The Reporter: Using AI to check the pulse of donor behaviour

Project name: Donor Behaviour Analytics Framework

Newsroom: The Reporter (Taiwan)

Solution: A reusable analytical framework using first-party data to explore connections between reading habits and donation behaviour

Organisation staff size: 21 - 50 staff

Tech stack: First-party analytics infrastructure, AI-assisted article classification, data analysis tools, and local AI infrastructure

Impact: Identified key reading patterns across donor segments to inform targeted reader-revenue strategies

Core lesson: Don't wait for perfect data. Build a research framework and choose reading metrics that fit your journalism


The Reporter is a non-profit, investigative news organisation based in Taiwan. During the JournalismAI Innovation Challenge it tried to reconcile its historical aversion to being guided by data with a growing need to understand its reader and donor behaviour. 

The Reporter historically followed the slogan, “We don’t see the data”. They did not rely on readership metrics to make editorial decisions, unlike most standard newsrooms. The slogan emerged from the organisation’s commitment to public and democratic-spirited investigative journalism. However, when a major donor of almost a decade decided to stop funding them, it created a critical life-and-death situation that forced organisational and cultural change.

The problem: Donor dependence

“It was a very important donor but when they decided to stop that, it was like wow! We felt the earth beneath us shift. So that's the first time we felt we needed to seriously look at the data and change our strategy,” shared Waiting Tseng, Product Manager at the organisation.

Financially, individual online donations have grown steadily since 2019 and, by 2023, funding had diversified such that no single donor contributed more than USD 100,000. 

The changes prompted the team to eventually apply for the JournalismAI Innovation Challenge. It was a crucial opportunity to secure the CEO's agreement to implement a data strategy and understand their reader and donor behaviour better.

Building the solution: Setting up the groundwork

Over the last two and a half years, Tseng and her team have been collecting and consolidating the data required to decode donor and reader behaviour. They stressed the importance of data collection and hygiene.

“At some point you have to start the process of collecting data so that you have enough of it to run experiments. While it may still be partially imperfect, a lot of things we already did we couldn't have done without it,” added Sven Lee, AI engineer. 

They pulled data points from different sources, including social media. However, the signals from social media data pointed to a large number of people listening to podcasts or watching videos rather than reading reports, causing hesitation and doubt with regard to the original question of linking reading behaviour and donations.

A quick pivot back to gathering first-party data, thanks to the programme coaches’ recommendations, led to some meaningful insights. By using first-party data rather than anonymised social media, they discovered positive reading signals linked to donation behaviour, prompting them to execute live experiments.

The opportunity: Pivots for success

While their initial hypothesis targeting conversion from one-time to periodic donors failed due to a small data sample of only hundreds of data points, they pivoted soon enough. They started comparing groups across 129 topics and found key insights. 

“We came to find that in a lot of ways for instance periodic and single donors had hardly any difference in their reading habits across most of the 129 topics. The frequency distribution of what they read was very similar even if how much they read and how often they read was different. This helped us avoid going down rabbit holes (for data analysis and explanation),” added Lee. Currently, the team is also running an experiment to further cement insights about the connection between reader behaviour and donation. 

Practically speaking, once the experimentation is done, The Reporter’s in-house data team will build an AI chatbot interface that will help the organisation’s non-technical marketing staff turn data insights into experiments and actionable strategies more quickly.

Lessons for newsrooms

  • Find the right reading metrics for your journalism: Different types of journalism require different ways of measuring reader engagement. For long-form investigative reporting, combining reading frequency with reading time and scroll depth can provide a more meaningful model for analysing reader behaviour.

  • Don't wait for perfect data: Data will never be completely perfect. Start by building a research framework and exploring what your data can tell you. This allows teams to keep learning, adjusting their direction, and improving data quality and research methods along the way. And AI can help accelerate this process. The project also highlighted the value of first-party data to help with internal models.

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.