness.
  • Opportunities to contribute.
  • Transparency.
  • Appropriate prices and asks.
  • Brand design and visual appeal.

“We need to make ourselves more visible and less institutional,” Goligoski said in conclusion. “Being clear about our mission and caring about what matter to our audience.”

INMA members can learn more about this at The Membership Puzzle.

Charging in a world of commodity content

With a completely different audience at The Weather Channel, Neil Katz next addressed audience members. As the global head of content and subscriptions, Katz explained how The Weather Channel is building a premium subscription business in a world of free information.

“We face a situation where a lot of our core content is commoditised,” Katz said. He then joked that the audience was about to see the most stunning piece of technology they would see all week, and showed a clip from 1982, when The Weather Channel started.

Although the might seem vastly outdated and laughable now, he said, “This was unbelievable tech in 1982. We were delivering  hyper-local weather reports before that kind of tech really existed.”

The Weather Channel provides 15 billion forecasts per day to nearly every country in the world. But competition today is fiercer than ever before, and not just from other publishers, television broadcasts, or weather apps.

“We now have to compete with smart speakers for getting the weather, smart homes, smart refrigerators, smart cars, and even smart potted plants. The window is probably our biggest competitor globally. People can just look outside and see what the weather is like.”

Neil Katz of The Weather Channel talked about how to build subscription revenue in an era of free content.
Neil Katz of The Weather Channel talked about how to build subscription revenue in an era of free content.

So the question is, how can The Weather Channel charge for something that everyone can get for free so many other places? “It’s the most preposterous sort of business to figure out. We came up with a credo that we think makes sense and feels very natural to us.”

  • Basic weather data should always be free, globally available and easy to access. And The Weather Channel should be the largest purveyor of it.
  • Weather technology and insights are proprietary, provide unique value, and are worth money.

“Not everyone will pay, but a percentage will really value that additional experience.”

Katz said 95% of its audience may access The Weather Channel information for free, and they put their subscriber revenue on that 5% who will pay for the additional information.

“One other thing that’s really important to us is that we never will put something behind a paywall that will fail to keep you safe. Our prime directive is to keep people safe, so we’ll never put major weather events behind the paywall.”

Go where the most loyal audience is

The Weather Channel determined its most loyaly audience is mobile users. 60 million app users visit 28 days per month on average. “Our usage is through the roof in terms of loyalty.”

He compared it to a car — everyone who buys one will get an automobile with an engine and the basics to drive, but if they want the fancy wheels and turbo engine and sunroof they’re going to pay more for it.

“We tried to think about what parts of our offering we could turbo-charge.”

The answer to figuring out what those parts would be was to determine what the people using the product most valued. The pricing model then became:

  • Free: 48 hours of weather forecasting with a one-hour precision window.
  • TWC Premium: 96 hours of forecasting with 15-minute precision, two times the power, and four times the precision.
  • WU Premium: 360 hours of forecasting, hourly precision, and 7.5 times the power.

“Another core technology for us is radar,” Katz said. “We do have a tremendous amount of free radar for people to use. For example, lightning detection, which is free for a 10-mile radius, but premium is 30 miles.”


Personalisation to a user’s needs is another part of the premium paid version. This enables the user to choose from a variety of templates and match the forecast in relation to specific outdoor activities they may enjoy (photography, fitness, biking, fishing, etc.).

“We’ll alert you to the perfect time for the things you love,” Katz explained.

The premium version is also ad-free.

Katz said another important aspect was to create affordable price points. The premium offerings range from $0.99 per month to $19.99 per year.

Subscription results and churn

Katz reported that The Weather Channel has added almost 500,000 subscribers in nine months. “We’re adding, on average, about 50,000 per month, and that pace seems to be going pretty well.”

The premium users take advantage of a lot more of The Weather Channel services, but Katz said the team isn’t clear if that’s cause or effect. “Is it because we took the ads out and the app is more user friendly — or because people are really using it more?”

When it comes to churn, he freely shared those numbers, saying everyone has churn and it shouldn’t be a big secret. He admitted their numbers aren’t great.

“We are losing about 10% a month.” That means that a $10 subscription is really worth $6.

Another question unique to The Weather Channel was whether major events drive more subscriptions for the additional features. For example, Hurrican Dorian drove 30,000 new subscriptions in a week and a half.

“Does news drive subscriptions? We like to say Mother Nature is really our chief marketing officer and also our editor-in-chief. She writes the scripts and we kind of follow along.”

Katz also shared whether free trials worked for The Weather Channel. “The answer for us so far is no, which really surprises me.”

They did two free seven-day trial tests and saw an increase, but in the end it was sort of a wash. “We didn’t really lose anything or gain anything.”

Change is hard

Katz admitted that it is hard to go from a free service to charging a little bit of money, and they received a lot of negative feedback when they started charging.

“What really made the difference was including technology that was exclusive to us, rather than only taking away the ads.”

However, 95% of The Weather Channel content is free, so when they lose a subscriber, they don’t lose a viewer — only a subscriber. “This tells me that we don’t have enough exclusive technology,” Katz concluded.

Behaviour as a metric

With a very in-depth look at how to best utilise audience data, Steve Lock of M Matters took the stage. Lok, the co-founder and chief marketing technologist of the company, formerly worked at The Economist.

“Why I’m here is to talk to you about the fact that data is kind of failing us over time. We’ve invested in more data, but what is it actually telling us about our customers?”

Data has too many signals and not enough meaning, Lok said.

“We started M Matters to help senior executives bridge the gap between marketing, product, technology, and data. Ultimately, at the end of the day we’re challenged to deliver more relevant experiences to our audience. But what are they actually telling us inside that data that we aren’t seeing on the surface?”

Customer data powers the predictive eco-systm and gives context to data relationships. These help publishers better understand what that data actually means and understand their customers better.

The problem isn’t that publishers don’t have enough data — it’s that they have too much. And it often isn’t telling the most useful information about the customers as people.

Steve Lok of M Matters shared the most important benefits of understanding customers with a CDP.
Steve Lok of M Matters shared the most important benefits of understanding customers with a CDP.

“It’s not good enough to have a single customer view. That doesn’t tell you about that person who was a dedicated reader every day, and then just stopped coming for a few weeks. If you knew why, could you throw a campaign at them?”

Human beings are both logistical and emotional creatures, Lok said. Good Customer Data Programs (CDP) take this into account and do behavioural scoring, using behavioural data from multiple data sources.

“They do content affinity,” Lok said. “The CDP’s job is to take all these signals that exist outside and combine them to create a virtual personality. Good CDPs do data science.”

He shared the three KPIs publishers can impact with a good CDP:

  • Efficiency: Suppress people you are wasting money on.
  • Discovery: Find new behaviour patterns which lead people to buy.
  • Prediction: Spend more money on people who are ready to buy.

“Prediction, I think, is the thing that’s really interesting, because a lot of what we do now is based on responding to things that already happened. Sometimes that is a little too late,” Lok said.

A good CDP helps a publisher gain insights as its customers develop behaviours, not after.

“We need something that’s going to take all those signals and write context for them, so you can finally get to the place where you’ll be the most relevant thing imaginable to the customer you’re trying to reach.”

He led the audience through the three most important aspects of what a CDP does:

  • Builds first-party profiles over time using multiple qualified data sources.
  • Lets an organisation group profiles into meaningful audiences, not just segmentation.
  • Activates these audiences with marcomms. “It’s not worth its salt if it doesn’t take your audience and push them into usable segments and platforms.”

Just as important is what a CDP isn’t, Lok said.

  • It’s not a DMP (information without identity).
  • It’s not a BI/DW (information without activation).
  • It’s not a CRM (information without behaviour).
  • It’s not analytics (information without individuality).
  • It’s not a file – information without real-time.

A CDP, rather, combines all of these — individual behaviour, plus mapped data, plus real-time machine learning — to create self-optimising, activated audiences.

“There are a ton of opportunities for us if we do a good CDP,” Lok said. These include content marketing, loyalty, value exchange, and data/insights. “There are tons of things that are actionable if you ask the right kinds of questions.”

The point, he continued, is that the CDP is doing a level of identity resolution that is extremely difficult for people to do. But the way that it actually resolves identity is much more human-like.

“It’s a huge benefit for us to create products that are value exchanges. Having some information from a leading perspective can give you information before you even start testing.”

A CDP tries to predict a person’s willingness to do something you would like them to do.

Propensity paywalls

He suggested that publishers would benefit from something he called a flexible propensity paywall. In contrast to a hard paywall, a propensity paywall would:

  • Be fully customer-centric.
  • Provide opportunities for different types of value exchange.
  • Ecourage content consumption.
  • Match the best prospects with their favourite content.
  • Provide cross-functional ad revenue opportunities.

Lok proposed that it needn’t be a question of ad revenue or reader revenue: “Why not both?”

With such a model and a great CDP, a publisher could start to slice apart its Web site and personalise it. “When we looked at this as a test for The Economist, nearly three-fourths of visitors to the site never saw the paywall at all.”

With the capability to slice and dice the way a Web site reacts to the kinds of users who are coming for one particular type of content, a publisher can then look at monetising those things as different products.

“It takes martech to understand your content contextually,” Lok said. “To understand and profile individual reading behaviour, and push modeled audiences to your channels.”

He warned of being aware of organisational silos, which hide the bigger picture and are unable to bring the best of martech, adtech, and data insight leadership together.

So how is this done?

Lok said he may have oversimplified how this is done, but not terribly. “You can absolutely do this. You need something that creates a custom Web or mobile experience, understands how to build these models, and understands how to build real content itself. And it’ll learn over time.”

One of the things CDPs are really great for is helping publishers sustain a level of product innovation. “What can you do if you can really understand what people might do in the future, not just what they’ve done in the past?”

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.