News media companies are working to balance the use of data to improve customer experience while respecting data privacy to strengthen audience relationships.
“We have to see the paradox that users expect us to know them, optimise for their personal preference, and respect their privacy,” INMA Smart Data Initiative Lead Ariane Bernard said. “We should appreciate that users do want to trust us with enough data that we can recognise them: They take actions like following authors or build recipe collections.”
Media leaders from Dow Jones, Datadog, The New York Times, and The Washington Post shared how they are using data to create personalisation and content recommendation features while building in structure and rules to preserve data privacy during INMA’s recent Transforming What We Build Using Data Master Class.
Personalisation has become essential for news media companies that want to get the attention of their customers. Dan Kent, director of product/personalisation at Dow Jones, said comapanies are no longer competing with other news media sources. In an age of pings, dings, and breaking news alerts, what companies are really competing for is people’s attention.
“Gone are the days where general alerts and random blasts are good enough,” he said. “We must be timely, specific strategic, and personal.”
Personalisation is all about relevancy, and readers today are demanding it, Kent said. In fact, a study by Accenture found 83% of consumers are willing to share their data if it means it will help create a more personalised experience.
That is being seen amongst Dow Jones brands, Kent said. A card asking readers about their special interests is the step of the onboarding process that has the highest level of engagement: “This demonstrates that they’ll invest some of their time and effort into it. And we need to reciprocate.”
Building a personalisation algorithm into a subscription model can yield amazing results for the users’ experience. There are many ways to approach user personalisation, and Erica Greene, senior engineering manager at Datadog, who shared three key features of a building personalised system.
“It is important to keep these in mind when evaluating and building your system,” she said.
Greene then explained a key concept in personalised systems is deciding on what she called a “candidate set,” or a box of things the system can choose from to recommend.
“It is important to have designed guardrails to keep personalisation accurate and create better user experience,” she said. “Doing this is more of an art than a science.”
With more than 150 pieces of content published daily to The New York Times’ Web site and an average of 70 slots on the homepage for readers to view stories, it’s easy for readers to miss a story they might be interested in. Even avid readers can miss stories when the homepage refreshes — which it does multiple times a day — and the problem is even greater for readers using smaller devices such as mobile phones.
“We think that using algorithms as one component of our promotion strategy — and when used thoughtfully can really help our readers discover relevant and useful content across our products,” Anna Coenen, director of data science at The New York Times, said. “And we also think that … they can be used to amplify our editorial judgment and not necessarily compete with it.”
Because user needs vary, it’s important to take a step back and look at all the possible use cases.
“The recommendation requirements are very different, depending on what product surface we’re talking about,” Coenen said. “So every time we’re thinking about adding recommendations to a particular part of the product, we’re thinking about what user needs we’re trying to address.”
The realisation that many of the technologies they had become dependent on were going away led The Washington Post to create more privacy-focused teams and look at how to build products with a more private future in mind.
“It means building not just new products, but reconsidering how we approach planning, architecture, and engineering of those new products,” Aram Zucker-Scharff, engineering lead/privacy and security compliance, said.
The team started building out early development practices and looked at such things as how the product will use user data, the best solution for storing the data, and creating a data checkout and deletion process.
Mid- and post-development practices included self-assessment processes, which gives teams ways to evaluate themselves and their projects at various stages. It also includes testing tools and walk-through documentation to make sure they are accomplishing company privacy goals and that it fits into the overall process.
“We’re looking at how we leak data to third parties and how we can minimise that,” Zucker-Scharff said. “We want to keep that data in. Because the more we can build a wall around our own user data, not only are we more private and more secure, but the more valuable our user data becomes as part of our marketing and ad sales efforts.”