How Torstar transformed from legacy media to data-fueled
Conference Blog | 04 September 2019
Transitioning a legacy media company to a data-fueled enterprise is challenging for publishers worldwide. In Canada, Torstar has rapidly implemented data centralisation, added data scientists and engineers, built out the data infrastructure to editorial and marketing teams, and is now using first-party data for advertising sales — all with journalism and data at the core of its strategy.
In a Webinar on Wednesday, INMA members tuned in as Torstar’s chief data officer, John Souleles, presented the details about the company’s two-year journey and shared potential lessons for publishers looking to perfect their own data transformation.
Two years ago, when this process began, Souleles said his challenge was how to organise Torstar’s data so that it could be used for transformation.
In 2017, the Torstar organisation was defined by:
- Being newspaper-centric.
- Print subscriptions.
- Free digital access.
- Brand/product focused sales approach.
- Disparate assets.
“We were operating in silos and in a way that did not look across multiple units,” Souleles said. “We wanted to move along from there.”
At the same time, the company wanted to hold fast to its mission of quality journalism. “We always value journalism, and that’s at our core for sure. We wanted to get to paid digital access, and we wanted to focus on having performance-based analytics around it — to set targets and have solution-based advertisement.”
So the question then was: How would Torstar take data and fuel it into journalism?
“Data underneath journalism was the key aspect of where we were moving towards,” Souleles said. “Our vision is to work with various communities and have them thrive and continuously grow along with Torstar.”
Underneath this the company had data, which was the pillar of everything. “Within two years we’ve pinned a lot of data under content, and it’s starting to form more and more decisions.”
Today, the company is much more customer focused:
- Customer and client-centric.
- Value on all journalism.
- Paid digital access.
- Performance-focused sales approach.
- Torstar 360 view.
To put this data to effective use, Souleles said Torstar had a lot of touchpoints and a lot of good reach, but the company was sitting in separate silos and not very effective. Transforming the company culture and democratising data quickly emerged as key factors for success.
How to become a data company quickly
One of the reasons Torstar was able to make this change within a two-year period was because the team leveraged data as a strategic asset by enabling users with actionable insights.
“We had to put in controls and make sure we had the right quality of data, and bring those silos together,” Souleles said. “We had to focus on action because if you can’t partner with the right stakeholders, this is all useless. We’re really focused on change management and using machine learning to predict customers’ paths to purchase. As we get better and better at understanding our clients’ needs and interests, it will be a lot easier to serve them.”
Key success factors for Torstar were:
- Identifying use cases.
- Building a strategy or roadmap.
- Hiring the right people.
- Executing on infrastructure.
- Implementing change management.
“We spent a lot of time at the beginning documenting the use cases of what we wanted to drive this change,” Souleles said. “You better be able to set a target for where you want to go. You’ll iterate along the way, but if you just wing it, it’s going to be harder and you’ll make a lot of mistakes. At least having a roadmap will make it a lot easier.”
Souleles added that change management was the most important part of this transformation.
When it came to monetising its data, Torstar had two key areas it was really trying to move: subscription recurring revenue and high-performing advertising. To do this, they really focused on identifying and documenting their use cases.
“We started with those [identified] KPIs and tied in the use cases,” Souleles said. “Once we documented that, we decided what data and analytics we needed to accomplish our goals. That’s how we prioritised the data we needed, with those use cases. We didn’t use all the data. The next step was to apply the right analytics to execute on that. Sometimes we had to invest in new solutions to execute.”
As the team learns from mistakes they iterate, get better at it, and then repeat the process.
“Today we are moving towards that 360-view of our subscriptions and a better understanding of our clients,” Souleles said. “We’re able to do same-day or next-day analytics pretty quickly. We’re focusing a lot on audience management, and our machine learning tools are in place to help drive subscriptions and revenue.”
The goal by 2021 is to have a real-time decision making processes. “When we put the right tools in the right hands, in a simplified way, we’re better able to make the right decisions. We’re going to expand our data capabilities by looking at our data sources. And moving toward a concept we call Next Best Advice by predicting what our clients need next.”
To execute on this road map, Souleles said he thinks of it as building blocks. “Each piece adds to the next and builds up the next level, and gives us a better return on our investment. There’s a lot of heavy lifting up front, but once you do that it’s easier to execute and iterate. We wanted to monetise on this plan pretty quickly.”
Investment in the right places
People are very important to this process, particularly getting buy-in from leadership. “Some of the folks on our team are trained to use data and ramped up pretty quickly,” Souleles said. “We built a data team of about 30 folks in six to nine months.”
When you have a road map and working in the cloud, with the right people and teams and you’ve done this a couple of times, it gets much easier, he added. Working closely with the team and peers to make decisions is a key factor. The fact that data has a seat at the table makes it a lot more effective.
“Once we looked at the infrastructure and data, it becomes a simple process at the end of the day. You turn that raw data into something the marketing people can understand and use. Data engineering is a big process, turning the raw data into something useful. Then comes analytics, execution and tie it back to action. A simple process, but a lot of work to get it done.”
Increased data adoption
Data has become a part of every discussion at Torstar. The CEO has a direct line to data, and every department works as a hub around it: finance, operations, editorial, marketing, sales, digital product, and human resources.
Increasing this data adoption in a way that results in optimal efficiency includes:
- Providing solutions to end users instead of managing ad-hoc requests.
- Finding things that are common and operationalising them.
- Utilising self-serve tools for the marketing, digital, editorial, and sales teams.
The other part of change management lies in understanding the business problem and the data that backs it. “We want to be able to service the customer needs, and if we can understand the business problems, then it’s easier to figure that out,” Souleles said.
Another key aspect is training and communication to create the analytic culture. “Imagine if all the employees of the company got 20% of the knowledge about data how effective the company would be.” They do a lot of lunch-and-learns that take it down to the detail level and train people how to use the tools.
“About a third of our employees have access to data services that caters to their needs,” Souleles said. “Not all of them use it, and that’s where change management comes in. Through proper coaching and training, more and more of them will use it.”
This expanded access to data includes improved visualisation tools that track business KPIs, have the ability to slice and dice by different dimensions and metrics, and offer self-serve data discovery and exploration.
The team has analysed more than 150 use cases over the past year and a half. Some have worked and some haven’t, Souleles said. He shared a few of these examples.
- Subscription. In subscriptions, data tells them what news and content drives subscription revenue, and therefore what news and content to focus on. On the acquisition side, it’s about targeting the right customer at the right time. This enables Torstar to pop up the right offer at the right time as the users move from registration into subscription.
- Acquisition. Data surgically helps the team acquire the right customer, using the right channels at the right price — and profitably.
- Retention. This is also an important aspect. “We have to be able to target the right customers who are likely to leave us with the right offer,” Souleles said.
One example of this is through locked content: “We needed to be able to lock articles in real time, and we were able to accomplish this in 45 days.”
Content science is a good example of how Souleles’ team is working with the editorial teams. They took about 300,000 Torstar articles and put them in the data lake, then compared pyramid tags to topics surfaced by Natural Language Processing. They came up with 45 models in four hours for each pyramid tag. “We know crime potentially drives more subscriptions, so here’s content tied back to that,” Souleles shared as an example.
Predictive models are used to predict the newsletter users who are more likely subscribe. “You don’t make the same offer to everyone. You target those who are more likely to subscribe than everyone in the pool.”
High performance digital advertising
Torstar went from using cookies to using 150 attributes and more than 50 algorithms to create first-party customer data. “It’s not just based on the cookie, the unknown user,” Souleles said. “It’s more about the known user, to identify what your clients need and the best targeted ads for them.”
Many unknown users come onto the Torstar site, he said, but they use a “look-alike approach” to create better data on them. “For example, have the age and gender of a known user, we can use the look-alike approach to help determine those stats on the unknown users.”
As a result, Torstar went from mass targeting to first-party data that gets down into more micro-information. “It helps us understand our customers so we can pitch the right products.” For example, with grocery content, the team can determine which customers are more interested in coupons and deals and which are more interested in shopping online to present different personalised offers.
Client insights are used with the sales force, using the look-alikes, to help the sales team determine where to sell advertising. “They can get a lot more targeted with smaller campaigns, which definitely a lot of our [advertising] clients will be more interested in,” Souleles said.
He warned this is much more difficult to executive if a company’s executive team doesn’t believe in data.
“Start with what metrics you’re trying to move, and then tie those use cases to build out a roadmap. Don’t be shy — try looking out three years and decide where you want to go and create the steps to get there.
“It’s not just about hiring the right people but creating a culture of engaged individuals. Start executing and partner with your stakeholders. Change management is the tough part. It’s sitting at the table, having a lot of discussions and doing a lot of presentations.”
INMA: What are some of your most important KPIs?
Souleles: At a high level, it’s really ad revenue and subscription revenue. Net adds, how many new subscribers are we adding. Ad revenues, it’s looking at click-through rates and other metrics driving advertising.
INMA: What are the results of how you’re using the data?
Souleles: We’ve used the data analytics to help drive subscriptions and reduce churn levels. A lot of that was tied to targeting the right customers and having a BI tool to understand how many new clients we were getting. We understand this data now in almost real-time so we can change immediately. We also understand what content our customers engage with.
INMA: How much has data impacted on ad revenue in term of increasing yields?
Souleles: When we first started this journey, we focused a lot on subscription and content. We’re starting to move toward advertising as we speak now.
INMA: Do you feel that ... data sharing with third parties, or selling data as a stand alone product, should/would be a priority or would you advise companies to keep their data to themselves?
Souleles: For us, data is really important. With privacy issues, sharing that data with a third party would not be part of our strategy at this moment. We have to manage the data in a secure environment.
INMA: What advice would you have for small publishers who can’t hire a 30-person team?
Souleles: Working with a cloud vendor will get you up and started pretty quickly. They become almost like your IT team. You definitely need a data engineer to manage that and possibly a data scientist. Leverage your top provider and free consulting to get those kinds of services. A data engineer and two data scientists will probably work.
INMA: What’s your view about using Facebook and Google services (i.e. log-in) versus a first-party data objective?
Souleles: They’ve got their first-party data. That’s why they’re winning the game. You’ve got to focus on getting your own first-party data. It’s very important that you start collecting known users and getting that data for yourself.
INMA: What were the challenges of the culture transformation?
Souleles: For me, I’m not an expert on publishing so ramping up quickly on that was important. Working with the editorial team and moving towards this was a fun experience. I can’t believe how quickly the editorial team has adapted and look forward to using the data, but a lot of that was because of leadership buy-in. It’s pretty exciting to see. On the sales side, there definitely are challenges.
INMA: What are the methods of getting first-party data and how do you use this?
Souleles: First-party data is really about getting people to sign up to use your service. Basically, give us some of your information and we’ll exchange by giving you a subscription. All that data is put into one place and creates the 360-view of the customer to create more personalised offers for them. We also take that first-party data and extract it to get the lookalike data.
INMA: What is personalisation starting to look like for you?
Souleles: I don’t think it’s one solution or one product that gives you personalisation. I think it’s a bunch of things working together that give you that. Offers in real time is a form of personalisation. In a year or two we might have content that serves up one-to-one. The content changes due to a user’s needs, changing throughout the day, and getting advertising that’s relevant to you. It’s multiple use cases working together to create that better customer experience at the end of the day.