Analytics: From Raw Data to Actionable Insights

Digital interactions generate large quantities of data reflecting the behavior and preferences of online consumers and audiences. How do news organizations use this data to maximize engagement and revenues?

If there’s one thing there is no shortage of these days, it’s raw data. This is especially true for companies in the business of creating content. Newsrooms and publishers know how important it is to promote content across a wide variety of digital formats— social media, podcasts, apps, e-newsletters, etc. — but more importantly, they know the value of the data resulting from these promotional channels; namely, analytics reflecting user behavior.

This data is potentially a powerful tool for iterating and improving content and audience engagement. However, the way it is aggregated varies across platforms, meaning that it is often not easily accessible or usable. This has driven a need for a framework with which to process and analyze the raw data available in order to create what have been called ‘actionable insights’.

What are actionable insights?

As defined by TechTarget, actionable insights “are conclusions drawn from data that can be turned directly into an action or a response..” Ruler Analytics adds an additional distinction: “Actionable insights have one main characteristic. They drive action that leads to results.”

The nature of actionable insights varies across industries, but often involves performance marketing, user insight, or problem-solving. In the media industry, the focus of actionable insights is largely on improving editorial processes with data-driven recommendations.

Though data analytics has historically relied on human interpretation, artificial intelligence (AI) and machine learning (ML) algorithms are quickly becoming a standard best practice for generating actionable insights. But, as TechTarget is quick to point out, “While ML algorithms offer great insights, they're fully dependent on the data they're fed during their learning and maturity stages.”

So what does it take for raw data to become an actionable insight? Let’s take a closer look.

Turning data into actionable insights

Though the specific goals and objectives of data analytics may vary across industries, the process of turning data into actionable insights follows the same basic trajectory:

• Prepare unstructured data — Raw data can be structured or unstructured. Structured data is presented in a traditional, easy-to-navigate tabular format that an algorithm has no trouble processing.
• But unstructured data is a different story. TechTarget reports an estimated 80% of all available data is unstructured. Common unstructured data formats include “raw text, social media comments, log files and call transcripts.” To effectively prepare and process unstructured data types, it’s essential to pick a platform that can integrate with multiple sources and data types.
• Test hypotheses — “Many organizations measure data and then wander through the results looking for some magical insight,” reflects TechTarget. “Establishing a hypothesis and testing it for accuracy will deliver actionable insights much faster.”
• Integrate insights — Once data analysis has been performed and a hypothesis is proven or disproven, it’s time to report the results and integrate any actionable insights into workflows and processes. Ruler Analytics recommends using visual aids like charts and graphics to articulate the data findings and illustrate the value of leveraging the resulting actionable insight.

Examples of data-driven insights in newsrooms and publishers

Now that we have an understanding of the process, let’s see how actionable insights are put to use in news media and publishing.

The Financial Times

For most media companies, the success of content hinges on reader engagement. Testing the assumption that the more engaged a reader, the less likely they are to cancel their subscription, the Financial Times extrapolated a key performance indicator (KPI) by which to measure success and predict future outcomes: recency x frequency x volume (RFV).

In conversation with the World Association of News Publishers, McKinley Muir Hydén, the Financial Times’ Lead Data Analyst of Audience Engagement, explained RFV “looks over the last 90 days to see how recently a reader visited us, how many times they visited, and how much they read over the period. We also have a range of metrics that try to balance each other, including pageviews, scroll rate, bounce rate and time on page, and how many people went on to read another article. RFV pretty much powers our commercial side. We actively report on it, and we use it to qualify initiatives and testing. We also have further models based on it, and it works for advertising, B2B marketing, and so on.” Into The Minds reports the RFV KPI is so precise, it can actually predict cancellation rates at the Financial Times.

The BBC

Historically, news media has been in the business of updating people on important information. But in today’s extremely online culture, consumers have come to expect more from their content.

While working at the BBC, Dmitry Shishkin was one of the first to champion six “user needs” newsrooms should consider. The BBC editorial team analyzed content output and page view data, and found “Update me” content — aka breaking news bulletins, summaries, and live updates — constituted 70% of the content being generated, but only contributed to 7% of all page views. To correct this imbalance, Shiskin helped the BBC begin incorporating additional user needs into the editorial pipeline, such as “Educate me” content designed to provide users with helpful instruction, “Inspire me” content to uplift users, and short-form “Divert me” content to provide some distraction.

And just as data revealed actionable insights on BBC users’ needs, so does it continue to allow the BBC to monitor the success of its content and iterate accordingly. In speaking to Em Kuntze for Medium, Shiskin advised publications analyze at least three months of content performance data before implementing the user needs model, and then closely monitoring the results to develop even deeper insights into what users need, and how to deliver it.