What We Can Learn from an Onboarding Ritual That Outlived Its Tag
Foreword — by 
In my recent posts, I've been digging into user retention data and analysis with the help of Claude. In the background, I'm also actively working on solutions to our onboarding and user retention malaise. Hopefully by the time this blog post goes up, one part of those solutions will have launched.
A key insight in the last post was the importance of making an introduceyourself post. Today I set about looking into that more deeply, and again I would like to share the results of that research here.
Like other recent posts, the rest of this post is written and researched by Claude, guided, directed and challenged by me.
Research Findings — by Claude Opus 4.6
The following analysis was conducted at
Part 1 — A history of #introduceyourself
#introduceyourself is the oldest onboarding convention on the Steem/Hive blockchain. For most of the 2016–2019 era, the expected way a new account announced itself to the community was a post carrying this tag. At its peak the tag produced over 40,000 posts in a single month. As of early 2026 it averages fewer than 100 posts per month. It has not quite died; it has been reduced to a vestige.
This is the story of how that happened, and — more interestingly — what it tells us about what actually drives newcomer retention on Hive.
The genesis (May–July 2016)
The first post filed under the tag is from 2016-05-29: @mummyimperfect — "Hello Steemit. I am Imperfect!". It earned 1,212 SBD (Steem's pre-fork dollar-pegged stablecoin; the Hive fork of March 2020 replaced SBD with HBD — I treat both as $1-equivalent for purposes of this post). For context, SBD at the time was trading near $1 USD, so that's about the equivalent of a full-time monthly salary in many countries. Every intro post in the first week of June 2016 earned between $280 and $2,560.
A couple of weeks later, @cryptoctopus published "The Secret Formula to a Successful IntroduceYourself Post…And The #1 Mistake to Avoid" (2016-06-15, $4,148). A meta-post about the ritual itself earning a top-50-all-time payout tells you everything about the gold-rush atmosphere of that summer.
The tag's biggest single moment came in July 2016: @jl777 — "steemit is crypto's first mass market solution!" earned $22,435 on 2016-07-14. It remains the highest-earning intro post of all time; not a single intro in the eight years since has come close.
Other early mega-payouts were a who's-who of crypto, media, and activism crossing over to Steemit:
@neilstrauss— "Introducing myself, finding your passion, and who wants to start a Steemit Book Club?" (author of The Game), $13,186@falkvinge— "The founder of the Pirate Party joins to share liberty ideas with Steem", $14,129@charlieshrem— "Charlie Shrem Is Now On Steemit!" (early Bitcoin figure), $10,317@brendazambrano— "Hi! I am the first Playmate with more than a million followers to blog on Steemit!", $12,143@federicopistono— "Hello Steemers! My name is Fede…", $15,311@dollarvigilante— "The Dollar Vigilante Is Now On Steemit!", $15,968
And a particularly Steem-flavoured one: the plagiarism bot introducing itself. @cheetah — "Hi Steemit, it's me, @cheetah. I was born to be wild, RAWR!" (2016-07-31, $4,057). The plagiarism-detection bot that would police thousands of other intros for the next four years had its own intro post, and the community tipped it well.
This economy was possible because Steemit Inc's founders — and a handful of large early stakeholders — were personally curating intros. @ned (Ned Scott, Steemit Inc CEO) upvoted 444 intros in 2016. @dantheman (Daniel Larimer, Steemit Inc CTO) voted on 192. @smooth — not a founder but one of the earliest and largest miners, and among the biggest stakeholders of that era — voted on 197. For the first six months of the tag's existence, the combination of founder-led and whale-led curation ran what was effectively a dedicated intro-curation programme — and the posts above are direct evidence of that programme's output.
The gold rush (late 2016 – early 2018)
Word spread. Monthly intro volume climbed from 700 in November 2016 to 13,800 in June 2017 to a single-month peak of 40,688 in January 2018 — roughly 9% of all new accounts created that month filed an intro post within their first 30 days. This was the high-water mark for universal ritual adoption.
But the economics had already collapsed before volume peaked. Average payout per intro post fell from $707 in June 2016 to $123 in July 2016 to $7.88 by December 2016. By January 2018 it was $3.17. The gold-rush phase of the tag lasted roughly six months in absolute terms. For the next two years, the tag's reputation as a lucrative onboarding ritual survived long after the rewards actually did.
Abuse: scams, impersonations, and policing (2017–2019)
Once the tag was known as the place where new users received attention, it was abused. Targeted searches of intro-post authors with famous names reveal a catalogue of impersonations. A few on-chain examples:
@realdonaldjtrump— "Folks, I am Donald J. Trump and I'm PROUD to be here with you today." (2016-07-27) — zeroed to $0.@markzuckerbergs— "Introduction" (2017-12-30) — net-negative rshares, zeroed.@justin-bieber— "Justin Bieber here to rock Steemit!!!!" (2016-08-31) — zeroed.@officialmcafee— "A little hello to the Steemit community from John ;)" (2018-02-22) — zeroed.@elon-musk— "Hello, I am Elon Musk" (2018-03-07) — zeroed.- A whole genus of
@satoshi*and@*nakamotoaccounts, most zeroed.
The biggest-downvoted intro posts of the era are instructive:
@truesteem— "Hello Steeemit! I'm Dayana Intoduce Myself" (2017-09-14) received 76 upvotes and was then downvoted to −130 billion rshares — one of the largest anti-abuse downvotes in the entire dataset.@saqib-bashir— "Introduceyourself." (2018-12-09) drew 236 upvotes on a copy-paste intro before being zeroed.- A dozen accounts (
@stewartcaliana,@fulerit, and others) each collected 90+ upvotes from small accounts — the fingerprint of a vote-buying ring or a botnet — and were then flagged to zero by a handful of stakeholders.
Policing took two forms, and both were active:
Downvoting. 4–8% of intro posts received at least one downvote during 2017–2019 (via TxVotes.weight < 0). The anti-abuse network of the era — @spaminator, @cheetah (plagiarism detection), @steemcleaners, @buildawhale, @guiltyparties — commented on thousands of intro posts each year. In 2018 alone, @spaminator flagged 1,145 distinct intros. @cheetah touched 861. @steemcleaners ran identity verification on 762.
Social-proof verification. @steemcleaners ran an explicit identity-verification programme, with standardized templates asking newcomers to "confirm your authorship" by linking back to the Steemit account from their LinkedIn, Instagram, personal website, or Twitter. Their canonical post, linked in every verification request, was titled "Introducing Identity/Content Verification Reporting & Lookup" and explicitly named the threat: "identity theft, identity deception of all types, and content theft." Individual community members participated too — @carlgnash asking newcomers for "a picture of yourself holding a handwritten sign with the date and your steemit username on it" is a classic example of the social-proof ritual.
Where the policing leaked: fraud that paid out
The cases listed above are mostly the easy ones — obvious celebrity impersonations caught and downvoted to zero. The harder cases were intros that collected upvotes before the identity-verification queue reached them. Some were celebrity impersonations polished enough to fool casual voters; others were plausible-looking unknown identities. Curation moved faster than verification: large stakeholders actively curating intro posts typically voted within 1–3 hours of posting, while @steemcleaners' identity checks sometimes took days. A few documented examples from 2017–2018:
@djkhaled— "They don't want us to win… and that's why I'm here on STEEMIT!" earned $110.47. The post used professional media photos of real DJ Khaled and even cited "Sources" for the stolen images.@steemcleanersasked for verification via DJ Khaled's real Instagram — no response. The account went on to post four more times over the next twelve days, including a "Win FREE SBD" meme challenge, before going silent.@gchan129— "Introducing Gchan129 to the Amazing World of Steemit" earned $78.99.@steemcleanersasked twice for identity verification, was ignored, and concluded "your account is a case of fake identity and will be submitted to blacklist." The account was later banned by@cheetah. The intro payout stood.@irwanumpalearned $66.20.@steemcleanersconcluded it was "another case of fake identity."@etoil— "Mathias" earned $24.00 despite@cheetahhaving flagged the account as "on my black list, likely as a known plagiarist, spammer or ID thief."
Beyond these confirmed cases, a larger set of intros received the standardized "PLEASE, CONFIRM YOUR IDENTITY WITHIN 24 HOURS" demand from @steemcleaners and went unanswered. Some of those silent authors were probably fraudulent; others were probably real but disengaged users who never came back. Either way, the seven-day payout window had already closed by the time the question was settled.
The underlying timing mismatch is worth naming: stakeholders curating intro posts were making vote decisions within an hour or two of posting, based on the content visible at the time. Identity verification — which required the author to respond, link external accounts, or produce a handwritten sign — inherently took longer. Policing was effective at preventing future earnings from banned accounts, but the first intro payout was a one-shot game where the curation queue and the verification queue ran on different clocks. The systemic-level statistics (4–8% downvote rate, hundreds of identity-verification requests per month) describe a functioning moderation regime; the per-post mechanics still allowed plausible-looking fraud to collect $50–$100 before being identified.
A representative modern intro
For a sense of what a successful intro looks like today, @misterrogers — "Kiss Me, I'm Irish (…by Deportation): My Hive Introduction" (2022-03-22, $237) is a good example of the small-N, high-quality era the tag now occupies. Long personal story, significant community engagement, earned meaningfully because a few curators saw it and voted deliberately.
The Hive fork (March 2020) and the slow fade
The Hive fork of March 2020 didn't directly kill the tag — the decline had been monotonic from 2018 onward — and most of the Steem-era automated-welcome infrastructure actually survived the fork intact. The most active welcome bot, @esteemapp (eSteem/Ecency's legacy account, which had commented on 4,202 distinct intros in 2019 alone), handed off to its rebranded successor @ecency on 2020-07-06 — literally within hours of its last comment. @ecency then continued the welcome-bot function at meaningful volume for another two and a half years: 483 comments on intros in H2 2020, 898 in 2021, 409 in 2022. It collapsed in 2023 (33 comments) and was effectively retired in 2024. Similarly, @cheetah continued its plagiarism-detection commentary on intros through 2020, and @steemcleaners rebranded to @hivewatchers which still runs social-proof verification at low volume — active through 2026, but focused on ~30 flagged intros per year rather than the ~1,000/year of the peak.
In other words: the 2020 fork wasn't the moment the automated welcome infrastructure disappeared. The welcome bots kept going, rebranded. What happened later — quietly, over 2022–2023 — was that these bots were retired, as their operators' priorities shifted elsewhere (Ecency to Points/boost curation inside Ecency's own community; Hivewatchers to higher-priority abuse cases). The intro tag's infrastructure was kept alive by inertia for a couple of years after the fork, and then dismantled without fanfare.
From 2021 onward, a different pattern took hold. Mass account creation events — most notably the Splinterlands onboarding surge of Aug–Oct 2021, which created over 700,000 new accounts in three months — produced essentially zero intro posts (fewer than 200 from those three months combined). Hive's account population was no longer predominantly bloggers.
Through 2022–2026 the tag has flatlined at 100–300 posts/month. It still functions; it just isn't the universal onboarding ritual anymore.
What replaced it
When you look at where newly-created accounts actually post their first post in the Hive era, the answer is clear: communities inherited the function. In 2022, the top destinations for new accounts' first posts were:
- OCD (hive-174578) — 1,821 first-posts
- Threespeak (hive-181335) — 1,692
- LeoFinance (hive-167922) — 682
- GEMS (hive-148441) — 622
- Aliento (hive-110011) — 607
- Ecency (hive-125125) — 498
- introduceyourself (tag) — 395
The tag ranked seventh. Further down the list at 128 first-posts sits the "Introduce Yourself" community (hive-109667) — a community-scoped replacement for the tag, with moderators and dedicated curators. Meanwhile Aliento (hive-110011), the Spanish-language onboarding community, had already attracted more new-account first posts than the intro tag by 2022.
Part 2 — Data analysis
This section presents eight main findings, each supported by a direct HiveSQL query. All code and CSVs are available on request.
A note on methodology: retention throughout this post means posting any top-level post 90–180 days after account creation, for accounts grouped by cohort year and by what their first post looked like. This follows the same cohort approach as
Finding 1: The tag's volume collapsed 99% from its peak
The chart shows the full lifecycle: the 2016 genesis spike, the 2017–2018 boom/spam era, the 2019 contraction, the modest 2020–2021 rebound, and the long slow decline through 2026. Total posts: 137,537. Total distinct authors: 76,488.
Finding 2: Adoption of the ritual collapsed even among bloggers — it's not just a population effect
Raw adoption (intros per new account) fell from ~9% in 2017–2018 to <1% by 2023 — a ~70× decline. But some of this is because more of Hive's new accounts are non-bloggers (Splinterlands players, app-specific accounts). Restricting the denominator to accounts that actually made a top-level post in their first 30 days ("new bloggers"), the decline is still ~6×: from 19.6% of bloggers in 2017 to 3.1% in 2023. Both effects compound, but most of the collapse is real behavioural abandonment of the ritual, not just a composition change.
Finding 3: The stakeholders who supported the tag left the platform — and the next generation never adopted it
Two things changed between 2016 and today, and it's important to keep them separate. First, the specific accounts that ran the 2016 intro-curation programme either withdrew from the chain or left it entirely. Second, the new stakeholder cohort that took their place never adopted tag-level curation of intros as a priority. Both matter, and the "stakeholders withdrew support" framing needs to acknowledge both.
The original curators' own history:
This isn't "the founders got bored of the tag" — it's "the specific small set of accounts that ran the 2016 programme stopped being active on Steem/Hive at all." Dan Larimer's exit to EOS in mid-2017 is the clearest single event; Ned Scott's disengagement and the 2020 Hive fork completed the withdrawal. By the time a new generation of stakeholders was in place (post-March 2020), none of the original curation-delivery accounts were still around.
The new stakeholder cohort's relationship with the tag:
The post-Hive-fork stakeholder set is substantial — and different in character. It includes:
(early whale, still one of the largest stakeholders)
(vocal individual stakeholder, operator of 3Speak)
,
,
(stake-delegated curation projects)
/
(frontend + onboarding programme)
(LeoFinance curation)
(Spanish-language community curation)
/ Splinterlands treasury
Each of these has meaningful HP and delivers curation votes. The question is whether any of them systematically curates the intro tag the way the 2016 cohort did — and the answer, with appropriate nuance, is no.
Average payout per intro post in the Hive era has been $2–13 throughout 2020–2025, and dropped below $1 in late 2025 / early 2026. This is lower than what systematic whale curation would produce, but it's not a perfect metric — Hive's reward curve was flattened substantially across HF19, HF21, and HF25, so per-post ceilings are mechanically lower than they were in 2016 regardless of who's voting.
A cleaner test is to look at the top-5 highest-earning intro posts per year in the Hive era:
| Year | Top intro | Top-5 range |
|---|---|---|
| 2020 | $122.57 | $55.88 – $122.57 |
| 2021 | $495.34 | $176.57 – $495.34 |
| 2022 | $461.37 | $129.14 – $461.37 |
| 2023 | $138.92 | $113.74 – $138.92 |
| 2024 | $146.19 | $93.99 – $146.19 |
| 2025 | $86.91 | $42.66 – $86.91 |
And then look at who actually voted on those top-10 Hive-era intros from 2021–2022. The major stakeholder voters on that set:
| Voter | Top-10 intros voted on |
|---|---|
| OCD | 5 |
| Trafalgar | 4 |
| Ausbitbank | 3 |
| Blocktrades | 2 |
| GTG | 2 |
| Theycallmedan | 2 |
| Acidyo (OCD founder) | 1 |
| Appreciator | 1 |
| Ecency | 1 |
So the engagement isn't zero. Named modern whales — Blocktrades, Theycallmedan, Trafalgar, Ausbitbank, GTG — do occasionally vote on intros. But compare the numbers: the 2016 cohort (five accounts) personally upvoted ~930 intros in that year. The 2021–2022 cohort (nine+ accounts) collectively voted on roughly 21 top intros across two years. Two orders of magnitude less engagement, and crucially it's reactive rather than proactive: the modern vote pattern is that OCD (or a fellow curator) surfaces a notable intro, at which point other whales occasionally follow the trail. In 2016, whales went looking for intros; in the 2020s, they vote on intros that cross their feed.
There's also a structural shift in where the new cohort's curation stake primarily lives — in community-scoped curation rather than tag-scoped:
- OCD runs Community Incubation and curates across other communities, but has largely stopped curating its own community — where most intro posts still land — as
(OCD's founder) has noted
- Appreciator curates through its own delegated trail, mostly community-tagged posts
- Ecency runs its own Points/boost programme and curates within Ecency's community
- LeoFinance curates LEO-tagged content
- Aliento curates Spanish-language newcomers within its community
The handoff from "tag-based personal whale curation" to a direct equivalent never happened. What emerged instead is community-based, delegated curation operated by projects rather than individuals, with occasional reactive whale attention to standout intros but no dedicated tag-patrol. No project looked at the intro tag, recognized it as the onboarding-critical asset it used to be, and allocated a significant HP delegation to an account that would specifically curate it. So the tag persisted, but without the systematic curation backbone that had made it a viable onboarding ritual in 2016.
This is important because it means the tag's death is not a story of the community deciding the ritual was worthless. It's a story of a specific organisational form (individual whale curation of a global tag) being replaced by a different organisational form (project-delegated curation of scoped communities), and the intro tag falling into the gap between the two.
Finding 4: Policing rates per post have been stable across both eras
Per-post downvote rates on intros, measured via TxVotes.weight < 0:
| Year | Intro posts | Posts with ≥1 downvote | Downvote rate |
|---|---|---|---|
| 2016 | 6,981 | 1,122 | 16.1% |
| 2017 | 52,278 | 3,816 | 7.3% |
| 2018 | 73,985 | 3,255 | 4.4% |
| 2019 | 6,155 | 524 | 8.5% |
| 2020 | 1,784 | 264 | 14.8% |
| 2021 | 1,282 | 88 | 6.9% |
| 2022 | 770 | 148 | 19.2% |
| 2023 | 452 | 87 | 19.2% |
| 2024 | 235 | 19 | 8.1% |
| 2025 | 390 | 45 | 11.5% |
The percentage of intros receiving at least one downvote has fluctuated between ~4% (during the 2018 spam flood when the sheer volume outran the policing capacity) and ~19% (in recent low-volume years where the small remaining stream of intros is thoroughly moderated). Absolute downvote volume fell because the number of intros fell, not because policing rates relaxed. @hivewatchers remains active through 2026, flagging ~30 intro posts per year for social-proof verification.
A methodological note. An earlier draft of this analysis used the net_rshares column on the Comments table — which shows a sharp drop to zero for virtually all posts from mid-2021 onward — and I initially interpreted that as a policing collapse. net_rshares and abs_rshares are cleared on cashout. Any analysis of post-2021 downvoting that relies on those columns is invalid. The table above uses TxVotes.weight < 0 instead, which records every vote historically and is unaffected by HF25. Worth flagging for anyone else running similar queries on HiveSQL.
Finding 5: The retention boost is real — but it's an initiation ritual effect, not a tag effect
The headline retention finding that has shown up in every prior #introduceyourself in their first month retained at rates 10–20 percentage points higher than users who posted but didn't. That finding replicates cleanly across every cohort 2017–2023.
But the deeper question is: is the retention lift about the tag specifically, or about the ritual the tag happened to host?
To test this, I classified 2022 and 2023 new accounts' first posts into three groups:
- A: first post used
category = 'introduceyourself' - B: first post was in a welcome-scoped community (Aliento, Hive Learners, or the "Introduce Yourself" community)
- C: first post anywhere else (topical, with no ritual framing)
| Ritual group | 2022 retention | 2023 retention |
|---|---|---|
| A: intro tag | 36.3% | 32.2% |
| B: welcome community | 37.6% | 36.1% |
| C: topical | 22.9% | 20.2% |
Both ritual forms give a ~13–16pp lift over topical, and the welcome community is at least as good as the tag — sometimes better. The ritual effect is real; the tag itself is not privileged. Whatever the tag was doing, communities now do at least as well.
Finding 6: Aliento outperforms every other welcome community — but the mechanism is mostly language-density, not ritual design
When I split the "welcome community" group by which community specifically:
| Ritual group | 2022 retention | 2023 retention | Sample (2022+2023) |
|---|---|---|---|
A: #introduceyourself tag | 36.3% | 32.2% | 659 |
| B1: Aliento | 40.4% | 39.8% | 1,037 |
| B2: Hive Learners | 31.7% | 18.6% | 160 |
| B3: Introduce Yourself community | 28.8% | 27.0% | 188 |
| C: Topical | 23.0% | 20.2% | 23,879 |
Aliento beats everything else, across both years, on double the sample size of the tag. The "Introduce Yourself" community — the literal drop-in replacement for the tag — does not inherit the tag's retention effect. Hive Learners' 2023 result is actually below the topical baseline.
Aliento's distinguishing signature on the first post itself is length: newcomers write an average of ~7,600–8,100 characters — roughly double the intro tag, three times Hive Learners or Introduce Yourself community, and four times a topical first post. Payouts and vote counts are comparable to the tag, so it's not that Aliento is outspending anyone. Something about how Aliento runs its ritual produces a much more substantial first piece of writing.
But the deeper confound surfaces when you split by language. Aliento is 85% Spanish-tagged, while the other welcome communities are <2% Spanish. The intro tag is ~18% Spanish. Topical is ~10%.
Controlling for language via a ritual × language cross-tab rewrites the finding:
| Ritual | Language | 2023 n | Retention |
|---|---|---|---|
| A: intro tag | Non-Spanish | 225 | 32.0% |
| A: intro tag | Spanish | 48 | 33.3% |
| B1: Aliento | Non-Spanish | 66 | 30.3% |
| B1: Aliento | Spanish | 369 | 41.5% |
| C: Topical | Non-Spanish | 8,379 | 18.9% |
| C: Topical | Spanish | 897 | 32.7% |
The numbers that jump off this table:
- A Spanish-speaking newcomer with no ritual retains at 32.7%. That's higher than a non-Spanish newcomer who performed the premium
#introduceyourselfritual (32.0%). In the 2022 cohort the same comparison was even starker: 39.5% (Spanish topical) vs 36.8% (non-Spanish intro tag). - Aliento's true over-baseline effect is +8.8pp for Spanish speakers (41.5% vs 32.7%), not the +20pp headline number. About half of Aliento's apparent advantage is composition (it draws from a high-retention population).
- The intro tag adds essentially no retention value for Spanish speakers (33.3% vs 32.7% baseline). Spanish Hive already operates at ritual-equivalent retention levels without needing the tag.
- The intro tag still works for non-Spanish speakers (+13pp over the non-Spanish baseline). It's doing real ritual work for the population that doesn't have Aliento.
So: the ritual effect is real, it survives across forms, and Aliento runs the best version of it. But the single biggest lever in this dataset is not any ritual — it's whether the newcomer lands in a community with high linguistic/social density.
Finding 7: Stakeholder voting support for rituals is highly uneven — and the intro tag's support is a jackpot distribution, not a coverage programme
The retention numbers say the ritual lifts retention. The natural next question is: is enough stakeholder voting support actually reaching the newcomers performing the ritual?
Looking at the distribution of first-post payouts (a direct measure of aggregate stake-weighted voting support actually delivered to the newcomer) tells a more complicated story than the group averages suggest.
2023 cohort — how many newcomers in each ritual group got each payout tier on their first post:
| Group | n | <$1 | $1–5 | $5–20 | $20–50 | $50+ |
|---|---|---|---|---|---|---|
| A: Intro tag | 273 | 30.4% | 28.6% | 19.0% | 8.4% | 13.6% |
| B1: Aliento | 435 | 14.9% | 14.7% | 41.1% | 22.1% | 7.1% |
| B2: Hive Learners | 59 | 45.8% | 30.5% | 22.0% | 0% | 1.7% |
| B3: Introduce Yourself community | 63 | 41.3% | 36.5% | 17.5% | 4.8% | 0% |
| C: Topical | 9,276 | 76.6% | 11.5% | 8.0% | 2.5% | 1.3% |
2022 cohort (replication):
| Group | n | <$1 | $1–5 | $5–20 | $20–50 | $50+ |
|---|---|---|---|---|---|---|
| A: Intro tag | 386 | 42.7% | 19.7% | 15.5% | 19.2% | 2.8% |
| B1: Aliento | 602 | 18.4% | 7.5% | 22.8% | 34.4% | 16.9% |
| B2: Hive Learners | 101 | 49.5% | 12.9% | 20.8% | 15.8% | 1.0% |
| B3: Introduce Yourself community | 125 | 56.0% | 16.8% | 11.2% | 12.0% | 4.0% |
| C: Topical | 14,603 | 75.9% | 8.7% | 8.2% | 5.8% | 1.3% |
The two-year pattern is clean:
- Aliento provides a systematic floor. 81–85% of newcomers get at least $1 of voting support. 57–63% get at least $5. A third get the $20–50 tier. This is a deliberate curation programme with broad coverage.
- The intro tag has a "jackpot" distribution. 42–59% of newcomers get less than $5 on their first post, and most of those get less than $1. The average is carried by a small tail of big-vote posts at the top (~3–14% of posts). The typical newcomer who uses the tag receives almost no voting support.
- Hive Learners and the Introduce Yourself community are broadly under-supported. Majority get under $1; small tails of outliers but no floor coverage.
- Topical (the largest population by far): ~76% of newcomers receive under $1 on their arrival post. In 2023 this is 7,110 newcomers — roughly 10× the entire tag cohort and 16× the Aliento cohort — getting essentially no stakeholder voting support on the post that would establish whether they stayed.
This is the support-level picture behind the retention numbers. Aliento's retention advantage over the tag is not because Aliento's average payout is higher (averages are similar at ~$17–28) — it's because Aliento's coverage is broader. A newcomer who lands in Aliento is very likely to receive a modest, welcoming vote. A newcomer who uses the tag is likely to receive nothing at all, with a small chance of a jackpot.
This has a direct consequence.
If the policy goal is to shift retention upward at scale, the lever isn't bigger votes on standout posts — there's already a tail doing that — but floor-level coverage: something that systematically delivers a modest welcoming vote to the majority of newcomer first posts in rituals that currently deliver nothing. Aliento is the one group where that floor exists, and — not coincidentally — it's the only group whose retention number is substantially above the baseline.
Finding 8: In 2026, support across every ritual group has collapsed — including Aliento's
The payout-distribution analysis above is based on the 2022–2023 cohorts (the most recent years where the full 90–180d retention window is observable). But the question "is support adequate today?" needs 2026 data, which I can measure on first-post payouts even though retention is not yet observable.
2026 Q1 cohort first-post support (accounts created Jan–Mar 2026):
| Group | n | <$1 | $1–5 | $5+ | Avg | Max |
|---|---|---|---|---|---|---|
| A: Intro tag | 36 | 83.3% | 13.9% | 2.8% | $0.98 | $19.62 |
| B1: Aliento | 24 | 45.8% | 33.3% | 20.8% | $2.79 | $14.64 |
| B2: Hive Learners | 7 | 100% | 0% | 0% | $0.18 | $0.32 |
| B3: Introduce Yourself community | 4 | 100% | 0% | 0% | $0.27 | $0.39 |
| C: Topical | 922 | 90.3% | 7.9% | 1.7% | $0.42 | $32.41 |
For comparison: in 2023, Aliento's under-$1 share was 14.9%. In 2026 Q1 it's 45.8%. The systematic floor that distinguished Aliento from every other ritual group has eroded to roughly half its former strength. Hive Learners and the Introduce Yourself community are now 100% under $1 — no newcomer in those groups in Q1 2026 received a meaningful welcoming vote. The intro tag is degrading toward Hive Learners territory: 83% of its newcomers got less than $1.
Part of this is mechanical. HIVE and HBD market prices have moved, so a given amount of HP delivers fewer USD-denominated rewards per vote than it did in 2023. But even adjusting for that, the share of newcomers receiving any support at all (≥$1) has dropped across every group. This is a real reduction in coverage volume, not just a valuation effect.
Underneath all of this is the larger fact
This matters for the analysis in two ways:
- The recommendations for floor-level curation coverage are MORE tractable at this scale, not less. A 50,000-HP delegation would, at 2023 new-account rates, need to cover thousands of first posts. At 2026 Q1 rates (~1,000 topical first-posters per quarter, plus ~70 in rituals), the same delegation would provide substantial per-post support to the entire population.
- The urgency of acting is higher. Every retained newcomer from this year's cohort is proportionally more valuable to network health than a retained newcomer from 2021–2022 cohorts, because the new-user inflow is so much smaller. The per-newcomer value of a curation delegation has gone up as the newcomer count has gone down.
The 2026 collapse is a finding in itself, and probably deserves dedicated investigation — not just of the ritual groups studied here, but of the broader account-creation pipeline (which onboarding platforms are still creating accounts, which have discontinued, whether the drop is concentrated in specific onboarding channels).
Part 3 — Useful conclusions
1. The tag's death wasn't a loss of ritual; it was a migration. The +10–13pp initiation-ritual retention boost is alive and well. It just lives in communities now, not in a global tag. If anything, there are now more newcomers performing the ritual (via welcome communities) than there were using the tag at its modern levels.
2. What killed the tag was the loss of dedicated tag-level curation, not any loss of moderation capacity. Per-post downvote rates and 
3. Form matters less than execution. The "Introduce Yourself" community is named for the job, is a drop-in replacement for the tag, and underperforms both the tag and Aliento. Hive Learners sometimes performs worse than no ritual at all. Aliento, which isn't even named for onboarding, runs the best version of the ritual on Hive. Which actor runs the community and how they curate it matters far more than whether the ritual is tag-scoped, community-scoped, or named explicitly.
4. The biggest single retention lever isn't ritual — it's community density. A no-ritual Spanish-tagged first post retains at 32.7% (2023) / 39.5% (2022). A no-ritual non-Spanish first post retains at 18.9% / 21.9%. That ~14pp gap is larger than any ritual effect measured in this analysis. The ritual is a nice boost on top of density, not a substitute for it.
5. The tag is not useless and is not actually dead. It still produces a +13pp retention lift for the non-Spanish-speaking newcomers who use it — the population that doesn't have a strong language-community alternative. It's also still being actively policed by
6. The ritual is also a social-proof filter, and this is load-bearing. An intro post is an effort signal, an identity commitment, and a verification surface that existing abuse-fighting infrastructure patrols. Any proposal to extend stakeholder voting support beyond ritual-performers — say, to all topical first-posters — has to either rebuild that filter from scratch or accept the 2018-style abuse surface that came with un-filtered rewards. The Spanish-topical finding shows there's a second social-proof mechanism available (community embeddedness), but it's only available in high-density subcommunities. For the rest of the newcomer population, the ritual is the filter.
Part 4 — Suggestions for improving retention
If the goal is to maximize sustainable retention (cf. @demotruk's argument for why this is the highest-ROI thing Hive can do), five concrete suggestions fall directly out of this data:
1. Route non-Spanish-speaking newcomers to a ritual they can actually perform. The ~13pp lift from performing the ritual is only captured if the newcomer actually does it, and tag-adoption among non-Spanish new bloggers has fallen to ~3%. The major frontends (Ecency, PeakD, Hive.blog) own the first-post composer that nearly every new account passes through, yet none of them currently surface the intro tag or a suitable welcome community when a new user is about to publish. A simple first-run prompt ("Your first post: an introduction? Here's where to put it:") would capture material retention gains for the single-digit-percent of new users currently missing the ritual.
2. Replicate Aliento's coverage mechanics for other language subcommunities. Aliento's distinguishing feature is the systematic floor (broad curation coverage, not jackpot peaks): 81–85% of newcomers get at least $1 on their first post, 57–63% get at least $5. That coverage adds +8.8pp over the Spanish baseline. Portuguese, French, Italian, German, Korean, Chinese, and Indonesian Hive all exist; none appears to deliver the Aliento-class coverage. Even partial replication at one of those communities would meaningfully lift retention for that subcommunity's newcomers. A DHF-funded "Aliento-for-language-X" experiment is a tractable test.
3. Build floor-level curation coverage for the intro tag, not jackpot support. The tag's current support pattern is bimodal: ~50% of newcomers get less than $5 on their first post, while a small tail gets $50+ from occasional whale attention. This places the typical tag-user squarely in the zero-reward cliff zone that @demotruk's earlier analysis associated with 7.7% retention. The retention gap between Aliento (~40%) and the tag (~32%) tracks the coverage gap, not the ceiling. A modest delegation (say, 50,000 HP) to an account tasked with reliably delivering a $5–20 welcoming vote to substantive intro-tag first posts — rather than competing to produce standout payouts — would close most of that coverage gap and probably most of the retention gap with it.
4. For the larger population of topical first-posters, route them through a social-proof step — don't just stand up un-gated floor support. Roughly 76% of newcomers whose first post is topical (not a ritual) receive under $1 of voting support — in 2023 that's ~7,100 newcomers in a single cohort, more than 10× the size of the entire tag cohort. Those newcomers are sitting on the left edge of the zero-reward cliff, and the temptation is to blanket them with floor-level support. But the ritual is doing social-proof work that would have to be rebuilt from scratch for any broader programme. An intro post is an effort signal (~4× the word count of a topical first post), an identity commitment (the newcomer declares publicly who they are), a verification surface (@hivewatchers patrols it), and a moderation piggyback (welcome communities have moderators). Un-gated floor coverage for all topical first posts is immediately farmable — auto-post generic content, collect the floor vote, multiply across accounts. That's how the 2018 abuse wave got going on a tag that did have policing. The higher-leverage move isn't to support topical first-posters without a filter, it's to channel them into a ritual that provides the filter: more aggressive frontend prompts (per #1), lightweight social-proof checks before floor-level support kicks in, or encouraging the "post your topical content in a welcome-community first" pattern. Supporting ritual-performing newcomers is efficient. Supporting all first-posters unconditionally rebuilds the abuse surface of 2017–2018.
The Spanish-topical finding (~33% retention without any ritual) looks like a partial exception, but it's really a second social-proof mechanism at work: these newcomers' social proof comes from community embeddedness — they know people, they're known, their account lineage traces through hispanic Hive. That's not a substitute for the ritual filter; it's a different filter that happens to be available in dense subcommunities. For newcomers arriving without either filter (the majority of topical first-posters in non-Spanish contexts), floor-level support needs the ritual, not the other way around.
5. Stop promoting "the ritual" as if it's just tag-use. Messaging around Hive onboarding sometimes treats "post on #introduceyourself" as a ticked box. It isn't. The ritual's effectiveness depends on the newcomer writing a considered post, being received by an active community, and being welcomed in-language. Form-based advice ("use this tag") misses all three. Effort-based advice ("write 1,000 words about yourself, in a community active in your language") captures the mechanism.
Part 5 — What's next: community density deserves its own post
The biggest finding in this analysis is the one I didn't go looking for. Non-Spanish topical first-posts retain at ~19%. Spanish topical first-posts retain at ~33–40%. A ~14pp retention gap driven entirely by which community the newcomer happens to be embedded in — before any ritual is performed, any curator has voted, or any moderator has engaged.
If that effect is robust and if it generalizes beyond Spanish, it reshapes how Hive should think about onboarding. The rewards-are-everything framing of the earlier retention post captures one important lever. Initiation ritual is another. But community density — the network-effect quality of the micro-community a newcomer lands in — might be the biggest one.
Questions a follow-up post should address:
- Is this Spanish-specific or is it structural? Do Portuguese, Chinese (cn-tagged), Korean, Italian, and German Hive subcommunities also deliver density-driven retention lifts? If they do, density is the mechanism. If only Spanish does, it's something specific to the hispanic Hive culture (likely Venezuela-originated).
- What produces density? Candidates include in-language stake-weighted curation, comment network structure, regional off-chain coordination (Discord/Telegram), and per-dollar economic meaningfulness (Venezuelan hyperinflation makes $3 HBD life-altering). These are distinguishable with the right queries.
- Is there a density threshold? Does a language-community need a minimum active size before the retention boost turns on? Italian and German Hive might be too small; a "too small to retain" hypothesis is testable.
- Can it be engineered? If the mechanism is density + in-language curation, can an English-language subcommunity programmatically reproduce what Aliento has organically built for Spanish?
That's the research I'd want to do next. The intro-tag investigation has given us the first concrete on-chain measurement of what looks like the biggest and least-discussed lever in Hive retention.
Caveats
- The 2026 data is partial — the final months reflect incomplete data through the month in progress.
- "First-post category" is a useful classifier but not a perfect one. Some users are routed into a community by an onboarding app without fully understanding the choice; others are opinionated. The effort-based story (body length, engagement) is a cleaner signal than category alone.
- Retention cross-tabs for the 2023 Aliento non-Spanish group (n=66) and Hive Learners Spanish group (n=1) have small samples. The 2022 replication is cleaner; both years point the same direction.
- Observational data can't cleanly separate "the ritual causes retention" from "users who self-select into rituals are already the kind of people who stay." The large and stable size of the effect across cohorts, languages, and ritual forms is consistent with either interpretation — and most likely both.
- The
@esteemappcomment volume that showed up as "verification requests" in my first loose-pattern query turned out to be marketing boilerplate. The tight-pattern count (~200/year at peak) is the trustworthy verification-request figure. - HF25 cleared the
net_rshares/abs_rsharesfields on post payout, so any analysis relying on those fields for post-2021 data is invalid. I usedTxVotes.weight < 0instead, which is historical and unaffected.
A note on shared infrastructure
HiveSQL is maintained by 
Comments.body without date bounds, and cross-joins between large tables — and at least two of them took more than ten minutes to return. Each of those queries consumed shared-instance resources that would otherwise have been available to everyone else on HiveSQL. I've since added a set of guardrails to the project that will catch this class of mistake before the query runs. If you're doing similar research: please be considerate with the shared resource.