It’s a pleasure to join you for the fourth and final newsletter of the CARL project and to share some of what we have learned at Prensa Ibérica about using data and technology to tackle the challenges media organisations face today.
One of the things we have particularly valued throughout CARL is the opportunity to exchange ideas and experiences with our European partners. And behind many of those conversations has been a simple but important question: how can we turn data into something genuinely useful for media professionals?
For this final newsletter, we thought we would share two sides of our experience: something that is already working for us — combining data and artificial intelligence to improve our processes — and something we are still trying to solve.
First, a little about us
Prensa Ibérica is one of Spain’s leading media groups, with a strong presence across different regions of the country. We started out in regional newspaper publishing and have evolved into a multimedia group encompassing print and digital publications as well as audiovisual operations.
That regional presence is still a big part of who we are. Through our different newsrooms and publications, we aim to stay closely connected to the communities we build with our readers, while adapting our journalism and business models to an increasingly digital environment.
And, inevitably, data and technology have become an important part of that journey. We use data to better understand how audiences interact with our content and products, to support editorial and business decisions, and to explore new ways of producing and distributing journalism.
Something that works for us: using data and AI to automate print production
A good example is ESCRIBA, a project developed by Ediciones Deportivas Catalanas, publisher of SPORT, together with Recursos en la Red.
Anyone who has worked on a daily print newspaper knows how complex the production process can be. Content has to be selected and prioritised, texts and images need to fit into predefined spaces, advertising has to be incorporated, and countless editorial and technical decisions need to be made — usually against the clock.
ESCRIBA explores how artificial intelligence can automate part of that process while maintaining editorial quality and, crucially, human oversight.
But one of the most interesting things we have learned is that successful automation doesn’t start with AI. It starts with understanding the process you want to automate.
Before automating print production, we first had to look closely at how a newspaper is actually produced and closed. Which tasks require manual intervention? Which decisions are repeated every day? What constraints apply to different page layouts? And which editorial rules must always be respected?
Once you understand the process, you can begin to translate that knowledge into data and parameters. Headline length, text volume, image proportions, available space, advertising, story types and editorial priorities can all be quantified.
The system can then use this information to identify suitable layouts, while generative AI can help adapt elements such as headlines, subtitles or text when the content doesn’t naturally fit the available space.
There is another important part of the process: measuring whether the automation actually works. Processing time, number of iterations, manual adjustments, hours saved and the quality of the final output all help us understand whether we are really creating value.
And this is probably the lesson that travels best beyond print production. Introducing AI into a newsroom isn’t simply about adopting a new technology. First, you need to understand your processes. Then you need to turn them into measurable information. Data helps us see where automation adds value, where human judgement remains essential and where there is still room for improvement.
And something we still need to solve: personalization
Of course, not everything works quite so neatly. One challenge we are still trying to crack at Prensa Ibérica is personalization.
Over the years, we have built up a great deal of knowledge about our digital audiences. Our data lake brings together information about how readers navigate our digital properties, how they interact with digital advertising, the subscriptions they have purchased, the newsletters they have subscribed to and other aspects of their relationship with our products.
We can identify readers relatively persistently, both at browser level through cookies and, when they are logged in, at user level. We have also developed activity-based audience segmentations for each of our newspapers, as well as interest-based segmentation using the IAB taxonomy.
In other words, for many of our readers, we actually know quite a lot: how they interact with our digital properties, what they read and what topics interest them.
And yet we don’t always use that knowledge to create a genuinely personalized relationship.
Most of our communications and messages still follow a largely generic approach. A reader may have spent months — or even years — showing us what they are interested in and how they use our products. But in many situations, they still receive essentially the same message as someone we know very little about.
We know personalization can work. Experiments within the company have shown that personalized communications can perform better than generic ones. What we haven’t yet fully solved is how to move from individual experiments to making personalization part of the way we work every day.
So perhaps the real question for us is no longer how to collect more data, but how to make better use of the data we already have.
How do we turn audience knowledge into relevant messages at scale? How do we make personalization part of everyday workflows rather than a special initiative? And how do we make it the natural choice whenever the data allows us to do so?
If we can answer those questions, we can make an important shift: from simply knowing our audiences to using that knowledge to build a more relevant and individual relationship with each reader.
And that feels like an appropriate thought on which to end our contribution to CARL: data becomes truly valuable when we turn what we know into what we do.
Next Steps in Project CARL
As Project CARL comes to a close, all partners are busy wrapping up the final project deliverables and reflecting on key takeaways for their respective newsrooms.
Teams are currently evaluating the data-driven approaches developed or adopted during the project to present at their national dissemination conferences, where they will share best practices, lessons learnt and pitfalls of the data driven approach with local peers.
Soon, partners will gather in Prague to celebrate our achievements, review our collaboration, and identify insights helpful for future initiatives. We look forward to seeing everyone in Prague!