For media and digital publishing companies, there are two main ways to attract revenue: from advertisers and from readers. This time, we want to focus on reader revenue and how your media company can convert casual readers to paying customers.
Many news websites find that the old strategies for driving reader revenue no longer work. Even that exemplar of digital-first news, guardian.com, recently reported a £3 million, or 4%, shortfall in digital reader revenue.
Today, the winners are those embracing new technologies—like propensity modelling—to capture reader revenue and gain the competitive advantage.
The challenge of reader revenue strategies
When it comes to reader revenue, the main challenge media companies face is predicting user behaviour and maximising subscription opportunities.
Since the mass migration of media companies from print to online in the 1990s and 2000s, news websites have found conventional strategies fall short. Looking at some of the traditional approaches for attracting reader revenue, some media companies focused on creating flexible ways to pay for and pause subscriptions, others like Swedish daily Dagens Nyheter reduced their output by at least 15%, and others homed-in on generally improving the user experience or shifting their value proposition.
But none of these approaches get to the heart of an individual reader’s nuanced preferences and behaviours. To achieve that requires an innovative solution: propensity modelling powered by AI is the way for media companies to understand user behaviour and revolutionise their reader revenue strategies.
What is propensity modelling?
Propensity modelling is a powerful application of artificial intelligence which provides media companies with a strategic advantage in predicting user actions around reader revenue.
It’s helpful to think about applying propensity modelling strategies in three key ways:
1. User segmentation: Propensity modelling enables you to better segment users into groups based on their historical behaviour. For example, Guardian Media Group employs this approach to tailor subscription offers to specific user segments according to each reader’s preferences.
2. Predictive analytics: By analysing user interactions, preferences and engagement patterns, media companies can predict the likelihood of a user subscribing. The Washington Post is one of many news websites which deploy predictive analytics in this way to forecast user subscription propensity.
3. Content personalisation: Propensity models can dynamically recommend personalised content to users, which is often hugely effective for increasing engagement. The New York Times, for instance, uses content personalisation to drive reader interest and subscription conversions.
So, those are some of the ways that propensity modelling can be embedded in your business. But how does it work and what technologies power it?
Propensity modelling technologies
Implementing propensity modelling within your media brand involves leveraging several cutting-edge technologies. You don’t need to be a tech and AI expert to enjoy the benefits—there are off-the-shelf, no-code solutions like Bridged Media—but it helps to understand the principles behind the tech. Here are some of the key components.
One element is machine learning algorithms. Algorithms such as ‘logistic regression’ (a process of modelling the probability of an outcome given an input) and ‘random forests’ (which combines the output of multiple decision trees to reach a single result) are widely used for building propensity models. Python libraries like Scikit-Learn provide robust tools for implementing these algorithms.
Another component is data integration, which is because propensity models require seamless integration of diverse data sources. Tools like Apache Kafka and Amazon Kinesis facilitate real-time data integration for up-to-the-minute predictions.
And finally, model evaluation, referring to the method through which media companies can evaluate the effectiveness of their propensity models. These measure metrics like ‘Area under the ROC Curve’ (shown on a graph which tracks the performance of a classification model) and precision-recall curves (showing the trade-off between precision and recall for different threshold) while tools such as TensorBoard aid in visualising model performance.
Measuring success of propensity modelling
When you come to implement your new reader revenue strategy using propensity modelling, here are some of the impactful Key Performance Indicators (KPIs) to include.
- Conversion rate: This measures the percentage of users who subscribe after being targeted by the propensity model.
- Subscriber retention: A measure of how well the model identifies users likely to remain subscribers.
- Lifetime value: Which predicts the value of a subscriber over their entire lifecycle.
Setting targets for these KPIs involves understanding your baseline metrics and progressively aiming for realistic improvements, all of which will be unique to your news website. You may want to target a 10% increase in conversion rates within the next quarter or enhance subscriber retention by 15% over the next six months—perhaps you could achieve even more.
Ready to embrace no-code AI tools?
When you embrace propensity modelling and other AI-powered technology, you may find greater business benefits and surprising trends. Take heart from The New Yorker which—after concerted AI-powered activities—is one of an increasing number of media companies where reader revenue now exceeds advertising revenue, by a staggering 65% to 35%.
Bridged Media offers a no-code AI solution, empowering media companies to seamlessly integrate propensity modelling into their reader revenue strategies.
So, if your media company doesn’t have the scale or investment capability of The New Yorker, Bridged’s tools eliminate the need for extensive data processing or dedicated AI resources, making it easy to adopt AI and effective at optimising reader revenue strategies. It’s out-of-the-box, off-the-shelf and, most importantly, a no-brainer.