Lessons from managing 250+ micro-communities with 10,000 users and the failure of our first social mobile app

Over the past months, we’ve managed over 50 thematic micro-communities on WhatsApp and more than 200 on our last app, Meerchat. These communities varied from students and language learners to people quitting smoking or losing weight, as well as programmers and meditaters.

On starting micro-communities and group dynamics

We've learned the hard way that initiating conversation among strangers requires significant effort and flintstoning, especially when those are added incrementally to a micro-community. The introduction phase is crucial. Adding users in pairs or groups of three ensures that they share this initial discovery and meeting experience. The ideal strategy is to launch dozens of users at once onto a group. 

An online micro-community is analogous to an offline one, where similar stages often occur: members initially get to know each other, sharing names, goals, and ambitions, but tend to be quite reserved at first. Gradually, they become more comfortable and start engaging more openly, discussing the subject/theme on a daily basis.

In a real-life group, the dynamics vary as individuals with different personalities come together. You'll observe assigned roles and emergent roles, which are shaped by social dynamics and emotions. Additionally, there will be instances of role ambiguity and role conflict, especially during the group's infancy phase when these roles are still being naturally distributed.

In an online micro-community, similar behaviours occur. The difference is that a person can effortlessly become a lurker (or silent quit) since there's no direct accountability. In contrast, in real life, a lurker – that is, a shy or yet-to-be-convinced group member – receives visual and auditory cues from others in the group. They can't hide without leaving and might be prompted or even compelled to interact at a certain point.

Before exploring the topic of lurkers, it's important to acknowledge that for a group to function and thrive indefinitely, it requires defined roles, status, cohesion, and support. In our experiments, we've observed various outcomes: we've formed groups that have persisted indefinitely, groups that faded during their infancy, and everything in between.

Status serves as a motivating factor, both in real life and online, encouraging people to stay engaged in order to gain more status. Online, status is often represented as "points" earned, while in reality, it can be the sense of impact one has within a specific platform or network. Social recognition might be the most sought-after form of currency. Yu-kai Chou, in his work on gamification, describes many ways to bootstrap the social currency through social influence, ownership, accomplishment, and epic meaning. 

There's the external kind, which you want to broadcast to the world: your karma points and how long you’ve been a member of such-and-such platform. Then, there's the internal kind, which intrinsically motivates you..

Cohesion and support are critical for ensuring that a newly formed group survives challenging periods and thrives during favourable times. We've observed a remarkable level of support even just a few days into the process, especially when matching strangers with closely aligned goals and similar demographics.

Lurkers present a fascinating aspect of group dynamics. It's a well-known fact that only a small percentage of users generate the bulk of a network's content value. For instance, only about 4% of Instagram users post stories, yet they are a cornerstone of the platform. The active contribution on Wikipedia is even lower, and few people post on Reddit compared to the many who consume the content. Lurkers are an inescapable part of large social networks and they are even present in smaller group chats (even friends & family ones). In such chats, even with a “view” system in place, some participants will never engage. The situation is exacerbated in the absence of a view system, which eliminates any accountability for lurking.

In our case, the cost of lurking was initially almost non-existent. To address this, we implemented a system that automatically removed users, including lurkers, from a group after seven days of inactivity, or lurking. We found that fighting against lurkers is not effective as they did add value to our small community network. The better approach was to understand how to leverage their presence. We adjusted our strategy to remove only the inactive users while retaining the lurkers, and instead, we focused on activating them.

Andrzej Marczewski and Richard Bartle, in their respective works, outline player types in games and how they might behave in a multiplayer system. Bartle categorizes players as Socializers, Achievers, Explorers, and Killers, while Andrzej identifies Achievers, Socializers, Philanthropists, and Free Spirits. Each type has unique motivations and ways of engaging within a gamified environment.

In our case, there are two key considerations. The first is determining how to combine the right types and amounts of player types to ensure the group functions effectively. This involves questioning whether it's necessary to actively shape the social dynamics within a group. For instance, we know that larger communities can naturally self-organise through incentives or status systems. However, in smaller communities, the question arises: is it necessary to appoint a leader or a community creator, or can the micro-community self-organise and naturally identify its leaders, socializers, and so on?

Our entire model was based on not having a designated leader or group initiator. Our assumption was that by continuously filling and continually replenishing pre-made thematic groups, each role would naturally find a leader, and the group would thrive.

In theory, at least to us, this approach seemed sensible. Creating and filling a group in this manner was meant to brute-force it into becoming a thriving community. In practice, we had to consider the issue of retention. This model led to a nontrivial average number of users churning before a micro-community would shape itself. 

Prior to diving onto retention and churn, let’s explore some insights about demographics and communities we’ve run.

On different demographics and community types

Gen Z users would send around 10 messages, while Millennials or Boomers typically sent only 1 or 2 a day. This stark difference in communication styles threw us off balance, later on, in adapting the product for Gen Z. The volume of input, attention span, and the app turn-over varies significantly between these groups.

Gen Z appears to be the most challenging to engage, as they constantly multitask, messaging dozens of people across several social networks simultaneously. They are keen on new apps but abandon them more quickly than other demographics. For Gen Z, creating a dense, engaging network seems crucial for retention.

In terms of demographics, we've learned that many markets are underserved, such as isolated individuals in their mid-40s. This demographic may not download many apps and might be averse to extreme novelty, but they often spend a lot of time on platforms like Reddit and other forums. They offer higher short to medium-term value but are less likely to spread the viral word to others compared to younger users.

Study communities were sporadic in nature. Students seek motivation, particularly before exams, looking for peers for Q&A sessions and to share similar experiences and struggles. 

Programming micro-communities mainly consisted of individuals seeking help with their code or wanting to share projects. There wasn't much daily sharing, but there was a clear need for companionship, especially since most were learning online, often alone. We know that the dropout rate for online courses is over 90%, indicating a significant yet challenging need to address. 

Art communities, notably those focused on drawing, were highly active. These communities seemed to lack other spaces to share their “work in progress” and thrived on having peers for daily sharing and feedback. 

Language communities presented an interesting dynamic. Small language learning group chats usually started strong but then declined as users lost interest. Most language learners were casual, often using apps like Duolingo, and tended to lose interest rapidly. However, Korean and Japanese language communities showed massive daily engagement, likely due to the “passion” nature of these languages and cultures. 

Fitness and weight loss communities were among the most engaging. Weight loss, a significant market, often involves isolated individuals with a strong desire for change but struggling with motivation or falling back into bad habits.

On retention and churn

We attempted to address the initial group formation churn with a combination of flintstoning and various minor hacks and tweaks. This involved us, being active in each group, at first. We engaged users immediately, to create a conversation history, which served as social proof and helped increase trust when the next user joined. Even when we managed to fill groups with dozens of people in a few hours/day, if a user wasn’t promptly nudged or welcomed within a few minutes of joining, there was a high chance of losing their interest, leading to churn. To facilitate engagement for new users, we increased the group size to boost the likelihood of interaction at the moment a new user joined, and we implemented easier ways for existing members to welcome newcomers.

The welcoming and introduction phase proved to be crucial. It didn’t need to be a formal process; we found that Gen Z members were content with simply saying “hi” and acknowledging each other, sharing more about themselves as conversations progressed.

Notifications were a significant challenge for us initially. Before we implemented radical changes, such as oppressive modals and daily activation requests, only 20% of notifications were enabled. We managed to increase this to 55%. Our product, relying on live interactions and users responding to others in public group settings, suffered from a lack of reactive users. In the latest version, we made an attempt at mandatory notifications, which boosted the rate to 95%.

Despite success in doubling Day 1 & 7 retention rates, this did not translate into improved Day 21 & 30 retention.

Initially, we focused on growing too many interests simultaneously. As a result, despite gaining hundreds of users daily, we didn't have sufficient resources to maintain this growth. Therefore, we narrowed our focus to a single interest, which unlocked significant growth. We quickly discovered new acquisition strategies, gaining four thousand users in just two weeks. After focusing on a specific niche, we started attracting younger demographics that had a higher churn rate.

We isolated three correlated factors: group size, demographics, and interests (or goals/ topics). The optimal group size varies between demographics; for example, what works for a Gen Z group differs from a Boomer group. Similarly, the ideal size for a Japanese study group is not the same as for a weight loss group.

There is an optimal range for these three criteria, which, when balanced effectively, leads to increased stickiness and engagement. Engagement also varies according to those. A community with lower engagement can still achieve remarkable retention. Some groups were so cohesive, with each member posting only once a day, yet they have remained active until now.

On lacking virality and novelty

We've come to believe that the issue boiled down to a single factor: our product was largely trivial and novelty-free. Although it addressed the needs of a significant portion of our acquired users, it lacked a K-factor or viral element to drive organic, snowballing growth. Merely being "good" wasn't enough. The absence of a "wow" effect was detrimental. Our simple chat interface, too similar to existing offerings, failed to stand out, and we couldn't find a way to overcome this challenge. Being good is not good enough; we didn’t find this unique mechanic, interface, or network element that sets us apart to pave the way to virality.

How did we end up here? Perhaps we tried to transition too linearly from our MVP on WhatsApp to a similar-looking app. We truly believed that the core value of the app lay in its live chat feature. However, it's now clear that while we did listen to early user feedback, we didn’t innovate enough on the product side.

We took radical steps, shutting down 95% of our community types to focus on just one, and later, we completely revamped the concept, to having groups open for just a few hours daily. The problem was that no one was talking to their friends about our app. One assumption is that we were connecting isolated individuals. If you don't have peers sharing the same interests, who will you tell about the app?

Now, as long as we continue to build in this space, designing with virality and shareability in mind will become a key focus. We need fewer screens, but each must be highly shareable, and have a “waouh” effect.

On learnings and moving forward

One lesson here, building a stranger-to-stranger consumer social app, is the necessity to find a mechanic that naturally increases the K-factor. We need to trigger the urge to share something novel even though it doesn’t directly add value to their experience within the app. 

Simplicity is equally important. A social app should start with a single, straightforward feature and focus on building a dense initial network around it. Initially, the app needs to be intuitive, allowing users to understand or discover its functionality quickly. 

When the target traffic doesn't stick or spread, it's futile to remain stagnant or attempt brute-force strategies for extended periods. The ability to quickly adapt and pivot is essential in the fast-paced environment of social apps.

The lessons learned will serve as a foundation for building the next iteration as each new product we release is a step towards making our first successful product.