Foreword — by @demotruk
Disclosure: I'm a regular HiveFest attendee — I've been to several editions since 2016 so I have a personal stake in this question. In the interest of transparency: I usually paid ticket prices that at least covered the per-person cost of hosting me, and often more. This is verifiable on-chain. Further, if a DHF vote comes up for funding HiveFest, I am likely to support.
Research Findings — by Claude Opus 4.8
This analysis was conducted at @demotruk's direction. I designed the methodology, pulled and analysed the data, and drafted the findings; @demotruk posed the question, directed the investigation, and challenged the results. Everything below is reproducible from public sources — HiveSQL, Google Trends (via pytrends), GitHub commit history, and the organiser's own announcement posts. HiveSQL data was gathered over the first week of June 2026; Google Trends data was retrieved 2026-06-07. All queries, scripts, and datasets are available on request, and every query was bounded by date and designed to be a light load on the shared HiveSQL instance.
The question, in their words
@roelandp has put an honest question to the community: should the DHF fund HiveFest XI — Barcelona, 17–20 September 2026 — and, alongside it, a new outward-facing day called ChainCulture — for roughly 45,000 HBD? It would be the eleventh edition of a festival that has run, almost without a gap, every year since 2016.
His go/no-go post does not oversell. He states the cost-benefit plainly: "I will not pretend this prints money for Hive. It does not, certainly not directly or measurably." What it does do, he argues, is point the community outward instead of inward — turning a decade of "us meeting us" toward thousands of people who have never touched Hive, as a funnel for developers, creators and makers, and as morale at a low moment.
The question lands in the middle of a treasury squeeze. With HIVE around $0.05 and the DHF under visible pressure, every HBD to be spent is scrutinised, and "does it earn its keep?" is fair to ask of any line item — conferences included. @roelandp is right not to pretend otherwise. So let's test it as rigorously as the data allows: treat the festival as a decade-long natural experiment, go looking for its return in every channel that can be measured — and if the growth he declines to promise never comes, ask what exactly the DHF is paying for.
First, what it actually costs (historically)
ROI has two halves. Start with the cost. At $0.05 HIVE and falling, the treasury is under real pressure and every HBD spent has an opportunity cost. The cost side deserves an honest number. For a decade, HiveFest has cost roughly $400–730 per person to host, and that figure has been remarkably stable across nine in-person editions on three continents.
Two things stand out. First, attendees never pay full freight at the bottom of the ladder. The cheapest tier has consistently paid roughly a third to a half of the true per-person cost — €111 against a ~€383 cost at SteemFest 1, €222 against ~€667 at SteemFest 2, $485 against ~$700 at HiveFest 8. Second, the pricing was never a flat fee: it is a stake-keyed, progressive ladder (Plankton → Minnow → Dolphin → Orca → Whale), where the lowest tiers pay a subsidised fraction and the regular and higher tiers pay 100% of cost "or even higher levels," in @roelandp's own words. The 2020 and 2021 editions were virtual and free — the cheap pandemic anomaly — and the two most recent editions only publish the ticket, not the full host-cost, so those bars are estimated.
Who covers the gap? Factually, and without spin: pre-fork it was Steemit Inc. plus a handful of whales; post-fork it is the DHF plus project sponsors. The DHF's direct contribution is small and easy to count — exactly two dedicated hivefest proposals ever drew treasury money, both for the virtual editions: 18,985.62 HBD in total (HF5 in 2020 and HF6 in 2021). @roelandp himself has received nothing through a personal proposal.
The in-person editions are subsidised too, but indirectly, through the @valueplan umbrella proposal, which @roelandp's ticket announcements credit for reducing prices ("reduced by 50%," "$700 → $485," "via the Hive DAO / ValuePlan"). ValuePlan is a large recipient that funds many things, and the HiveFest slice is not separately itemised on-chain, so it can only be bounded — roughly $30–85k of DHF money per in-person edition. Real money, ongoing, and rightly scrutinised in a downturn.
But look at who actually covers the bill. Attendees have consistently covered between a third and half of the true per-person cost at the cheapest tier — and the subsidy share has shrunk over time, not grown: from ~70% subsidy at SteemFest 1 down to ~30–45% at the recent in-person HiveFests where data is available.
In absolute terms, the total event cost has ranged from ~$42k (SteemFest 4's lean Bangkok edition) to ~$217k (SteemFest 2 at the peak of the bull run), with attendees themselves funding roughly half through that progressive ticket ladder. The virtual editions cost the DHF less than $10k each. For the recent in-person HiveFests (8–10), headcounts weren't published and ticketing moved off-chain, so totals can't be reconstructed — but the one post-fork in-person edition with full data (HiveFest 7, Amsterdam 2022) came in at ~$73k, with attendees covering roughly 55–100% of the per-head cost.
So the festival is modestly priced and largely self-funded — but "modest" is not "free," and "self-funded" is not "costless to the DHF." The right question isn't "can we afford it?" It's "does it earn its keep?" And we can test it unusually cleanly — because the chain's data is public and queryable, and because the festival is, in every respect that matters, the same festival throughout.
Same organiser (@roelandp). Same format. Same annual cadence. And a single continuous numbering that runs straight through the 2020 fork: SteemFest 1–4 → HiveFest 5–11. The event simply became HiveFest when the chain became Hive. That sameness is what makes the comparison fair: whatever the festival returns, we can measure it before the fork and after, holding the event itself essentially constant.
The full series:
| # | Event | Year | City |
|---|---|---|---|
| 1 | SteemFest 1 | 2016 | Amsterdam |
| 2 | SteemFest 2 | 2017 | Lisbon |
| 3 | SteemFest 3 | 2018 | Kraków |
| 4 | SteemFest 4 | 2019 | Bangkok |
| 5 | HiveFest 5 | 2020 | virtual |
| 6 | HiveFest 6 | 2021 | virtual |
| 7 | HiveFest 7 | 2022 | Amsterdam |
| 8 | HiveFest 8 | 2023 | Rosarito |
| 9 | HiveFest 9 | 2024 | Split |
| 10 | HiveFest 10 | 2025 | Kuala Lumpur |
| 11 | HiveFest 11 | 2026 | Barcelona |
A note on method. All on-chain measures are Hive-native. I split every series at the fork date, 2020-03-20, and the festival cohort is the 1,347 distinct accounts that ever published a post tagged steemfest* or hivefest*. That cohort over-samples public posters and misses silent attendees — a caveat I'll return to — but it is a consistent, reproducible proxy across all ten past events.
I went looking for growth. It isn't there — and it never was.
If HiveFest generates a return, the obvious place to find it is growth: new users, buy-side demand, development that compounds into adoption. I tested each, in both eras, and tried hard to harden every apparent signal until it either survived or dissolved. They dissolved.
Account creation — null in both eras. An event study (detrending each event window against a 28-day local baseline, which removes both the free-signup level of the Steem era and the 2017–18 bull-run ramp) finds no signup bump above local noise at any festival, on either chain. SteemFest z-scores ran −0.12 to +0.25 (median +0.02); HiveFest ran −1.04 to +0.67 (median −0.32). Conferences never acquired users — not even on free-to-join Steem.
Price / buy-demand — null, with one tempting exception. Token returns around events (STEEM/BTC before the fork, HIVE/BTC after) point in opposite directions by era — Steem −12% over the following 30 days, Hive +7% — and both are mostly noise around the market cycle. The one move large enough to tempt you is HiveFest 6: HIVE/BTC was down 18% going into the event, then ran +94% in the month after. And here's the honest part — that's a HIVE/BTC figure, HIVE beating a Bitcoin that was topping out, so "it was just the market" doesn't explain it. Something HIVE-specific happened. The obvious suspect is also right there on the bill: Splinterlands was a headline act of HiveFest 6, and late 2021 was the peak of the play-to-earn gaming mania that Splinterlands led. So did the festival move the price? Almost certainly not. The SPS boom was a year-long, ecosystem-wide wave — HiveFest put Splinterlands on stage because it was already the chain's juggernaut, not the other way round. One idiosyncratic spike, landing in the single biggest dApp boom Hive has ever had, can't carry a "festivals drive price" claim when the other three pre-fork events and every later HiveFest show nothing at all.
Development → growth — looked like a clean break, then died under hardening. This was the most promising lead, and the one I most wanted to be real. Controlling for BTC and ETH returns, monthly development activity correlated with active-author growth at r = 0.60 before the fork but only 0.12 after — looking like a clean before/after split, and the single most tempting result in the study. But splitting the pre-fork window in two killed it: the coupling is entirely the 2016–2018 adoption ramp (r = 0.77), and it had already collapsed to 0.04 during Steem's own 2018–2020 bear market — before the fork, on the same chain. Like-for-like, post-peak Steem and post-fork Hive decouple identically. It was a bull-ramp effect, not a fork effect. (What survives is real but chain-agnostic: a decade of sustained engineering — 15,000 to 29,000 commits a year, every year, across ~570 repositories. Effort never collapsed. It just stopped converting to growth once the hype phase ended — on both chains.)
Power-ups — null both eras, even at the cohort level. Powering up (transfer_to_vesting) is the cleanest on-chain signal of commitment. Festivals don't drive a chain-wide power-up wave; and when I split the series by festival-cohort membership and ran a difference-in-differences against the rest of the chain, the festival cohort powered up only +4.6% more than baseline around events (z = 0.15), and even that is carried almost entirely by a single event (SteemFest 3). No commitment spike.
Four channels, two eras, the same verdict each time: nothing. And notice where the nulls are. Every one of them is an external / growth channel — the things "ROI" is usually meant to capture. The festival did not acquire users, did not move the token, did not convert development into adoption, did not even spike on-chain commitment among the people who attended.
Here is the reframe that took me four dead channels to arrive at: a conference was never a user-acquisition channel — not even on a free-signup, booming Steem. Judging HiveFest by signups or growth-ROI was the wrong frame from the start. Steem's growth ran through price → search → signup; the festival's product was never signups.
Which raises the real question. If the return isn't external growth, then the festival is producing something — people keep organising it, keep paying to attend, keep coming back. What is it?
So what DOES it produce? The core that holds the chain together.
Turn the question around — stop asking what the festival sends outward and ask what it concentrates inward — and the nulls invert into one of the strongest signals in the whole dataset. The festival is an internal engine, and its output is measurable and large.
Fork loyalty — they carried Hive across its existential moment. At the March 2020 fork, every Steem account was airdropped a matching Hive account and forced to choose: stay on Justin Sun's Steem, or move to community Hive. Because HiveSQL only sees Hive, every post-fork post is, by definition, a visible record of someone who chose Hive. Among everyone who had posted in the 12 months before the fork — people who demonstrably faced the choice — pre-fork SteemFest authors chose Hive at 2.2× (by 2021) to 2.7× (by 2025) the rate of engagement-matched non-festival authors. And the edge holds within every activity tier, which rules out the obvious "festival-goers just post more" objection: among one-post casuals the lift is largest, and even among the heaviest 101+-post power users it is still 1.7× to 2.3×.
Leadership pipeline — roughly 21–24× over-represented. Festival-goers are about 1.1% of active authors. But they are 23% of active witnesses and 26% of DHF proposal creators — a 21–24× concentration. This isn't a long tail of marginal names; the witness overlap is the top-20 core that literally runs the chain: gtg, pharesim, timcliff, yabapmatt, stoodkev, howo, emrebeyler, and more.
Onboarders — about 4.5× more likely to bring people in. 36% of the festival cohort has created at least one account for someone else, versus 8% of the active baseline. And they don't just dabble: festival-goers run Hive's largest onboarding operations — oracle-d, Ecency / good-karma, steemhunt, theycallmedan, hivewallet. The cohort created over 31,000 accounts, 72% of them through six bulk services that are themselves operated by festival attendees.
Retention — high in both eras. This was established in the in-person retention post: festival cohorts retain at high rates on both chains.
And there is a quiet fifth finding hiding in plain sight in §1: the funding model is itself communal. That progressive ticket ladder, where the cheapest tier pays a third to a half of cost and the higher tiers pay 100% or more, means wealthier attendees literally subsidise the rest — reinforced by explicit "support a stranger" mechanisms in both eras (SF1's support tickets; HF10's donate-extra-at-checkout). The attendees fund a large share of their own gathering, and they fund each other. The festival's own economics are a microcosm of the community it convenes.
One honest caveat, stated once. Every one of these is vulnerable to selection-versus-treatment: maybe leaders, loyalists, and onboarders attend festivals because they are already the committed core, rather than the festival forging them. The baseline already conditions on being an active poster, so the concentrations are over and above general engagement — but causation isn't proven, and probably runs both ways. Festivals likely select for the core and reinforce it. Which is exactly why cutting one matters: either way, you are removing the ritual that binds the people who hold the chain together.
And one loose end. There is a puzzle in the onboarder data I want to flag now and come back to. Festival-goers are prolific onboarders in both eras — but the timing differs. After a SteemFest, the cohort's account creation visibly bumped — even SteemFest 3, held in the depths of the 2018 bear market, when chain-wide account creation was collapsing. After a HiveFest, the same people stayed just as prolific onboarders overall, but that post-event bump never appeared. The willingness never went away. After the fork, the wave it used to produce simply stopped coming. Hold that thought.
The turn: everything glided across the fork except one thing
So I went back to that loose end, because it is the only asymmetry in the whole study that the nulls couldn't explain. After a SteemFest, the cohort created a visible wave of new accounts. After a HiveFest, the same people stayed just as prolific onboarders — but that post-event wave never came. What single thing changed across the fork that would let the willingness survive while the wave disappeared?
First, rule out the obvious. Did the fork break the festival's internal function? No — and emphatically not. Retention is high in both eras. Loyalty, leadership, and onboarding propensity are all healthy on Hive. Even the festival's reach — distinct festival authors as a share of the chain's active authors — declined as a smooth glide (14% → 8% → 5% → 4.8% across SteemFests, then 3.9% → 1.8% → 3.9% across early HiveFests) with no discontinuity at the fork at all. SteemFest 4 (4.8%) flows straight into HiveFest 5 (3.9%) and HiveFest 7 (3.9%). The chain shrank and the festival's relative footprint drifted down with it — gradually, with the market cycle, not at the fork event.
So the break isn't inside the festival. It's in what the festival's recruiters were recruiting into. An onboarding wave needs two things at once: people willing to bring others in, and others who can actually be brought. The festival reliably supplies the first — in both eras, its attendees are among the chain's most committed recruiters. The second isn't the festival's to supply: it depends on whether someone an attendee enthuses about the chain can go home, look it up, find it, and arrive. That second half ran through one thing — the chain's name and its front door — and that is the one thing that broke cleanly, on the date of the fork. The variable this post has been circling finally has to be named, and it is the name: at the fork the chain's identity went from a findable word to an unfindable one, and the stream of newcomers an attendee's enthusiasm could turn into accounts dried up with it.
The chain had a working discovery flywheel — and it ran on a searchable name. Pre-fork, "Steemit" was a destination people typed straight into a search bar: it was searched roughly 20× more than "PeakD." That is the loop the recruiters plugged into — a rising price drew searches for "Steem" and "Steemit," those searches hit a front door people could find, and some fraction stayed. Enthuse about the chain in 2017 and your friend could go home, search it, and land on a destination they could locate, navigate, and stick to — an account waiting to be made. Post-fork, the loop is severed at the mouth: Hive's flagship front-end is search-invisible, the name returns everything-but-the-chain, and there is no findable door. The attendees never stopped being recruiters — cross-sectionally they are still Hive's most prolific onboarders. But a recruiter can only create an account for someone who shows up, and after a HiveFest fewer people showed up: an attendee can enthuse all they like, but a friend who hears "join Hive," searches for it, and lands nowhere never becomes an account to create. That is the resolution of the loose end — the willingness survived the fork; the findable chain it used to feed did not.
And here is the tell that the severed joint is specifically the name — not branding, not the front-end, the word itself. The festival's own name carries the identical fracture. This matters precisely because it shouldn't: there is no "SteemFest flywheel." Nobody ever grew the chain by searching the festival's name; the festival's discoverability was never a load-bearing part of any funnel, so it had nothing to gain or lose by being findable. Which is exactly what makes it clean evidence — when the same naming pathology turns up here too, in a place where it earns nothing, it can't be a growth mechanism doing its job. It's the disease leaving its fingerprint where it has no reason to be.
"SteemFest" was a unique, clean, searchable term: zero before 2016, then spikes precisely at events 1 through 3, scaling with the market to a normalised 100 at the November 2017 bull-run peak. "HiveFest" sits at the noise floor — non-zero in only three months across 22 years of data, and two of those three aren't even the event (one predates HiveFest entirely; one falls in a month with no HiveFest); it collides with a music "HIVE Festival," and the volume is so low that Google returns no related queries at all. The honest nuance sharpens it rather than softening it: SteemFest 4 in 2019 also scored zero, so the pattern is not "Steem always buzzed." It's that SteemFest responded to a discovery engine — hot market → 100, bear market → 0 — while HiveFest never clears the floor in any condition at all: bull or bear, virtual or in-person. One word was wired to a funnel; the other is disconnected from every funnel. Same word-level break, same date, as the chain that named it.
Let me show you, rather than tell you, how real this is. While researching this very post, I could not look up HiveFest ticket prices. Searching "HiveFest tickets" returned a German techno festival and a music event — not the blockchain conference. I had to pull every historical ticket price directly off the blockchain, out of @roelandp's own announcement posts, because the open web doesn't reliably know the event exists. The name problem obstructed even this analysis of the name problem.
So the loose end resolves — and it resolves upward, to the chain. The festival's people stayed willing recruiters in both eras; what vanished after the fork is the post-event wave of new accounts — not because anything in the festival broke, but because the chain-level funnel it fed — searchable name, findable front door — was severed at the fork. The festival's own unfindable name doesn't cause that loss. It diagnoses it: a smaller, self-contained fingerprint of the exact same naming wound. Which is the thread the next section pulls — because once you see that the break is the name, you stop seeing a festival problem at all.
It was never the festival. It's the whole chain.
Here is the part that should change how the DHF debate is framed: the severed pipeline is not HiveFest's failing. It's Hive's.
The festival is a microcosm of the entire chain. It produces real, large, measurable internal value — loyalty, leadership, onboarding, cohesion — and converts almost none of it into external growth, because the discovery layer that would carry that value outward is broken. That is not a fact about conferences. It is the same wall that the broader discoverability research documented at the level of the whole chain: the price-to-search correlation that ran at r = 0.90 on Steem collapsed to r = 0.21 on Hive. The same wall that defeats every DHF dollar spent on growth. The same wall that let Splinterlands bring in hundreds of thousands of accounts with essentially zero spillover to the token flywheel.
The fear that DHF spending earns no measurable return is correct about the symptom and wrong about the cause. Internal value on Hive is real and abundant. External, growth-shaped return is gated — and the gate is discoverability, and the name keeps the gate shut. You cannot spend your way past it at the level of any single line item, because the thing that's broken is shared by all of them.
The name "Hive" keeps showing itself as a problem even when we're not looking for it. HiveFest is a decade-long natural experiment — same event throughout — and we came to it asking about ROI, not branding. Yet every external channel is null in both eras, and the only clean fork-break is the one the name predicts.
Back to Barcelona
So, back to @roelandp's go/no-go on HiveFest XI — and on ChainCulture, the outward-facing day he proposes alongside it. Whether 45,000 HBD clears the bar in a squeezed treasury is something stakeholders can decide. What this analysis can offer is the other half of the cost-benefit: if the DHF chooses to fund Barcelona, here is the return it is actually buying — and, just as importantly, the return it is not.
What the money does not buy: growth. It won't deliver one, and @roelandp is right to say so plainly ("I will not pretend this prints money for Hive ... certainly not directly or measurably"). The honest finding of this whole exercise is that no conference, on any chain, in any market, ever did. Anyone funding HiveFest expecting signups, buy-side demand, or a development-led adoption bump should know the decade-long record says it isn't there.
What the money does buy: the internal core it forges and re-forges every year. This is the ROI, and it is unusually legible for a line item this size. The loyalty that carried Hive across the fork (2.2–2.7×). The leadership layer it concentrates (21–24×). The onboarders it gathers (4.5×). The cohesion that doesn't show up in a token chart but is the reason there is a chain to chart at all. On a modest, bounded, largely self-funded budget, that is a return that is actually visible in the data — which is more than can be said for most growth-shaped spending. Whether it's worth the price is a judgement; whether there's a return at all is not.
But understand the limit. No event-level budget can fix discoverability. ChainCulture's instinct is exactly right — point the community outward, toward the thousands gathered next door rather than inward at the reunion. But it aims a funnel at the one joint this whole study shows is severed: a chain whose name can't be found and whose front door can't be searched will leak most of those fresh eyes straight back out. That is a chain-level problem — a name problem — and judging HiveFest or ChainCulture by the growth the name makes structurally impossible would be a category error: it risks defunding the one thing that reliably works while leaving the actual disease untouched.
Because that is what this is. The DHF's anxiety about whether its spending returns anything and the HiveFest go/no-go are not two problems. They are two symptoms of one disease, wearing two masks. Internal value is real; external conversion is gated by discoverability; and the gate stays shut as long as the name does.
HiveFest doesn't have an ROI problem. Hive has a discoverability problem. The return on Barcelona — whatever the community decides it's worth — is real, measurable, and inward-facing. The gate is still what needs fixing next.
Appendix — method & data
- Festival cohort: 1,347 distinct accounts that ever published a depth-0 post tagged
steemfest*orhivefest*(data/lead_festival_cohort.csv). A proxy for attendance that over-samples public posters and misses silent attendees. - Baseline community: 25,574 distinct authors active in 2025 (
data/lead_active_2025.csv). Over-representation figures are relative to the festival-goer share of this already-engaged baseline (1.09%), so they are over and above general engagement. - Fork date: 2020-03-20, used as the era split everywhere.
- Event dates verified from official posts (
data/events.csv). - Market controls: BTC, ETH (backfilled to 2016), STEEM, and HIVE daily prices; native token is STEEM pre-fork and HIVE post-fork.
- HiveSQL notes: power-ups via
TxTransferstype='transfer_to_vesting'(full history; the verified-output table only starts mid-2021); witnesses viaTxWitnessSetProperties.owner(noWitnessestable exists, so witnesses who only ranwitness_updateare missed); proposals viaProposals.creator. Real stake isvesting_shares. - Development data: 589 curated Hive-ecosystem repositories; 201,274 commits across 571 of them, via the GitHub stats/contributors API.
- Caveats: pytrends returns relative (0–100) interest and can vary slightly between pulls — directional findings are robust, individual points less so; both festival terms are low-volume, which is itself part of the finding. Selection-versus-treatment is unresolved for all internal-value findings (cohort may be selected by the festival, reinforced by it, or both). The onboarder event-timing result (§3/§4) is suggestive, not a hardened causal break — the cross-sectional 4.5× is the robust onboarder finding.
- All queries, scripts, and CSVs available on request.
*Data: HiveSQL (queried June 2026, all queries date-bounded and designed for minimal load on the shared instance); Google Trends via pytrends (retrieved 2026-06-07); GitHub commit history; and @roelandp's own SteemFest/HiveFest announcement posts.
