• crowd funding announcement

• using statistics/ML to measure a way to increase retention

• general growth metrics of farcaster

Crowd funding

Purpose - I'm analyzing Farcaster data to help builders grow high-quality users and retention through data analysis and statistical modeling. My research will be shared openly to benefit the entire Farcaster community.

Kinds of Questions i am looking to answer -

• Which activities increase quality user retention on Farcaster (e.g., welcome messages)

• Impact of airdrops on user engagement and retention

• Correlation between onchain activities (token/NFT holdings) and social behaviors

• Data-driven strategies for maximizing quality user growth

• etc

i will try to answer as many questions as i can working full time.

Why builders should care -

Builders want more high-quality users on Farcaster for their projects. By looking at the data from all the possible angles, they will have an easier time in deciding what THEY can do to maximize quality user growth. I believe that there is a lot on the table has not yet been explored from which builders could learn and explore.

My ask -

I believe that farcaster does have the potential to be the next big thing if done right. Which is why i believe that investing my time in Farcaster is the best course of action. i have been analyzing farcaster's data for almost 2 years however it would be impossible for me to go full time in analyzing Farcaster if i can afford to sustain myself.

If my crowdfunding reaches $2,000, I'll dedicate myself full-time to analyzing Farcaster data for a month, with the option to continue further.

Donate to the crowd fund here - https://warpcast.com/yesyes/0xb27126f1

I will weekly share all the insights i find in the form of an article such as one below(but more detailed) -

Impact of quoting/tagging a new user

Basic question being asked - Does quoting or tagging a new user in their first week of joining increase the likelihood of them being retained after 30 days?

rish asked me to check the retention of people onboarded by big accounts. I decided to check that. But then i realized that i could also just generally check the impact of mentioning or quoting someone in their first week joining. This will be helpful in deciding whether it's a good idea to pursue this behavior or not. So i decided run some analysis on that -

I check for those fids that were mentioned or quoted by a user in their first week of registering. And then i check whether they had casted after 30 days(but before 60 days). This analysis only takes into consideration those fids that casted at least once in their first week of joining. This contains fids from 100k to 900k for individual analysis.

Let's look at the performance of some top accounts -

interesting that linda had 21 users mentioned for sub 800k fids(and still 20 retained users. meaning the 4 users from 800k to 900k were not retained at all)

It's generally safe to say being that tagged or quoted by a big account greatly increases their chances of sticking around.

The general MoM casting retention of a new account is about 65%(see retention details later)

Now, Taking analysis this to a more general level -

I looked at the data of fids from 100k to 800k FID (FIDs 801k to 900k were left out for testing and cross-validation purposes in future analyses.). The fids that casted in the first week of joining were considered in this analysis. There were a total of 327,788 fids.

Out of these -

• 109k were mentioned or tagged by another user in the first week

• 218k were not mentioned or tagged by another user in the first week

out of the 109k tagged users

- 75k casted at least once after a month

- 33k did not cast even once between the 30 to 60 days.

out of the 218k non-tagged users

- 122k casted at least once after a month

- 95k did not cast even once between the 30 to 60 days.

Looking at the retention from this -

Those that were mentioned had a retention rate of 69%. Those who weren't mentioned had a retention rate of 56%.

Presenting the data in tabular form -

• Users who were mentioned or tagged by others in their first week showed a 13.2% higher retention rate compared to those who weren't mentioned

• While mentioned users represent only one-third of new users, they account for 38.1% of retained users

Given the large size, it is safe to say that this difference in rates is statistically significant. However, i will further perform chi-squared test and logistic regression to further understand how significant the difference is and whether mentioning people early should be encouraged more.

Chi Squared test

Here is the contingency table for the above data -

Without getting into the nitty gritty - A chi-square test is a statistical method used to determine if there's a significant association between categorical variables.

Null hypothesis - There is no significant association between being mentioned in the first week and activity after a month.

Skipping the calculation part, Here are the results of the chi-square test -

Chi-Square Test Results: Chi-Square Value: 5277.2128

P-Value: 0.000...(far less than 0.05)

Cramer's V: 0.1269

What this essentially means is that since p-value is much lower than 0.05 then it is safe to say that there is a significant association between being mentioned in the first week and activity after a month.

(hence we reject the null hypothesis)

However, a low cramer v value suggests that the association is not too strong.

Logistic Regression

I also decided to run logistic regression on the data to determine a rough increase in odds after being mentioned/tagged in the first week of joining.

Here are the results of training the model

Again, the p-value in this Regression is also very close to 0 which means that, again, the results suggest that quoting/tagging someone in the initial days does cause statistically significant increase in retention!

Here is the odds ratio - 1.7635

what this essentially means is that users who were mentioned in their first week are 1.76 times more likely to be active after a month compared to users who were not mentioned. This represents a 76.35% increase in the odds of being active.

Conclusion of the analysis

Overall, it is obvious that quoting/tagging someone in the first week of their joining increases their likelihood of being retained a month later. The difference it makes is statistically significant though not super substantial.

Still, it is advisable to quote/tag a new user if you want them to stick around. It will definitely help in retaining them.

General metrics

Now, let's look at some general metrics regarding Farcaster's growth.

DAUs return back to their baseline

The DAU figure seems to have returned to baseline after seeing a nice spike in March. The spike was caused by new and returning users(because the storage rents also went up but not that much).

This spike was most likely caused by the aggressive marketing push by Farcaster. Stuff like 'Make money on Farcaster' and the push for airdrops brought eyeballs towards farcaster.

Unfortunately, These users weren't very sticky and faded after a while. Understanding why these users left could be key to understanding how to make Farcaster grow.

Personal opinion - I think most people tend to leave the app because they fail to get any engagement here. It could also be that they don't find the content here engaging. This is a cold-start problem that every new social media faces. Regardless, i believe giving these new people a quote or tag if you know them will be of great use in retaining them!

It is also possible that using onchain metrics could help people in find like-minded people(miladies want to follow miladies, people who play onchain games likes pixels,runiverse,etc also want to other players, collectors of a particular song/article would want to follow people with similar taste, etc. The key would be look at assets held because of taste and not due to financial interest.

I hope to quantify and verify this stuff in the later articles.

new user storage rents show a troubling picture

New user sign ups are at the lowest point since going permissionless. This is a bit concerning. What's interesting to note is that Farcaster seems to have switched from 'make money on farcaster' to advertising its mini-app ecosystem as well its builder friendly nature. So perhaps, this marketing will pay off more in the long term.

General MoM retention -

Notice that caster cohorts are getting big again after being tiny around october. The retention of the recent cohorts seem strong but this does not yet include April data which will probably shed some important light on the retention of these cohorts.
We will soon be revisiting this chart with April data. Stay subscribed for that!

This was a demo article for my crowd fund. I hope to release longer, more frequent articles with deeper insights if i get crowd funded. Please consider crowd funding if you found this article insightful and want more insights like these.

This article will provide a dive into Farcaster’s data related to –

•        Farcaster’s recent growth

• Solana data of collected wallets

•        Analysis of the recent airdrops($HIGHER)

Note - This is a partial article. Full paid article($9) is available for my hypersub, along with many other benefits such as regular posting of Farcaster analyses and access to GC! So please Subscribe!

Reach out to me on yesyes if you have subscribed and i dont reach out to you soon!

The daily new sign ups have exploded recently, however they are still not near February’s numbers. It is safe to say that the recent sign-ups have been due to the airdrop announcements.

Also, if you look at the monthly chart of the same, you’ll see that the month of march is not that special despite the spike in sign ups –

As you can see here, the DAU figure on 19 March 2025 with 64k users got close to the largest peak of 73k user.

It is interesting to note that this spike in DAU happened before the spike in new users (which happened on 23 March) which suggests that this spike in DAU number is due returning users and not new users.

(note - The DAU is calculated as the number of distinct users that are posting or commenting)

Monthly Active Casters of March(215k) present a very positive figure. This is the highest MAU figure since July of 2024 which had 238k MAU and the month of march isn’t over yet! Things are seemingly headed in the right direction. Although, one issue is that a good chunk of this number is due to the airdrop announcements so it remains to be seen whether these new and returning users will be sticky and high quality or not.

It is a safe to say that the recent spike is due to the airdrop announcements because there has been a sharp uptick in the wallet connecting activities.

The first peak in February is due to the introduction of Warpcast wallets(warplets), the second one is due to the airdrop announcements

It becomes more obvious when you look at just solana wallet connections. Lots of new connections in the recent period

I will provide Monthly retention analysis at the end of March to my subscribers on hypersub.

I will also closely take a look at who these re-activated users are.

So please subscribe and join my group chat if you haven’t yet!

The increase in connected Solana wallets on Farcaster also gives us an opportunity to look at the data of the activities and the nature these wallets.

There are 155K solana addresses connected to Farcaster belonging to 154K FIDs.

•         SOL- 49,460 ($ 6,776,020)

•         USD Stablecoins – $ 5,528,218

•         Other tokens- $ 9,648,549 (the dune prices table is messy so this is taking the lowest prices of tokens 10 days ago)

•         Total balance – $ 21,952,787

It is important to note that the SOL balance of users who casted in the past 10 days is a little more than 10K.

This is actually a pretty positive figure because most active users tend to lurk a social media rather than posting on it.

Again, this uses lowest token prices from 10 days ago.

As you can see, Farcaster users love FARtcoin. Other top holdings include – JitoSOL,$WIF, ai16z,etc.

The distribution of money across these tokens is very interesting. Fartcoin is almost in a different league as compared to the rest of the tokens.

Now, let's who are these people on Farcaster that are holding heavy bags of SOL, Fartcoin,JitoSOL, etc.

This section is only available to hypersub subscribers!

NOTE - Dune dex table does not yet have the pumpswap data. But this list will be updated as soon as the pumpswap data is added(which is expected soon)

Of the given projects, Raydium easily remains at the top, however even it has 14k distinct wallets in the past 30 days which suggests that most wallets are not active traders(atleast not outside pumpswap)

Farcaster team has recently started focusing in helping teams of different tokens  airdrop their tokens to Farcaster users. Farcaster has had multiple tiny airdrops($HIGHER,$jam, etc) and more airdrops are planned in the future.

This section will focus on the analysis of the airdrop of $HIGHER which started a little more than three days ago.

(note –the intent of this article is not to shun anyone for selling or transferring their airdrops, but to merely check the effectiveness of airdropping to FC users)

Since then, The $HIGHER airdrop has been received by 1000 wallets, each getting 600 $HIGHER.

How has the holding performance been?

i will look at metrics such as - How many are still holding the full sum, how many fully sold, how many were holding higher prior to the airdrop, have airdropped users bought more, how much was sold etc

•     688(69%) out the 1000 wallets are still holding 600 $HIGHER or more

Rest of the airdrop related insights are only available to paid subscribers!

If you are a project thinking of airdropping to Farcaster users then please subscribe to look at the performance of this airdrop! It would help me a lot to provide regular free and paid farcaster and web3 related insights!

In 2024, Farcaster grew by 712,000 new users—but its biggest growth month may have little to do with Frames or Degen. Such insights and more will be looked at in this article. Let's dive into the data.

This is a partial article. Subscribe to my paid hypersub to get more alpha and insights -

• Total new users gained in 2024 – 712k

• December 2024 saw a resurgence with 36.5k new users, the highest since July.

• Interestingly, June(with 133k new users) is the month with the most new users not February(which had 127k).

• June’s growth may have been due to a perfect mixture of a variety of events – popularity of degen, frames and mainly due to the 150M $ fundraising announcement in late May.

• Enough evidence that the funding announcement played the main role will be presented later.

Total FIDs that casted at least once this year (including deleted casts) – 533,989

• DAU peaked in July 2024 at 73,700 but dropped to 36,000 by year-end.

• Despite the drop, year-end DAU was similar to February’s peak of 40,000

Note - DAU/ Daily active Casters here mean unique users that either posting or commenting

• MAU peaked in July 2024 at 238k and ended the year at 195k (down 18% from the peak).

• MAU performance was stronger than DAU and WAU, indicating that users are still engaging with the platform monthly, even if not daily or weekly.

  ---

• Daily ticker-posting casters peaked at 28,000 in February but dropped to 7,000-8,000 by year-end.

• The recent addition of clickable tickers has not changed things

• Channel Dominance:

  • Early 2024: Over 70% of posts happened in channels.

  • September 2024: Less than 50% of posts were in channels.

• Channel Changes:

  • Channels became membership-only in October, leading to a significant drop in usage.

  • Introduction of public channels failed to revive engagement.

•   The month of April has the weakest 6 months retention(20%)

•     The months of may and June were big yet they have the best 6 months retention after January(which was the  smallest month)

•    June, not Feb, was the biggest cohort month of new casters

• Popular Farcaster channels include mainstream topics like frames, base, memes, farcaster, and degen

• Many less prominent channels gained temporary popularity, including perl, farcastles, zora, drakula, goat, paidgroup, masks, scoop, wildcardclub, airstack(moxie), lunchbreak, and scout-game

• The data reveals numerous trends on Farcaster, with most remaining contained within the platform regardless of their intended reach

• After October's channel changes, top channels experienced a significant decrease in post volume, though the number of unique posters remained relatively stable

• New trending channels emerged post-changes, including itookaphoto, rodeo-club, aether, pvpgame, moxie-korea, spotlight, and yoink, which have maintained steady performance

• Trending words follow similar patterns to channel popularity trends

• $degen emerged as a trend in January, peaked between late February and April, then saw reduced usage from late June through August

• Other notable trending terms and their peak periods:

  • frame/frames: Late January

  • farcastle-related terms: Late February

  • Drakula-related terms: March

  • goat: Early April

  • moxie: Late July

  • basenames: August

  • trump: November

Some of the major trends and changes that led to FC's growth were

• Degen (main boost in Feb and April)

• Frames(main boost in April)

• 150 million dollar announcement(boost in May, June)

• Moxie( boost in July) - Most people fade Moxie but it did play a minor role in boosting the numbers of new users

The weekly active casters in the week of the funding announcement shot up from 84,865 to 119,422 and kept increasing for next few weeks.

How am i SO sure that it was the 150M funding announcement(and not $DEGEN or frames) that brought in the new users? And what does it mean for Farcaster's marketing strategies?

Subscribe to access the full article -

The full 24 page Article Includes -

• Looking at AI agent mindshare on FC

• A deeper look into trending channels, words and trending tickers over the year

• Analysis of how different trends attracted different types of users

• More Retention charts and analyses!

• And more!

I will be looking deeper into - AI agents on Farcaster, FC X solana data among other things. So subscribe now to get the alpha!

You can also subscribe to this paragraph to get free partial articles!

The analysis will be mostly restricted to sub 100k fids. Farcaster hit that number soon after going permissionless. Though many of these 'FIDs' aren't actual users i.e. they don't own storage(not that they're spammy).

caveats - The farcaster network is continuously growing and evolving. this means that there might be several causes for the changes in user behavior over time. Ensuring that the correct cause is attributed to the correct impact is an important but difficult task. As such, several different methods are used to explore farcaster's data and a decent attempt is made to normalize the data.

All of the analysis is done in UTC

Directly answering what was asked in poidh.

Analysis should look at # of threads with comments by 3 or more unique FIDs before February, then the same metric after February (excluding bots and tipping comments) - analysis should be restricted to only power badge users or pre-100k FIDs to normalize data - the fundamental question is "what were the odds that, if a real user responded to a real user pre-February, that a third real user would then join that conversation" - after that question is answered, we'll want to see how those odds changed post-February

All the analyses don't count tipping comments(ones with $ sign) unless stated otherwise.

Firstly, i look at the daily number of replies received to a comment on a post. In this case all three people(the original poster, the commentor and replier of the commentor) are different people and have fid under 100k. this is what the data looks like. As you can see- The number of replies have gone down. Interestingly enough - the sub 100k fids are more active now than they were in december/jan so the number being lower is all the more surprising

One thing to note is that this data also includes those cases where the third person is actually the fourth commenter or is replying to a reply of a reply of original and so on. Their data is included as long the replier, the commenter he is replying to and the original poster are all different people

From here on out i will call a reply to a comment as a 'reply-comment' or 'second comment'

The reply to the original post may be referred to as 'first reply'

Okay what if the number of replies by sub 100k fid to sub 100k fids are down because the people are replying to greater than 100k fids as well?

This is indeed the case -

But If you look at the percent of 'reply-comments' vs total posts then it paints an interesting insight.

As, it can be seen - the percentage of reply comments have been reduced as compared to the total number of posts. In short - a sub 100k fid person is less likely to reply to a comment(Irrespective of the commentor fid). So this essentially means- although there are more active sub 100k users now than there were before. The percent of the 'reply-comments' have gone significantly down since early February which is when reply-bumping was removed

What do these 'reply-comments' look like? - If i only look at the data for sub 100k fids then the 'reply-comments' aren't spammy and surprisingly most of them are pretty interesting.

Though if i allow any fid then the comments become very spammy

What percent of comments by sub 100k FIDs received a reply by another sub 100k fid? The percentage has drastically gone down. So while sub 100k are more active than before, they are less likely to reply to other's comments. You might be at a disadvantage at having your content remain undiscovered if you mainly use farcaster by commenting.

Though it should be mentioned that your odds of receiving a reply(from any user - not just sub 100k) are practically the same even if tipping comments are filtered. Though i am assuming that some spam still goes through(see the spammy comments later) so the actual overall odds are lower.

If no FIDs are restricted then this is what the data looks like -

Now let's take a look at what these unreplied replies look like for sub 100k fid users -

Without tip filtering-

With tipping filtering -

What are the chances that a third distinct user would reply that happens to be mutuals with the first-replier -

if no fid restriction-

If only the first replier's fid is restricted to sub 100k-

Number of tipping comments over time -

Replies(excluding self replies) vs original casts by sub 100k FIDS(excludes tipping comments)

What is the weekly number of people that an average sub 100k fid replies to. The median is currently sitting at 3 increased from 2 since jan 24. This ofcourse only counts those users that are atleast replying once that week. So if a user is inactive in a week then it isn't taken into calculation.

The median number of users replied to has gone up from 2 to 3.

What the data looks like if there's no FID restriction(tipping comments still excluded)-

The median number of users replied to sit around 3-4

Now, i'll look at different feeds randomly.

Looking at people's chrono feeds with and without tipping at random points

Kenny's feed -

Keccer's feed -

My personal conclusion to the the data shown are that -

• Farcaster went heavy in killing reply bumping in february but decided to change the algo again somewhere around in may to somewhat show replies though still not to the same extent as it once used to show. It is very clear from the data that the algo is being tweaked to somewhat favor replies again

• I am paraphrasing here but i remember Dan mentioning in one of the late night night crew taverns that reply bumping is not being reinstated because it would only be helpful to power users of the app and not the general users. Farcaster currently is trying to properly tweak its algorithm for the general audience

• The data shows that the relationship between people on farcaster is getting closer to being that of consumer-producers as opposed to that of being friends though it is still better than most social medias today . This is not necessarily a good or a bad thing, it is just the direction in which farcaster is going.