At the INMA Smart Data Initiative’s first meet-up in February, Yves Van Dooren of Mediahuis revealed its Belgian newsrooms turned to measuring and reporting aggregated time spent on articles by readers, as this metric was universally understood by editors and well accepted.

In the follow-up call, Van Dooren explained focusing on time instead of pageviews might help in reducing click-bait, while aggregating time spent by all readers on individual articles captured popularity of the articles, too. 

Mediahuis also found aggregated time spent on the site by individual readers correlated with the likelihood they converted to paid subscribers and renewed.

Benefits of measuring time

Research by news publishers, such as this Deep.BI modelling, confirms time is predictive for a reader subscribing or renewing. In comparison to other signals, time is more predictive than the number of pageviews but slightly less than, for example, visit frequency. 

Academic research, in general, confirms benefits of measuring online engagement with time: 

• Scientists at Yahoo Labs found it a good indicator for user interest.
• Researchers at Tsinghua University saw it as a proxy to for content quality. 

Time is also widely used in digital media beyond news. For example:

• At Facebook, time spent interacting with posts helps rank the News Feed.
• At Google, time spent on sites informs its search results.
• At Netflix and Spotify, play time guides content, product and marketing decisions. For example, a study by Caitlin Smallwood, vice-president/data science at Netflix, found the total hours spent watching was the most predictive for member retention, well ahead of movie or show ratings. 

Risks of measuring time 

“Although time is a popular measure,” admitted Mounia Lalmas-Roelleke, head of tech research at Spotify, in her seminal book “Measuring user engagement,” “there are significant drawbacks in using it.” 

• Tracking challenge: Popular software, such as Google Analytics, tracks reader behaviours on the browser side and that method has limits. For example, it doesn’t measure the time a reader spent viewing the last page of the visit. So, if the reader views just one page, the time reported is zero. It matters because on average, 67% of all visitors to news sites view just one page in a month.
• Observation challenge: Readers often open multiple articles simultaneously in their browsers and switch between tabs. It’s difficult to figure out which tab captured the reader’s attention.
• Analysis challenge: Time varies much across pages, sections, and sites. Research by Lalmas-Roelleke found time spent on reading depended on the type of a reader, e.g., whether she is a generalist reading everything or a specialist interested only in some topics. It depended on the type of a page or a section, e.g., readers typically spent less time on a sports section and more on a home page. And it depended also on the type of a task, e.g., whether the reader searched for something or just killed time browsing.

Best practices in time analytics

To mitigate the challenges of time measurement during single sessions, Lalmas-Roelleke recommended measuring engagement across multiple sessions — for example, the aggregated time spent by a reader in a month. A version of this is used by German regional publishers who collaborate on audience analytics. 

Another solution could be to analyse time in connection with other metrics, even the vilified pageviews: 

• For example, The Financial Times in the United Kingdom developed Quality Reads metric that measured pageviews qualified by the threshold of time and scroll depth. For a pageview to be counted as a Quality Read, the reader needed to spend at least 50% time required to read the whole article estimated by the number of words and scroll to at least 50% of the page’s length.
• Time could be also integrated into article or user scores used by journalists or to inform algorithms that curate home pages or trigger paywalls. For example, Die Welt in Germany and Dagens Nyheter in Sweden used time in their complex article scores presented on the newsrooms’ dashboards and regular reports. 

There are benefits in measuring not only engagement time but absence, too: 

• News sites want readers to visit repeatedly and develop a habit. Measuring absence time may trigger engagement actions, such as notifications or e-mails, as Aftenposten in Norway discovered.
• Measurement of absence time between visits helped Der Groene Amsterdammer in the Netherlands map readers’ paths to conversion and attribute influence of individual articles on the decisions to purchase a subscription.

What’s your North Star engagement metric? How did you choose one? What are the benefits and challenges? E-mail me at: grzegorz.piechota@inma.org.

From the editor: Welcome to the data-positive world!

With the new Smart Data Initiative, INMA aims to help transform the online news business and make journalism sustainable. In my new Smart Initiative Product newsletter, I will be sharing insights and best practices in democratising data and tying it to editorial and business objectives.

I address this newsletter to decision-makers, or simply data insights users, in the boards, newsrooms, marketing and sales, product, technology, research and — surely — data analytics departments themselves. 

Read more on the Smart Data initiative. Check the resources via the Initiative’s page. Sign up to this newsletter today and tell your colleagues about this new resource.

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