Unlocking paywall growth with AI

How Alma Media kickstarted subs growth with an AI-infused paywall.

Johanna Suhonen of Alma Media talking at the Future of Media Technology conference
👨‍💻
These are live-blogged notes from a Case Study session at Future of Media Technology conference. Prone to error, inaccuracy and howling crimes against grammar and syntax.

Johanna Suhonen, VP of Content Business, Alma Media

Kauppalehti is a business title with 700,000 readers looking for reliable, in-depth reporting — and willing to pay for it. The site’s been on a journey with its paywall, moving through different models over time.

COVID was a boon for their subs, but since then, growth in subscriptions has flattened. They’ve experimented with content, paywall design and pricing, but with no real impact:

Timeline of Kauppalehti’s paywall evolution, from 2012 launch to 2025 smart paywall with stricter access limits.

And then they started applying a propensity paywall — but they lacked the conversion data to make it work.

This year, they’re publishing 65 articles a day, 15 of them premium, and an archive paywall that locks everything after four days. The free content was available through a metered registration system, which kicks in after two free articles. It still wasn’t working, so they started talking to Mather about applying their AI tools.

Sophi-stication

In partnership with Mather, they applied the Sophi AI propensity system to their content. The results?

  • Subs 37% up
  • Registrations 153% up
  • Logins up 269%

But they’ve also got useful insights into content strategy and paywall process — leading to streamlined newsroom workflows. At the moment, they’re in a test phase, when Sophi recommends paywall decisions, and they are manually implemented. But that’s moving to automated as they bring the system into production.

They’re building on this start with more ongoing testing, some adaptive modelling to target special paywall cases, adding more data to the model — and rolling it out to other titles in the group.

Getting buy-in from the team

Was there pushback from the organisations? Yes — the organisation always wants to build it themselves first, and only goes to vendors when they struggle. And there were worried about vendors’ ability to support niche titles in small countries with language issues. “Does it speak Finnish?” was a key question.

 A woman with long blonde hair, wearing a white blouse and black trousers, stands at a podium with two microphones. She holds her hands together while speaking, wearing a smart watch and necklace. The backdrop glows with purple and blue lighting, suggesting a conference or keynote presentation.

The newsroom was used to AI experimentation, and were open to the idea of an AI-driven paywall, but were more concerned about the impact on their workflow. They do have an editorial team that is engaged with the sales targets, which helped. But they were worried about being blamed for a failure from Sophi. But the results have convinced them…

Content & Data

The content strategy hasn’t fundamentally changed. However, they can see more clearly the value of articles to different segments of the audience — including smaller ones. It’s likely to change the strategy in the long-term, and make it more focused on readers’ expressed content needs.

The data project was the biggest challenge. There was a lot of work needed to put that into a form they could work with. Collecting first party data is crucial for us and everyone. It was one of the big goals for implanting Sophie to get more registrations.

“This is not the final phase of paywalls, but we are building something new.”