Note: This post does not talk about transaction fee revenue models like ecommerce, app stores, etc. They are a big part of software monetization but I chose to focus on the other ones for today.

Traditional social media / ad based companies

Monetization is a function of time spent i.e. more time spent on the product == more revenue generated by the company — sort of like a linear graph (it’s not actually linear, I’m just over simplifying for the sake of this comparison). This analogy applies to companies like FB, IG, TikTok, etc.

(y = mx+c; m=1, c=0 in illustration below)

Take a look at Facebook’s ARPU over the years (source):

Here is how time spent on FB family products has changed in the last few years (source). Although FB main blue app time has decreased, it’s been matched by an increase in time spent on IG — at least till 2020, and relatively stable since then. Don’t have Messenger, WhatsApp and Oculus in this image because ad revenue from those products is still negligible for $META even though their time spent has dramatically increased in the last few years.

What this means is that in this case, companies are incentivized to keep increasing user time spent in the product to increase their profit margins. The more data they have about users, the more engaging content they can serve them. More the users engage, greater the ARPU because monetization scales linearly as a result of time spent in the product. Linearly scaling monetization means that these companies / products are on a never ending hunt for user attention.

Subscription products

Compare this to subscription products. Subscription products fall into a few broad categories. Products that have

• a free ad supported tier e.g. Spotify

• no free tier and users must convert to paid e.g. Apple Music, Peloton

• products that have discrete usage tiers e.g. iCloud storage

Subscription products like Netflix and Spotify also need user data to serve engaging content (think Netflix recommendations or Spotify’s Discover Weekly) to keep users coming back to the platform. However, the monetization function does not seem to be linear here. Imagine the follow scenarios:

• user who does 4 Peloton bike rides per week vs a user who does 6

• listens to 1 hour of Spotify vs 1.5 per day

• watches 30 mins of Netflix per day vs 45

In all three above examples, both users (for each Peloton, Spotify and Netflix) have crossed the barrier to remain a subscriber. They are engaging enough to build a habit of using the product and not unsubscribe. Monetization follows a function similar to

y = nlog(x) || y = {n, if a <= x <= b

where ARPU increases up to the point of the highest subscription tier and then stabilizes. The function can be discrete or continuous based on free tier vs not, and number of tiers.

This results in two main themes for subscription drive companies:

1. they are not hunting for user attention infinitely and ARPU is fairly stable once users are retained

2. their revenue is far more dependent on total subscriber numbers (declines in which is why Netflix stock has plummeted)

For subscription products with free (probably ad supported) tiers, once the free tier user population reaches a certain usage retention, they are very likely to self select into the paid tier to make their own experience better. So overall, ARPU tends to normalize asymptotically around the subscription fee.

In some cases, tiered subscription products (e.g. cloud storage, cloud computing, etc.) get prohibitively expensive in the higher tiers. This somewhat pits users and products against each other since price conscious users / entities are trying to use the product less while the service provider is trying to get more usage and thereby more revenue. This is why large corporations start building infrastructure inhouse because it becomes cheaper to build than buy the service once a certain scale is reached.

web3 monetization

web3 products largely don’t monetize with ads today. Instead of being separated into subscription discrete tiers, their usage price is determined by the supply and demand of the market (more like a commodity e.g. oil). This results in more elastic and transparent pricing for consumers but also increased volatility — for both consumers and businesses. The experimentation in this space is just beginning with new tokenomics models tested all the time — all in pursuit of finding a monetization scheme that works long term for all participants in the ecosystem.

web3 / crypto mixes technology, art, culture, economics and politics. I imagine web3 monetization will leverage all these axes to create a new business model — a mix of subscription (web3 domain registration fee), usage (chain network fee), time spent (web3 ads), financing (storing assets on platform), etc.

What models am I missing? If you have thoughts on the above, you can find me @rish on Farcaster.

1. differentiating from other social media products (web2 and web3)

2. growing the community while preserving quality of signups

Differentiating from other social media products

Context on social media moats

One of the real moats of large scale social media products is the “media” i.e. content you can only find on that product / platform / protocol. The size of the network and the availability of media feed off of each other. You can’t find Twitter content on FB (even though both allow text broadcasts with variety of attachment types), TikTok content on Instagram, etc. 

The media / content advantage is gained by a mix of first mover tactics coupled with great, and ideally unique, user experiences that enable good media creation. The content creation and consumption experiences get commoditized over time and thus, it becomes important for newcomers to:

• move first to provide new creation experiences (e.g. Snapchat stories) that builds an inventory of new media on the platform

• maintain an edge by continuously evolving new media creation trends (e.g. adding AR filters to Snapchat visual content)

Users flock to new media experiences that enable new interactions and even fight against network effects (e.g. Instagram creators starting TikTok profiles from scratch to move to TikTok like video format) for such media creation and consumption.

Farcaster’s moat can potentially be web3 native media creation and consumption at the protocol level that attracts new users as web3 mindshare grows in mainstream audiences. Unclear what the full extent of this will look like over the next few years (we are still in the skeuomorphic era of web3). However, Farcaster is already on this path and the NFT Feed today is one such feature on the primary client. 

Problem 

As Farcaster scales, new users (especially non crypto native ones) won’t join and/or retain on the protocol/platform simply because Farcaster is decentralized. They will join if Farcaster provides an experience that is not available on traditional social media*. 

Potential solutions

Web2 social media has optimized the in-app funnel for most web interactions (e.g. signup on a form, read an article, etc.). If Farcaster can build the web3 versions of funnels (ideally less skeuomorphic and more web3 native), then it can enable new experiences that are not possible in web2 social media products. Long term examples of this are hard to imagine immediately, some near-term examples include: 

• minting NFT from the cast if mint page link is shared

• donating to Gitcoin grant directly from grant link shared in cast

• submitting a bid on partybid

• adding money to a PoolTogether pool, etc.

To support experiences like the above, Farcaster needs:

1. “Dapplets” in casts — rich web3 link attachments that allow quick interactions with the dapp right within the cast (similar to how Spotify might allow playing the track right from the link attachment)

2. A standard that dapps can adopt to broadcast such information when dapp url is crawled, such that the cast can then embed a “dapplet”

3. Farcaster client that supports importing a full mnemonic (instead of just connecting a read-only address), allowing web3 txns seamlessly via imported address

Creating and gaining adoption of a new standard (#2) is a hard problem in itself. However, given increasing developer traction on Farcaster already, Farcaster standards might get strong adoption. Even without #3, read-only dapplets will be an improvement from where we are today. 

*PS - Some might say that new users will join because community on Farcaster is >> community on Twitter. That is very true today and a big reason why we are all here. However, I think quality will change as network size increases and that network size and node quality are always somewhat at odds with each other. Separate note on that below. 

Community growth

*Terminology (might not map fully to Farcaster protocol spec): Nodes: User, cast, onchain activity, etc. Edges: actions — like, follow, recast, watch, report, etc. *

Generally average node quality will drop as network increases number of nodes. Measuring this “quality” is a hard problem in itself. Facebook uses (or at least used to) a metric called meaningful social interaction (MSI) that measures the quality of a node in the graph (let’s say Node1) similar to Google’s Pagerank i.e. how many nodes share an edge with Node1 and what is the quality of nodes that share an edge with Node1 (measured recursively). This works — somewhat. When there are billions of nodes and manifold more edges, bad nodes reinforce themselves as much as good nodes do leading to, what we call, “filter bubbles”. Unclear if there is any real solution to this. Even physical networks of people create filter bubbles by choice (see coastal regions vs central US). If Farcaster nodes reach the size of FB, it’s unlikely that node quality will stay the same as today. Only way to preserve that is to create filters. Nodes will create their own filters and Farcaster clients will support such creation.