Meredith Digital: Entering a new era of data-driven publishing

About Meredith Corporation

Meredith Corporation has been committed to service journalism for 115 years. Today, Meredith uses multiple distribution platforms, including broadcast television, digital, print, mobile, and video, to provide trusted content and experiences on a wide-range of topics that include entertainment, style, food, home, lifestyle, health, wellness, travel, and luxury, and to deliver messages from its advertising and marketing partners.

Industries: Media & Entertainment
Location: United States

As its volume of content explodes, Meredith Digital is using machine learning to automate content classification, applying a custom universal taxonomy with Cloud AutoML and Cloud Natural Language.

Google Cloud Results

  • Helps build readership and loyalty by enabling more relevant and engaging audience experiences
  • Classifies text content across entire portfolio of media properties in months versus years
  • Gains greater awareness of new trends and customer interests

Matches human teams on accuracy, precision, and recall

With renowned brands such as PEOPLE, Allrecipes, InStyle, Better Homes & Gardens, Martha Stewart Living, and Food & Wine, Meredith Corporation is a leader in creating content and experiences across media platforms and life stages in key consumer areas of interest. One of the largest brand powered media companies in the world, Meredith has successfully grown its portfolio during more than a decade of digital transformation.

From 2010 to 2018, magazine readership across Meredith brands has grown 5 percent while the company has kept up to speed with evolving readership and content consumption trends for its more than 40 brands. It has done so by not only listening closely to what readers have to say and are interested in, but also by creating a culture of inclusion and shared perspective within the company.

“Every voice in the company is a voice of thought leadership,” says Grace Preyapongpisan, Vice President Business Intelligence at Meredith Digital. “In the name of driving incredible consumer experiences, I’ve seen our company take what’s working within small and forward-thinking groups and scale those practices across the entire company.”

Today Meredith’s National Media Group reaches more than 175 million American consumers every month across channels, including print, digital, social, and TV, reaching 80 percent of U.S. millennial women. Meredith Digital’s powerful brands deliver valued content to millions of people online, reaching 140 million online adults in the United States each month, or more than half of the U.S. adult online population. Its celebrity and entertainment sites reach nearly 70 million unique visitors a month, followed by food and home at nearly 60 million, beauty and fashion at 15 million, and parenting at nearly 10 million.

Content classification that enables advanced personalization by connecting relevant terms and topics is one way that some of Meredith’s digital brands innovate to create great customer experiences. For example, at Allrecipes, a food-focused, user-submitted recipe and social networking site, the team developed and maintains a robust custom taxonomy for content classification using a controlled vocabulary of standardized terms that helps monitor food trends.

For example, one of the nodes in Allrecipes’ food taxonomy is Cooking and Serving Style, with child nodes such as One-Dish Meals, On the Go, and Family Friendly. Food editors wanted to build a collection and monitor the trend around “bowls,” so they created a child node under Cooking and Serving Style for dishes served “In A Bowl.” The team classified relevant content, published relevant recipes, and monitored the results. Currently, recipes classified as “In A Bowl” are among the fastest growing, even though the number of recipes is still relatively small.

“Cloud AutoML Natural language met our needs more than other solutions we considered. We can better identify and respond to emerging content trends, creating more relevant, higher-impact audience experiences.”

Grace Preyapongpisan, VP, Business Intelligence, Meredith Digital

This taxonomy has proven so effective at providing a consistent set of content terms and identifying relevant trends that it’s being universally extended across Meredith’s brands. But scaling at this magnitude demanded much greater efficiency. For Allrecipes, teams of people manually classified recipes and reviews to inform how content would appear in recommendations, navigation, and search results and offer choices in line with a customer’s tastes and dietary preferences. This wasn’t a scalable process to apply to all of Meredith’s brands. Meredith turned to machine learning (ML) to make content classification more repeatable and scalable.

To train its custom ML models to automatically predict text categories without requiring in-house ML expertise, Meredith Digital started using Cloud AutoML Natural Language.

“Cloud AutoML Natural Language met our needs more than other solutions we considered. Cloud AutoML was relatively easy to use and we were able to quickly get up and running. Also, our initial Cloud AutoML tests produced good results compared to open source tools like scikit-learn,” says Grace. “We can better identify and respond to content trends, creating more relevant, higher-impact audience experiences.”

“With manual content review processes, it took us years to reach the unique level of content classification maturity that we achieved at Allrecipes. Cloud AutoML Natural Language will help Meredith’s other media properties get there in months with automation.”

Grace Preyapongpisan, VP, Business Intelligence, Meredith Digital

Solving the content classification problem

Meredith Digital put Cloud AutoML Natural Language up against its highly skilled manual review teams, using high-quality training data based on their processes. Teams of people and Cloud AutoML Natural Language classified the same set of approximately 10,000 recipes, and the results were then audited without the auditors knowing which review process had been applied.

“In our very first test, Cloud AutoML Natural Language provided automated content classification that was at least as good as our experienced manual review teams,” says Grace. “The results closely matched human teams on accuracy, precision, and recall, which are the primary metrics for evaluating classification models. With manual content review processes, it took us years to reach the level of content classification maturity that we achieved at Allrecipes. Cloud AutoML Natural Language will help Meredith’s other media properties get there in months with automation.”

Staying in tune with consumer trends

In certain cases, Meredith doesn’t need to train custom ML models to review content and user comments. To derive insights from unstructured text using pre-trained models, Meredith uses Cloud Natural Language, an easy-to-use REST API. For example, Meredith uses entity detection and sentiment analysis to extract actionable insights on user experiences from social media and to detect new topics of interest to fold into its taxonomy.

“Especially in consumer-interest topics such as food and beauty, we need to stay on top of trends to stay relevant and cutting edge,” says Grace. “We see Cloud AutoML Natural Language and the Natural Language API as opportunities to monitor sentiment in conversations and surface topics that we don’t even have awareness of right now.”

Meredith’s stakeholders who are interested in reporting on food trends now have an interface, and a data discovery framework that helps them better understand which dishes are most popular and which are showing year-over-year growth. For example, if primarily vegetable-based recipes are growing in popularity, editors can drill down to see what other attributes those recipes have and apply those insights when developing new content and strategies.

“More than increasing overall volume, machine learning helps us ensure that we’re asking our editorial team to focus their attention in the places that will resonate most with our readers,” says Grace.

“Meredith continues to innovate in the ways we develop, deliver, and manage content. Cloud AutoML Natural Language will shorten the time to insights with automated content classification based on specific business needs, our custom taxonomy, and custom models. The speed at which we’re moving is phenomenal.”

Grace Preyapongpisan, VP, Business Intelligence, Meredith Digital

Building readership and loyalty

As more publishers realize the value of structuring their consumer data around hundreds of topics and terms, Meredith is ahead of the game. The company is already using content in innovative ways that build loyalty for its brands. For example, if readers are looking at rainbow piñata cake recipes, they might be planning a birthday party. Considering readers’ specific context opens up additional editorial and marketing opportunities and helps Meredith create a more personalized and meaningful reader experience. Meanwhile, with better informed content strategies from rich data, editors can focus on getting the most relevant content in front of audiences to provide the most inspiring reader experience, rather than reacting to well-established trends.

In 2015, Allrecipes launched its first major redesign in nearly a decade. Enhancements included a mobile responsive design, elevating the social features of the site, such as adding favorites and cook profiles, and introducing a personalized content feed on the homepage. Since then, Meredith’s iterative efforts to surface fresh, personalized content on the homepage have been quite successful. Following the redesign, the number of return visits to Allrecipes.com has increased 10 percent, homepage views have risen 20 percent, and daily visits to the homepage have grown 10 percent.

“Meredith continues to innovate in the ways we develop, deliver, and manage our content,” says Grace. “Cloud AutoML Natural Language will help us shorten the time to insights with automated content classification based on specific business needs, our custom taxonomy, and custom models. The speed at which we’re moving is phenomenal.”

About Meredith Corporation

Meredith Corporation has been committed to service journalism for 115 years. Today, Meredith uses multiple distribution platforms, including broadcast television, digital, print, mobile, and video, to provide trusted content and experiences on a wide-range of topics that include entertainment, style, food, home, lifestyle, health, wellness, travel, and luxury, and to deliver messages from its advertising and marketing partners.

Industries: Media & Entertainment
Location: United States
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE