Choosing a voluntary membership model in reader revenue comes with a plethora of problems. Daily Maverick, the South African newsroom, has faced many of them head-on. The three key components of a membership funnel include acquisition, engagement, and retention. The Daily Maverick team found that while they had strategies in place for all three components of the funnel, with a growing audience, these strategies required consistent iterations.

The problem: Moving from reader interest to action

There was a significant drop-off from people clicking to convert and those who actually completed their membership signup. While the global industry standard for paywalls is 11%, Daily Maverick’s voluntary membership was sitting at 3.5%, according to Fran Beighton, Head of Growth at Daily Maverick. 

“When we dug a little deeper, we saw that often our messaging was so emotional by the time they [readers] got to our standard landing page, something didn't resonate with them and they failed to complete (the signup),” explained Beighton.

Equally important to them was the focus on community to anchor their membership, said Beighton. “Community can not be a one-way street; Daily Maverick has benefited from our members so much but they need to also feel how the community can change their lives and businesses,” added Beighton. 

Building the solution: A 360° view of membership

Keeping this in mind, the team came up with an idea for a holistic membership strategy, aided by AI, to target the three components of a membership strategy, and with a renewed focus on community and engagement. 

It led to optimising the landing page for acquisition, maximising engagement by creating the Daily Maverick Connect platform, and an Impact Tracker tool to demonstrate the impact of their journalism. The team behind the Revenue 360 project included the reader revenue team, audience development team, head of operations, and the editor-in-chief, said Beighton.

First, the landing page

A crucial element of the entire puzzle was in optimising their landing page for conversions. “Even with a voluntary membership, we had an added issue. There's no impetus to absolutely 100% signup. It's not like it's an e-commerce store where you want to buy a pair of trainers and you find other offers or testimonials. That element of urgency for the potential member isn't there. So, we have to be really specific and careful about what is on the landing page and what's just a waste of time,” said Beighton. 

They added heatmaps to the page to find out and discovered that 75% of the people didn’t go below the fold. “What's really interesting there, is that it showed us where we need to concentrate the information, where we need to add behavioral nudges, to explain more, while still keeping it clean and not overwhelming,” added Beighton. 

They changed the landing page from a static page that encompassed all ideas and started delivering tailored messages to specific audience targets. Initial testing showed a 55% increase in conversion. It also led to ongoing iterations and improvements.

The team built a custom A/B testing module specifically for this, which intercepts traffic to https://ab.dailymaverick.co.za and then redirects to set URLs based on the parameters set, with a tracking component. For the landing pages themselves, they integrated the pages into their Wordpress / Woocommerce setup to enable integration into their Maverick Insider system, and included Hotjar and Heap tracking.

Next, fostering community connections

The team also recognised the importance of deepening meaningful engagement with their community to eventually help in more acquisition and retention, said Beighton. This is where the Daily Maverick Connect social networking platform comes in. 

“We want our members to be able to network with each other, whether that's finding new clients or new jobs or a new book club. That could exist by itself, but the purpose behind it is to engage with each other and with Daily Maverick,” explained Beighton. The tool, which was in beta-testing with staff at the time of this interview, will be first opened to an initial closed group of members before wider expansion. 

The biggest challenge they faced while creating this platform was in continually engaging users without it becoming addictive, intrusive, or overwhelming. Beighton confirmed that the team has many safeguards in place to avoid toxicity, specifically by training an AI bot to prevent misuse and misbehaviour. The tech stack for the Connect platform includes using Discourse and Docker. 

“We are using ModerationAPI's AI model which we tuned based on its response to test inputs in order to encourage the behaviour on the site that we want to promote,” said the team. “The AI Moderation tool we have implemented, will allow us to moderate comments at scale, with minimal human intervention, by passing them through custom prompts and training the system on what we regard as acceptable and non-acceptable comments. This will allow DMC to perform optimally, and conversations to happen on the platform without a slow and cumbersome moderation regime,” they added.

Last, impact for retention

The team also created an impact tracker tool, ImpactEngine, that uses an LLM trained on past stories to scan parliamentary notes (Hansard) and identify how Daily Maverick's journalism effects change in parliamentary meetings. This tool serves as a research asset for journalists, allowing them to track the progression of stories, find experts, and design their reporting with intentional impact. It also generates reports on the effect of specific teams’ work which is then shared with readers to demonstrate the organisational impact.

“We didn't anticipate how useful it would be for the very construction of journalism,” said Beighton. The tool will also be used to cover local committee meetings, which Beighton says is crucial for a “resource-constrained” industry without enough people to cover towns and municipalities with “boots-on-the-ground”.

The technology behind the ImpactEngine include SvelteKit, Ollama (Mxbai-embed-large) OpenAI (GPT 4.1), MongoDB and Qdrant. Most of the "AI" logic is built around vectorisation, so LLMs are not particularly important for this project, shared the team. They selected GPT 4.1 as it has a large context window and is good at creating the types of reports that ImpactEngine outputs. The biggest challenge they faced was in terms of time taken to ingest data cleanly into the system, well-chunked, and with sufficient metadata to be useful. 

Lessons for newsrooms

• AI empowers smaller newsrooms: Recognise the significant possibilities AI offers to smaller news organisations, specifically its power to fast-track project development and overall technology adoption.

• Value interdisciplinary teams: Consider it a "huge mistake" to approach AI projects without an interdisciplinary team. Since the first users of internal-facing tools are journalists and staff, breaking down silos and collaborating across the board is essential for success and user-friendly design.