For someone like me who worked at Snap, LinkedIn, and built a social media analytics company (sold to WPP), the potential of Clubhouse was directly visible. The “drop-in audio chat” app can be called almost ‘mainstream’ after Oprah, Zuckerberg, Elon Musk, and even Drake appeared here. But the future success of Clubhouse depends on how it can help users find content and enable engagement. The answer is “through the use of user data,” but it might not be as easy.
‘What is Clubhouse?’ People have described Clubhouse as:
- Discord for old people (here)
- Next Snapchat for audio (Rex Woodbury here)
- Zoom with community aspect (here)
- Gen Z alternative for radio (here)
- First AirPods social network. (here)
- Experience of meeting strangers on ’90s chat lines (here)
- Ambient social interaction layer on top of…well everything (Bubba Murarka)
In essence, Clubhouse users can join audio rooms and chat with others. Some rooms host informal chats between friends, while others might have thousands silently listen to a panel with Oprah. Clubhouse that is now evaluated at 1bn USD has seen incredible user growth in just a few months. Many attributed this success to our need to find social interaction during the pandemic. A need that text messages and carefully crafted videos in Twitch or youtube could not fulfill. While true, this need can only partly explain this exponential growth.
Clubhouse has lowered the barrier to participate in conversations. Writing blog posts and creating good looking videos demand effort and skills. Joining a voice chat is simpler than shooting a selfie. Setting up a one-time room is easier than planning a podcast series. In short, the app vastly simplifies talk radio content creation.
Ten years before the internet browser was invented (yes, we are taking the 80th), Talk Radio became popular. The reasons are very similar to Clubhouse’s success: ease in distribution and proliferation of content. AM Bands allowed for more bandwidth to distribute content. Additionally, deregulation shifted content production from local radio stations to at industry scale content creation. Four decades later, the internet brought us the same two effects: Distribution of content at zero marginal cost and, in the case of Clubhouse, no COGS for radio shows or panel discussions. Clubhouse follows the rules of Internet 1.0 as @benthompson explained in social networking 2.0:
MORE FOR YOU
“Remember that the essential characteristic of v1 digital products is that they simply copy what already exists offline. For Facebook that meant digitizing connections between friends and family, and for Twitter it meant broadcasting conversations as if you were sitting at a bar. Such literal translations, though, have limits […]
What truly makes a category is v2: products that are only possible because of the unique properties of digital.”
In turn, that means that Clubhouse is not using the “properties of digital” to their full advantage. At the core, any social app solves two issues (1) to find and (2) enable social interactions. Play around with Clubhouse, and you will quickly realize that finding interesting conversations is not easy. You curate your own content by following people. This social graph is not unlike the graph Twitter used as a first approximation for your interest.
As we all know from Twitter, a social graph is a clunky way to curate your feed. A good Friend of mine told me recently: “My Twitter account is now set up so that I get the information I want.” He has been on Twitter since 2008! To be forced to update and adjust your following frequently is the opposite of ‘user delight.’ But worse, while at Twitter, we might quickly scroll over a message we don’t like, we might get stuck in a long, dull conversation in Clubhouse before we know it.
The challenges of Data.
To find the right discussion for you, Clubhouse would need to understand all conversations. Meaning it would need to live to transcribe them. Technically this is feasible. Speech to text algorithms has become very powerful. Take i2x.ai as an example (disclaimer: I am an investor): the founder @BrehmMichael created software that coaches call center agents in real-time by analyzing their conversations.
Clubhouse is currently not transcribing conversations. I assume that they do not want to impact the authenticity of the moment. It is the same ephemeral appeal that Snap had initially. Please note that neither Snap messages nor Clubhouse conversations are guaranteed to be ephemeral. (For example, recordings from a sexual moaning session in Clubhouse surfaced on Twitter.)
The power that data could have for Clubhouse growth can be seen if you look at TikTok, an algorithm first company. Unlike Clubhouse or Facebook, it does not mainly rely on the social graph but has built a powerful recommendation engine. As @eugenewei wrote in TikTok and the Sorting Hat:
“Bytedance has an absurd proportion of their software engineers focused on their algorithms, more than half at last check. It is known as the algorithm company, first for its breakout algorithmic “news” app Toutiao, then for its Musical.ly clone Douyin, and now for TikTok.”
TikTok’s machine learning magic is a supervised learning system that will find videos that are interesting to the user. It will measure how long you look, how fast you scroll, and many other features of your behavior for the algorithm to work. Additionally, the algorithm will evaluate the actual video clip: content, speed, sound, colors, words, and many more features. For this algorithm to work, TikTok needs data – loads of data. To give you some scale, only to create a system that can identify what is in an image, we need 14 million classified training photos (like imagenet)
To collect the data has been a product choice. At its core, TikTok’s has been designed with this data collection focus in mind. Today, the algorithm is by far superior to the recommendation algorithm from Facebook. Moreover, this data has also created a sizable moat that not even Facebook could cross so easily.
As Clubhouse grows, their design choices will be necessary. How can they record the liking of a conversation? Clubhouse is a passive medium. That means users might not directly react to a conversation. Streaming companies like Spotify or Deezer can tell you that a delay between the user’s dislike and an action such as ‘skip’ makes it way harder for the algorithm to work correctly.
To tip content creators could be one way of measuring user delight, but this might not be common – especially not in the beginning. At my own company, I found that the spread of content across different platforms was the best signal for high engagement. This insight seems to hold also for Clubhouse. Already today, you can see the best parts of Clubhouse conversations on other networks such as Twitter and Facebook. Therefore, I expect the fiercest competitors to be Twitter and Facebook as they do not only bring the existing social network graphs but also measure how content spreads.
Despite those challenges, the trend Clubhouse started is undoubtedly not a covid19 fad. It’s the unbundling of content creation and distribution. Talkshow Radio is no longer a domain of a few corporations, and the internet has made distribution virtually free. Whether or not Clubhouse will become what @benthompson calls a super-aggregator will depend on their ability to create a multi-sided network with decreasing acquisition costs. Therefore, it’s no surprise that Clubhouse announced that they would invest in creator tools and an ability for content creators to get paid. In order to scale, they will need data and engagement metrics; otherwise, they will become something like a substack for audio – which would be a success in itself but below clubhouse’s potential.