Executives from three media companies shared data strategies and insights at the 2017 INMA World Congress on Monday.
“We’re very lucky at the South China Morning Post to be at the very start of this transformation,” said Chief Executive Officer Gary Liu.
The company recognises data as an asset to treasure, Liu said, but added that it is only an asset if it is taken advantage of. “Being able to capture and collect that data doesn’t do us any good if we don’t know how to utilise it,” he said.
Liu then gave a brief overview of the components driving SCMP’s data strategy. With a content management system (CMS), data warehouse (DWH), and data management platform (DMP), South China Morning Post can power algorithms and platforms. These algorithms help spot changing consumer behaviours and inform necessary changes.
“We actually need to change user experience of our news products,” most importantly using CMS, DWH, and DMP together powers machine learning, he said. “Machine learning means a lot of things, but the output is Artificial Intelligence.”
This makes it easier for computers to understand how data associates with one another. Machine learning can free up journalists to do the work they are meant to do, Liu said — adding that this holds governments responsible.
On the editorial side of operations, Thomson Reuters’ Reg Chua said machine learning and AI can have a big impact on process and outcome. As executive editor of both editorial operations, as well as data and innovation, this is something Chua is thinking about from multiple perspectives.
“The goal for me is to think through a very data-centric view of news that I think drives quality, it drives efficiency, it drives data,” he said.
Automation has proven benefits for speed and breadth of content creation, Chua noted. It can even support cost effectiveness, though he warns that companies should consider that cost effectiveness will be realised in the long term.
A big question asked by media and outsiders is if automation will replace human reporters. Chua’s opinion is that this is a stupid question.
“Things are good at what they’re good at,” he said. “You shouldn’t make cars do horse-type transportation, and you shouldn’t make horses do car-type transportation.”
Machines are good at two important things, Chua shared: computing (which is used for automation of insight) and scale (which powers personalisation and news-on-demand).
At Thomson Reuters, using a machine to write a story produces an article filled with statistics and trends. A reporter then pulls some of that information to use in their own work. A machine cannot decide the most important fact in the context of the story, Chua said. This is where a real person is crucial.
“The real value here is the marriage between machine and people,” he said.
Automation can also be used to create more personalised news experiences. Automation pulls in information, like stock figures, and can also report the status of a consumer’s stocks — perhaps even what may have happened if they had not sold a certain stock a few days prior.
“That’s what machines can do,” Chua saud. “They can explore all the counterfactuals.”
Using automation to create personalised news-on-demand can also give consumers updated news when they want it, no matter where they are. A consumer can tune into stock reports while driving, or read about it on their phone while standing in a long line.
“That gives you the ability in theory at least to build real engagement with readers,” said Chua, “because you’re giving them something they’re interested in.”
Overall, he said, machine learning and automation can help news media companies evolve with consumers’ changing news habits. New workflows, specifically in revisiting data and structure, creating competitive advantage, and pushing ideas, can push companies along the right path.
“The whole point of this is that there is a fundamental shift in what news is,” Chua said.
Before the machine age, publishers wrote a story to make it relevant to as many people as possible. Now, Chua said, they can create news the way people want it: “This is a way of driving engagement.”
At The New York Times, machine learning is being used internally, said Laura Evans, senior vice president of data and insights: “We’re using machine learning to advance our understanding and improve efficiency.”
Algorithms are used to fill in the gaps by identifying cross-topic opportunities in content. Audience data is also used to create content segments based on interest.
These continual cross-references are valuable to the company, Evans said. “What you can create with that is what we call sort of an unlimited query engine.”
This allows the company to create opportunities for itself to reach new people, and to keep improving itself at the same time, Evans said: “That’s the power of machine learning — being able to bring these things together and have that constantly adding to itself as you add more and more topics.”
At New York Times’ Slack, the Blossom bot helps the company internally identify and promote articles that may go viral by identifying trends on its own Web site and also on external Web sites, such as Facebook. By telling the team when to strategically post a specific article to Facebook, the content has a better chance of reaching more people.
“Our strategy is really to create an efficient approach to data,” Evans said.
The data opportunities, both internal and external, can be invaluable for media companies — but sometimes a question of resources can leave publishers unsure whether they should invest on building internal structures, or look to existing platforms and capabilities to power data strategies.
For Liu, the “buy vs. build” question does not mean that both cannot happen at the same time. “I think that likely for most news organisations, it’s going to be parallel: buy versus build.”
The first step is to consider what will get your company there the fastest; but then each company must determine what will work best in the long-term. The key to understanding this, Chua said, is to understand what is possible to do internally.
Evans added that the parallel path is a learning opportunity for a media company’s internal data team. “I think a lot of people in the data field have learned a lot from the parallel path.” She recommends storing data to refer to later and capturing what you can as fast as you can.
From a newsroom perspective, Chua said, size does not matter. A company does not need huge engines or access to huge chunks of data. “You can actually start generating this stuff with very little effort.”