In the past four years, Canada’s Torstar has transformed its operations using data. From subscriptions to registrations to advertising, data services have reinvented the company’s approach and made it a leader in the use of machine learning, AI, and personalisation.   

Torstar’s data transformation journey began in 2017. Although it has made significant changes in its operations, it still hasn’t reached the milestones Chief Data Officer John Souleles imagined it would have reached four years later.

“Back in 2019, I thought that by 2021 we would have a lot of the personalisation and capabilities [we needed] in place,” he said. However, COVID-related delays pushed some changes into 2022, and Souleles said they are now starting to see “the light at the end of the tunnel” to accomplish the final stages of Torstar’s data transformation.

Analytics improve operations

The use of analytics affects every segment of operations. It has improved subscriptions by providing:

• Predictive analytics that can determine future outcomes of reader behaviour. This allows for more efficient targeting in areas of subscriber acquisition, churn management, and user engagement.

• Algorithms that understand what content drives user engagement and then segment users into different groups for more efficient targeting.

• Content science, which provides actionable insights to help newsrooms in their content strategy development.

• Visualisation to track subscription performance and compare KPIs to budgeted targets.

“Our drive really is three areas: content, customer, and client,” Souleles said. “We work closely with the editorial team to drive insight from analytics to … optimise their newsrooms and subscriptions. We work really closely with marketing and digital to acquire better customers, and then we also work with the sales department to leverage that first-party data and create audiences as third-party cookie goes away.”

From a subscription sales perspective, the data has helped provide improved forms of analysis, including:

• Geospatial, which allows for precise audience targeting based on geographic location. It also gathers customer intent signals.

• Sales performance analytics and sales support to enhance sales management across all media sales.

• Marketing research insights to aid with strategy development and execution.

• Advanced audience insights to provide solutions to clients, allowing them to target the right audiences with the right message and the right time.

“We spend a lot of time analysing data, predicting churn, and [looking at] who we should acquire,” Souleles said. “We have a lot of algorithms we use as signals to be able to identify customers’ behaviour.” That data then informs their opportunities and strategies. It allows them to create more customer stickiness and reduce churn.

Improving accuracy 

One challenge Torstar faced was the way clearing cookies made it difficult to identify anonymous users that were engaging. They created a new model that allowed a persistent cookie to track user behaviour with each repeat visit. This provided greater user history and more accurate predictions.

“That helped lead to better customers [with] a higher propensity to subscribe,” he said. “And we were able to offer more engagement tactics for low users to drive more visits.”

As Torstar continues its data journey, it is moving toward more advanced capabilities. It is partnering with universities to discover ways to resolve some of the machine learning bottlenecks it has encountered.

The company has made tremendous progress in four years, and Souleles compares its data journey to building a house: “You have to start from the ground up.”

The foundation began with putting in a data warehouse, governance, finding the right people, and building the data architecture.

“Now, we’re really focused on the top floor,” he said. “This is where we start getting into recommendation engines, personalisation, leveraging our AI capabilities, and automating our marketing processes. The foundation is there, but now the opportunity is going to expand a little bit more.”  

This case study originally appeared in the July INMA report, The Guide to Smart Data Strategy in Media. 

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