Malaysiakini: Using AI to spot subscribers before they churn

Project name: Re-engage

Newsroom: Malaysiakini (Malaysia)

Solution: A predictive analytics tool that scores subscriber churn risk, allowing the subscription team to proactively engage readers before they lapse

Organisation staff size: 51 - 100

Tech stack: Machine learning churn models (gradient boosting, random forest), first-party subscription data, CRM and email marketing tools

Impact: Identified high-risk subscribers and key retention packages, directly informing outreach and pricing strategy

Core lesson: Churn isn't necessarily an editorial problem. Understand your subscription data before reaching for AI-driven personalisation


Subscriptions generate 40 per cent of revenue for Malaysiakini, a leading independent news site in Malaysia. During the JournalismAI Innovation Challenge, the team addressed rising subscriber churn. Though initially assuming content was the main driver, they quickly found engagement and payment behaviour were equally critical.

The problem: The bump that didn’t last

Malaysiakini pioneered reader revenue in Southeast Asia back in 2003. It was a long journey before subscriptions became a substantial portion of its income during the COVID-19 pandemic. As network advertising revenue dwindled, the team tightened the paywall, improved checkout flows, reduced friction and experimented with new digital offerings.

Political turbulence played a role as well. Malaysia went through four prime ministers in three years, and that tumultuous period gave Malaysiakini its own version of the “Trump bump” in subscriber growth. But as the political situation stabilised and news consumption habits shifted, churn climbed to unprecedented levels.

The subscriptions team tried every strategy available to win back former subscribers and engage those nearing expiry, ranging from broad campaigns around Press Freedom Day and National Day to, very reluctantly, price discounts.

“The subscription team wasn’t short of ideas. But what they were short of was data, specifically who are those considered a churn-risk,” said Andrew Ong, who led the project as Malaysiakini’s product lead before becoming its editor-in-chief in August.

The stakes go beyond revenue. Malaysiakini is pivoting its newsroom towards user needs to get closer to its audience. Its subscribers already form a dedicated community: they can comment on stories, a premium feature Ong believes is unique in Malaysia, and subscribers and even former subscribers alike turn up to its events. 

However, Malaysiakini lacks a dedicated tech team, and issues with a previous CRM and CMS vendor had left it with many unknown gaps in reader data.

Building the solution: Different models for different readers

With support from JournalismAI, Malaysiakini brought in a data scientist for the first time and began building Re-engage. The team worked from the two datasets that were still intact, one running up to June 2023 and the other a full set for 2025. They used them to score each subscriber’s probability of churning, combining techniques such as gradient boosting and random forest. To test the model, they checked its 2025 predictions against what subscribers actually did in 2026. It held up well, giving the team confidence in its accuracy.

Early on, the team realised one model would not fit all of its subscribers. Malaysians generally prefer to pay online directly through their bank accounts, and banks there don’t support automatic recurring payments the way services like Stripe do. Helping readers set up recurring payments would require the subscriptions team to learn  dozens of different banking apps. As a result, Malaysiakini serves two very different groups: one-off subscribers who renew manually, and recurring subscribers.

So the team built separate models for each, and found vastly different behaviours, stickiness and churn probabilities. Manual renewers make a conscious decision every year, while recurring subscribers often leave for other reasons, including failed payments.

The project also hit setbacks. Malaysiakini had to migrate its CRM and mailing system mid-project, and a significant amount  of user data failed to transfer. The pipeline tracking reader behaviour, such as time spent on site and pages read, collapsed in the process. The team relaunched data collection in September 2026 and will rebuild the model once enough new data has accumulated.

The opportunity: Reaching readers at the right moment

The models surfaced several key insights:

  1. One-off subscribers who haven’t visited for 30 days have a greater chance of lapsing. Recent engagement matters far more than long-term trends.

  2. Package type matters. Of Malaysiakini’s monthly, one-year and three-year packages, subscribers stick around longest on the three-year plan.

  3. The biggest retention opportunity may lie in converting manual renewers to recurring payments.

“If somebody is inactive for 30 days, don’t wait. Send them a love letter. If you can, personalise it, because it may take two, three back-and-forths, maybe more, for them to commit their credit card,” said Ong.

The findings are already reshaping strategy. Malaysiakini is due for its first price increase in about eight years. Ong had been advocating a Spotify- or Netflix-style model, with lower monthly prices replacing annual and multi-year packages. The data pointed the other way: since multi-year subscribers are the stickiest, the team now needs to work out how to move readers up to longer, more expensive packages.

Originally, the team wanted outreach to be fully automated. It is now moving towards more personal communication instead. “We don’t trust the AI enough to represent us,” said Ong. “Our subscribers are in the five figures, but the ones who interact are not that many, and some of them end up being very good news sources for us. So it’s important for us to keep the relationship warm.”

The team had planned email experiments for readers on its churn-risk watchlist, testing, for example, whether focusing on less active readers, or tailoring outreach by package, has the most impact. However, the disruption caused by the CRM migration weakened how those campaigns were designed, so the experiments still have work ahead.

Moving forward, Malaysiakini is working with its new CRM provider to integrate the model directly, ensuring the subscriptions team receives a weekly updated list of subscribers needing attention. Since the watchlist updates dynamically as interventions succeed, regular refreshes are essential. The team will also re-validate the model once enough new data is gathered.

Lessons for newsrooms

  • A good data scientist is worth the investment: Working with one for the first time was eye-opening, especially for a newsroom without a tech team. Translating machine learning into newsroom language takes time, but with patience (and perhaps a beer or two), the conversations get easier.

  • Analyse churn root causes: Retention issues often stem from payment methods or user experience rather than editorial content alone.

  • Let the data challenge your assumptions: Segmenting subscribers by payment type and package revealed patterns a single model would have hidden, and overturned the team’s plans for a monthly-first pricing model.

  • Keep humans in the relationship: Use AI to identify who to talk to, but not necessarily to do the talking. Your most engaged subscribers can also be your best sources.

  • Protect your data through system changes: CRM and platform migrations can quietly break the data your models depend on. Plan for them, and check what has transferred.

Explore case studies from the 2025 cohort

The 2025 Innovation Challenge Report, Prompting Sustainability, collects case studies from all 12 grantees and reveals their journeys, challenges, and opportunities.

Read about their experiences, discover best practices, and find inspiration in their innovative approaches to integrating AI into journalism.


The JournalismAI Innovation Challenge, supported by the Google News Initiative, is organised by the JournalismAI team at Polis – the journalism think-tank at the London School of Economics and Political Science, and it is powered by the Google News Initiative.