Hello! This is Smart Data, a newsletter for INMA members on creating value with data analytics for media companies.
I am INMA’s researcher-in-residence. E-mail me at: email@example.com.
DATA-POSITIVE CULTURE. Concessions to journalistic autonomy help introduce data to newsrooms
“When it comes to creative people, you’d better ask how to use data to free yourself from doing things that you don’t want to do rather than trying to make data do the job for you.”
New insight: Last week, at the INMA Product and Data For Media Summit, author Caroline Carruthers was asked about how to help form a data-positive culture in the newsroom.
Across the world, newsrooms have been adopting data and metrics to make the work of journalists more effective and efficient. The change has been accelerated by:
The economic pressures on efficiency.
Effectiveness-boosting management ideas such as customer-centricity.
Advances in technologies, making work in the newsrooms easier to measure.
Still, the pace of the change varies due to the newsroom’s managers and staff’s scepticism, shortage of skills, or internal processes and resources. Sceptics believe knowledge and creative work requires autonomy to produce quality outputs, and worry that metrics and algorithms undermine professional judgment and artistic creativity.
“A lot of times people don’t like data because they don’t understand it, or we turn it into something big and scary. It is especially true with very creative people,” said Carruthers, co-author of The Chief Data Officer’s Playbook with Peter Jackson.
“The best way is to use data to free these people from the mundane and boring, so they can actually use their talent and skill in the best way possible,” she suggested.
The challenge: In a newly published book All the News That’s Fit to Click, Rutgers University Professor Caitilin Petre argued that metrics are a form of managerial surveillance and discipline. Their adoption in recent decades followed the principles of scientific management, or factory work optimisation techniques, as imagined 100 years ago by an engineer Frederick Winslow Taylor.
In the worst-case scenario, Professor Petre found, metrics could facilitate the regime in which “journalists are reduced from expert arbiters of news-worthiness to mere executors tasked with unquestioningly following the dictates of quantified representations of audience popularity.”
Are modern newsrooms really like factories? There’s a difference. “Journalists possess professional status, ample reserves of cultural capital, and a highly visible public platform — resources they can mobilise to resist metrics-driven performance evaluation if they choose to do so,” observed Professor Petre.
To avoid resistance, she reported, data initiatives and products gained journalists’ trust and acceptance by concessions to journalist’s autonomy.
Inspiration: At the INMA Summit, Caroline Carruthers pointed to international efforts to develop COVID vaccines as an example of a recommended approach.
U.S.-based Moderna used data and machine learning to accelerate the pace of research on proteins that train a body to respond to the virus. Automation helped them increase the number of tests of different sequences of the mRNA proteins from about 30 a month to about 1,000 a month without significantly more resources.
“Humans are really good at creativity and flexibility and insight, whereas machines are really good at precision and giving the exact same result every single time and doing it at scale and speed,” explained Moderna’s Chief Data Officer Dave Johnson in an interview with the M.I.T. Sloan Management Review.
“What we find [to be] the most successful projects are where we kind of put the two together — have the machine do the parts of the job that it’s good at [and] let the humans take over for the rest of that,” said Johnson.
Case studies: News publishers ran initiatives that increased efficiency and effectiveness of the work of journalists, while preserving their autonomy.
Content reviews helped spot what’s not worth doing: Funke Media Group in Germany analysed supply and demand of articles to help identify topics for which “we generate a lot of content which are not well received by our readers.”
Algorithmic text generation expanded coverage saving reporters’ time: NDC Media Group in Sweden automated generation of reports from 60,000 local football matches, using data on results, pictures, and quotes crowdsourced from coaches.
Automated curation boosted effectiveness saving editors’ time: The Globe and Mail in Canada handed over curation of its home page and social channels to an AI system so its editors could focus on finding and telling stories.
The bottom-line: Don’t use data to tell journalists what to do, but rather show them what is not worth doing. Automate the mundane and boring tasks to free their time and energy, and use their judgment and creativity to do things machines cannot.
Caitlin Petre, All the News That’s Fit to Click: How Metrics Are Transforming the Work of Journalists, Princeton University Press 2021.
NEW METRICS. How U.K. start-ups go beyond engagement and measure impact of journalism
Sifted, an U.K.-based news publisher covering European tech, uses machine learning to quantify and maximise its readers’ knowledge about fintech, venture capital, and other subjects.
How it started: Backed by the Financial Times, Sifted has been reporting on the U.K. and European start-ups since 2017. It reaches 1 million unique visitors monthly, offering a mix of breaking news and in-depth articles for paying members.
Sifted measures engagement with a conventional set of metrics such as RFV scores for users, and pageviews, dwell time, and scroll depth for articles.
“These metrics look at our readers’ experience from the publisher’s point of view. We wondered how to measure the value readers get from reading,” said CEO Caspar Woolley in an interview with INMA.
What is new: Last winter, Sifted teamed up with Crux, a U.K.-based tech start-up, and launched the Knowledge Tracker feature.
Scoring articles: Crux mines Sifted’s articles using Natural Language Processing to score their relative importance within a topic and the amount of new insights they deliver.
Scoring users: It tracks the reading history of each user to estimate her knowledge level in any given topic and gains from reading additional articles.
Recommender system: It recommends new and old articles, ranked based on their knowledge score, so the user can maximise the value of time spent reading.
“I like the gamification element: People can see their score increasing while reading and how any new article on the topic decreases the score,” said Woolley. “The next step will be a dashboard for users, so they can see scores for all favourite topics and perhaps even compare with friends or teams.”
Why it matters: According to Woolley, the knowledge-based recommender system increased the number of articles viewed by readers by 55%. Those exposed to the Knowledge Tracker also registered at a higher rate, and the uplift was 16%.
Assessing the impact of journalism is a widely debated subject in itself. Although many agree the purpose of journalism is to have an impact, they agree less on how to measure it.
The work of journalists is the easiest to measure by activities and outputs.
The quality of journalists’ outputs is trickier to measure and includes checks of adhering to the craft standards, metrics of user engagement, and recognition by peers, e.g., citations and awards.
The impact of journalism on individuals, society, and institutions is the hardest to quantify.
Measuring individuals’ gains in knowledge and attitudes, e.g., with surveys measuring awareness, understanding or their well-being before and after the news. Sifted with Crux found a way to automate such measurement for readers, although without considering effects of other sources of information.
Measuring actions by stakeholders, e.g., affected individuals signing a petition or making a donation, or decision makers who influence the status quo, e.g., making statements, changing decisions in individual cases or changing policies. Some of these can be measured with media monitoring tools.
The big picture: Sifted’s new knowledge-tracking metric is an example of a broader trend in business, in which data leads to new revenue models.
In a 2020 book The Ends Game, two marketing professors, Marco Bertini and Oded Koenigsberg, described how smart companies stopped selling products (“the means”) and started selling value instead (“the ends”). They asked: “Would you rather pay for health care or better health? School or education? Car or transportation?”
Professors Bertini and Koenigsberg broke down revenue models by the stage of a customer journey that companies charged for and analysed how much value customers derived from purchases at each stage.
Access models: Companies charge for “the means,” their physical products and services, and a promise “the ends” customers desire will follow. For example, an advertiser buys an ad in a printed newspaper.
Consumption models: Companies still charge for “the means,” but differentiate prices based on actual use. For example, one pays only for ads viewed or clicked by a consumer.
Performance models: Companies charge for “the ends” or outcomes. For example, an advertiser pays only for the sales attributed to an ad.
What’s next: Professors Bertini and Koenigsberg saw the new consumption and performance models rise as a direct result of digitisation. Easier than ever, companies can collect, analyse, and interpret data on when and how consumers use products and services and on how well these offerings actually perform.
In the context of reader revenue, newspaper publishers indeed shifted from selling access to printed goods to selling news online as a service.
Many subscription models are based on metering use of content or features. All leading publishers track engagement, and they see conversion and retention is mostly an outcome of engagement.
While Sifted has found a way to measure gains in knowledge of its readers, its paywall is still triggered by a simple article count and pricing is based on access.
Bottom-line: The case of Sifted in the U.K. showed how advances in data analytics led to a discovery of new metrics, helping to assess the impact of news products. In other industries, the impact metrics led to innovative revenue models.
Want to learn more? Sifted’s CEO Caspar Woolley is speaking at the INMA Product and Data For Media Summit on Thursday, October 14.
- Marco Bertini, Oded Koenigsberg, The Ends Game: How Smart Companies Stop Selling Products and Start Delivering Value, M.I.T. Press 2020
FROM OUR SLACK: Best reads about applications for machine learning and AI
Every day, on the Slack channel for the Smart Data initiative, I post links to articles that showcase the best practices and inspire applications of data.
Here’s a selection of the last week’s reads:
AI strategy: “Without data, AI can’t exist and with poorly managed data, AI won’t thrive,” said Schibsted’s Agnes Stenbom at the INMA Smart Data Master Class, during which she outlined the building blocks of the group’s data strategy and a separate AI strategy.
Use cases of AI in journalism: The New York Times’ Marc Lavalle discussed applications of AI technologies, such as natural language processing and computer vision, to tell stories better.
Use cases of AI in marketing: Kevine Payne of The Black Enterprise magazine summarised the ways AI tools can help get and keep customers, e.g., by qualifying leads, improving check-out user experience, or automating customer service.
Use cases of AI in sales: At last month’s INMA Advertising Master Class, News Corp. Australia’s Suzie Cardwell, South China Morning Post’s Ian Hocking, and Times Internet’s Partha Sinha told how they improved reporting ad campaign’s results to advertisers.
About this newsletter
Today’s newsletter is written by Greg Piechota, researcher-in-residence at INMA, based in Oxford, England. Here I share insights and best practices on creating value with data analytics and incorporating a data-positive culture.
This newsletter is a public face of the Smart Data initiative by INMA, outlined here.
E-mail me at firstname.lastname@example.org with thoughts, suggestions, and questions. Sign up to our Slack channel.