Sold out doesn't mean funded.
Private disclosures from 20+ founders put refunds at 60–95% of announced capital in the launches they described. We analysed 16 launchpad policies and 21 recent sales to examine what happens between “sold out” and settled capital.
The Refund Gap is the difference between the raise a token sale announces and the capital the project keeps once the refund window closes.
The Refund Gap 60–95% refunded · range disclosed by 20+ founders
Sold out — 100% announced
5–40% settles
The announced 100% stays in the public record. Nobody restates it.
Disclosure: AlphaMind operates a non-refundable, committed-capital launchpad. That is why we collected this data — and why every external number below carries a source you can check.
Three findings that explain the Refund Gap.
of the announced raise came back as refunds — the range founders disclosed to us across 20+ refundable launches, 2024–2026, clustering around 80%. Disclosed, not audited: no launchpad publishes an average.
11 of the 16 sales with verifiable price data traded below their sale price while the refund window was still open — and 15 of 16 ended below sale price in the window or after it. Public trackers, 2024–2026.
of documented refunds or cancellations were still displayed as completed raises on major third-party aggregators when we checked, 24 July 2026 — including MultichainZ: $1.44M announced, fully refunded, still shown as raised.
Basis: 16 launchpad policies read in full · 21 sales traced, 16 with verifiable prices · 6 documented cases · 20+ private founder disclosures · prices and policies verified 23–24 July 2026.
For founders, a sold-out raise is not yet a funded one. For participants, the window covers the first hours, not the decline that follows. For the market, the announced number can outlive the capital that actually settled.
- In AlphaMind's July 2026 Refund Gap Report, founders of 20+ refundable token launches (2024–2026) disclosed refund rates of 60–95% of announced capital, clustering around 80%; no launchpad publishes an average.
- In 11 of the 16 sales with verifiable price data — 69% — the token traded below its sale price while the refund window was still open (public trackers, 2024–2026).
- In all six documented cases of refunded or cancelled sales, the announced raise was still displayed as a completed raise on major third-party aggregators when checked on 24 July 2026.
- Of the 16 launchpads whose policies AlphaMind read in July 2026, 12 operate a refund or withdrawal mechanism — six by default on every public sale, six per sale or on conditions; two run no speculative window as written and two publish no refund policy at all.
We are not neutral about refundable sales.
AlphaMind operates a committed-capital launchpad — one of the few without a speculative refund window — and we are one of the sixteen rows in this report's own policy table. We collected this data because founders kept describing the same private picture: a raise that looked complete in the announcement and lost most of its capital inside the refund window. We believe demand should be validated before a sale, not unwound after the token starts trading. Read this as advocacy research, openly: our position is declared, every external number carries a source you can check, and figures that come from private disclosures say so. The dataset refreshes quarterly. Corrections: research@alphamind.co.
Watch a sold-out raise fall through zero.
$500,000 − $450,000 − $50,000 = $0, before any campaign cost.
Founders pay on the announced number and live on the retained one. The waterfall below is arithmetic, not a case study — it exists to show which line items survive a refund and which do not. An announced raise on a refundable launchpad is an option book — the project finds out what it raised only after the market prints a price.
For the $EYWA launch we raised from public about $1.5M in total: $1M on AlphaMind and roughly another $500K on two well-known refundable launchpads. On listing day bitcoin was down a few percent and the token briefly traded about half a cent below the sale price. That was enough: the refundable platforms returned around 95% of what they had collected — and still took their fees. They earned their commissions; we received almost nothing from those rounds. The committed raise is what actually paid for our exchange listings. Even then, money was so tight we borrowed $100K from friends for a week to fund market-making liquidity at launch.
How AlphaMind addresses the seven risks.
Vesting existed on paper, not economically. Some participants could claim 10%, sell, then refund the rest. Dashboards showed “fully raised” while funds never reached the project.
A non-refundable agreement is not non-refundable settlement.
The refund era normalised launchpads holding project funds until after listing. In a public 2025 statement, Phron AI said that two of its four launchpads never transferred funds after token generation — and thanked the two that did:
“We want to thank AlphaMind and Poolz Finance, both respected their commitments and supported us through every step of the raise and post-TGE. True professionals.”
— Phron AI, public statement, 2025 · the statement
The only committed capital is capital that transfers when the sale closes. Everything else is a promise with a timing risk attached.
20+ token sale and launch campaigns. The first IDO — $EYWA — reached a $1M hardcap. A public memo is published for every launch before the sale.
You get a written answer to all seven checklist questions before anything is signed. The full checklist ↓
In 11 of 16 verifiable sales, the token traded below its sale price while the refund window was open.
The refund was worth more than the allocation at some point inside the window — and 15 of the 16 sales with verifiable price data ended below sale price anyway, in the window or after it. The window that made refunding possible now runs as short as one hour.
Helios on CoinTerminal, RWA Inc, TAIX AI, Kima and EARN'M held or beat their sale price inside the window; their in-window highs run from +1% to +281%, past the right edge of this axis, so they are shown as rings inside the hatched zone rather than plotted. At least four of the five closed below sale price after the window expired.
Be clear about the trade. The refund protects the individual who acts inside the window — and it does so by draining the raise out of the project that everyone still holding is betting on. It expires before the real decline: windows have shrunk from 14 days to as short as one hour, and 15 of 16 verifiable sales ended below sale price once they closed. What you get instead on a committed-capital launch is information before commitment: a public memo before every sale, and access earned through quests and Karma — no launchpad-token buy-in, no staking wall.
Before your next refundable sale, get three answers: (1) What window does this sale actually run — the announcement, not the policy docs. (2) The exact timestamp the window closes. (3) Whether claiming any tokens forfeits the refund.
Access on AlphaMind is earned through quests and Karma — no launchpad-token buy-in, no staking wall.
Build KarmaThe dataset refreshes quarterly — the next Refund Gap refresh is due Q4 2026 and drops on X first.
Follow AlphaMind on XThe full report PDF ships free in our Telegram — no email, no form, nothing gated.
Join the TelegramThe refund settles the capital. Nothing settles the record.
In all six cases we documented, the announced raise remained visible after the sale was refunded or cancelled.
“Not published” means exactly that: no launchpad or aggregator publishes retained capital, so we do not estimate it. Announced figures are as displayed on third-party listings on 24 July 2026; refunds and cancellations are as announced by the launchpad or the project.
Journalists: per-sale records, archived snapshots and the underlying tables are available on request — research@alphamind.co.
In a market where most launches trade down, a refundable raise converts market-wide price risk into a capital drain that only the project absorbs.
Two ways to structure a public round.
Read across each row. Neither model is risk-free; they differ in when capital settles and who carries the uncertainty until it does.
Fees and marketing are paid on 100% of the announced raise; runway comes from the aqua band. The split is not knowable at signing.
The uncertainty moves earlier: demand has to be validated before the sale rather than unwound after it.
AlphaMind operates the committed-capital structure described in the right-hand column, alongside pre-sale demand validation and a published memo per launch. Two of the sixteen launchpads in the policy table run no speculative refund window as written — we are one of them; two more publish no refund policy at all. Committed capital removes the refund gap, not the risk of a weak launch: no launchpad structure guarantees a raise, a listing or a price.
Seven contractual questions, and where each one leaves you.
Each question exists because something in this report went wrong without it. Read down: every answer you cannot get in writing moves you one step deeper into exposure.
The ladder ships as a one-pager inside the report PDF, with the 16-launchpad policy table and all 21 sale-level records.
Get the One-PagerAlphaMind answers all seven in writing before a campaign starts, and publishes a memo for every launch.
Apply to Launch Your TokenWhich launchpads are refundable: all 16 policies, as written.
Sixteen platforms; identical policies are grouped. Expand any row for the source note behind that reading.
Refunds are available at 12 of the 16 launchpads we read — at six of them on every public sale by default, at six more per sale or on conditions. Two run no speculative window as written. Two publish no refund policy at all.
Colour carries accounting state and nothing else: white is announced, aqua is retained, pink is refunded or lost, hatched grey is unverified. Every other distinction — policy type, above or below a threshold, answered or unanswered, shrinking or unchanged — is drawn with fill, outline, dash, position or a direct label, never with colour.
“Guaranteed Full Refund” badge as the default on public IDOs, 24 hours to 7 days.
Source note
Read from Seedify IDO listing pages and platform docs, 23 Jul 2026. Length varies per sale; the badge itself does not state which.
Grace period on every IDO. Policy reads 7–14 days; recent sales ran as short as one hour.
Source note
Policy text from ChainGPT Pad docs; the one-hour figure comes from a single 2026 sale announcement, not from the written policy. Read 23 Jul 2026.
Platform-wide 14-day right of withdrawal — voided once tokens are distributed or listed.
Source note
Terms of service, withdrawal clause. Read 24 Jul 2026. The voiding condition is the operative detail: distribution can precede the 14th day.
Claim-or-refund at token generation; 24 hours standard, 12 hours seen. Unclaimed allocations auto-refund.
Source note
Formerly Ape Terminal. Read from platform docs and sale pages, 23 Jul 2026. Auto-refund means inaction returns capital — the default is a refund, not a claim.
“Refundable (3 days)” marked per SHO rather than platform-wide; 2020 rSHO terms ran for months without a stated day count.
Source note
Read from DAO Maker sale pages carrying the “Refundable (3 days)” badge and from the 2020 rSHO documentation, 23 Jul 2026. The badge is applied per sale, which is why DAO Maker is classified as per sale rather than default. The original rSHO cover was milestone-based and stated no fixed day count, so it is plotted at the far end of the scale in Fig. 03 and labelled “months”.
Applied per sale. Policy reads 7 days; 12-hour windows observed in practice.
Source note
Read from Polkastarter sale pages and its refund-pilot announcements, 24 Jul 2026. The 7-day figure is the written pilot policy; the 12-hour windows are what recent sales ran.
Chosen per sale from three published policies: a refund if the price falls below the IDO price within 24 hours or within 14 days of token generation, or an optional 5-, 7- or 14-day claim window in which an unclaimed allocation is refunded in full.
Source note
Read from Enjinstarter's published protection policies (Enjinstarter, May 2023) and its Launchpad Protection Protocol (September 2022), checked 24 Jul 2026. The platform states which policy applies to each IDO before Register Interest opens, and says some sales carry none — which is why it is classified as conditional rather than default.
24 hours to 5 days by pool tier; the policy is triggered on underperforming sales.
Source note
Read from Kommunitas docs, 24 Jul 2026. The window is tier-dependent and is triggered on underperforming sales rather than offered on every sale — which is why it counts as conditional, not default. Kommunitas has announced triggered refunds of 80–100% on individual sales.
Two platforms running claim-or-refund on every public sale. Red Kite's written 2022 policy sets 30 minutes, while 2025–26 sales ran 24 hours; BSCS documents a tiered 24–72 hour refund policy.
Source note
Read from each platform's docs and sale pages, 23–24 Jul 2026. Both are counted as default because the mechanic is available on every public sale. Red Kite's 30-minute figure is what the written policy says and the 24-hour figure is what recent practice ran — both are stated because the document and the practice differ. BSCS's window is tier-dependent. The bar pins to the 1-hour floor of the log scale.
“Risk-Free IDO” label applied per sale, 24 hours; USDC auto-refunds seen on 2026 sales.
Source note
Read from Poolz sale pages carrying the “Risk-Free IDO” label and from 2026 refund transactions in USDC, 23–24 Jul 2026. The label appears on selected sales rather than on every public sale, which is why Poolz sits in the per-sale group in Fig. 08.
SAFU 48-hour guaranteed refund, with a per-sale opt-out.
Source note
Read from TrustPad's own “100% guaranteed refunds” documentation, 24 Jul 2026. The written window is 48 hours and individual sales can opt out, which is why TrustPad sits in the per-sale group in Fig. 08.
No speculative refund window. AlphaMind's refunds are integrity refunds only — in both launches that did not proceed, participants were returned 100%, labelled “Refunded” on our launches page, transactions on-chain.
Source note
Read from each platform's terms and docs, 23–24 Jul 2026. Fjord Foundry's purchases are final, with a refund only if a minimum raise is not met. AlphaMind's own integrity refunds — a launch that does not proceed — are labelled on our launches page and the transactions are visible on-chain.
No refund policy found in either platform's official documentation.
Source note
No refund policy found in official docs as of 23–24 July 2026 — BSCPad's helpdesk is closed; Coresky's docs are silent on refunds. Classified as unverified, not as non-refundable.
The 16-launchpad policy table with verbatim clauses and archive links, all 21 sale-level records, the window-shrinkage data — every written policy against what 2026 sales actually ran — the legal chapter and the diagnostic ladder as a one-pager.
Get It on Telegram Posted free in our Telegram channel — no email, no form. Prefer email? Write to research@alphamind.co.Refund Gap questions, answered directly.
What is the Refund Gap?
The difference between the raise a token sale announces and the capital the project keeps after refunds are processed. Announced numbers are published everywhere; retained numbers are almost never published at all.
What percentage of a crypto launchpad raise gets refunded?
Founders who ran refundable launches in 2024-2026 disclosed refund rates to us in a 60-95% range, clustering around 80%, with at least six disclosures of 100%. That is the range of rates reported in those disclosures, not an audited average - no launchpad publishes one.
Do refund windows actually protect investors?
They protect the participant who acts inside the window. In 11 of the 16 sales with verifiable price data the token traded below its sale price while the window was still open, and in 15 of 16 it ended below sale price in the window or after it. Windows have also shortened to as little as one hour, so the protection can expire before most of the volatility arrives.
How do launchpad refunds affect founders?
Fees and marketing are typically paid on the announced number while runway depends on the retained one. As a worked illustration: a $500K target that fully subscribes, a 10% success fee charged on target and a 90% refund rate reconciles to $0 before any campaign costs are counted.
Is a launchpad refund window a legal right?
It is a platform policy, not a statutory right. In the EU, the MiCA withdrawal right runs for 14 days but ends once the token is admitted to trading, so a post-TGE refund window is a contractual construct rather than statutory protection. Terms differ per launchpad and can void on distribution or listing, and a non-refundable agreement is not the same thing as non-refundable settlement. Nothing here is legal advice.
Which crypto launchpads are refundable and which are not?
Of the 16 launchpads whose terms we read in July 2026, 12 operate a refund or withdrawal mechanism. At six it is the default on every public sale: Seedify, ChainGPT Pad, Legion, CoinTerminal, Red Kite and BSCS. Six apply it per sale or on conditions: DAO Maker, Polkastarter, Poolz, TrustPad, Kommunitas and Enjinstarter. Fjord Foundry and AlphaMind run no speculative window; BSCPad and Coresky publish no refund policy at all. Policies change; the report links each source.
What is the difference between a speculative refund and an integrity refund?
A speculative refund lets a participant exit because the price moved. An integrity refund returns 100% when a launch does not proceed. AlphaMind runs the second and not the first.
Where does the data in the Refund Gap Report come from?
Sale prices and in-window lows from public trackers (CoinGecko, CoinCodex, CoinPaprika, CoinMarketCap, GeckoTerminal), verified 23-24 July 2026; refund policies read from official launchpad terms and docs on the same dates; founder figures from private disclosures collected by AlphaMind, 2024-2026, anonymized.
How long do crypto launchpad refund windows last?
Written policies range from 30 minutes to 14 days across the 16 launchpads we read in July 2026. Observed windows are shorter: recent sales ran windows as short as one hour, and five of the six platforms we could trace across time shortened their window - Legion's unchanged 14-day right of withdrawal is the exception. The operative number is set per sale, so check the sale announcement, not the policy page.
Does a sold-out token sale mean the project is funded?
Not on a refundable launchpad. Allocation there works like an option: participants can reclaim capital during the refund window after token generation, so a sold-out raise can settle far below its announced number - founders of 20+ refundable launches disclosed refund rates of 60-95% to us. On a committed-capital launchpad the sale settles as it is collected, so sold out and funded describe the same event.
What is a committed-capital (non-refundable) launchpad?
A launchpad where sale participation is final at purchase: no speculative refund window exists, so the announced raise and the settled raise are the same number, and demand has to be validated before the sale rather than unwound after it. Of the 16 launchpads in this report, Fjord Foundry and AlphaMind run no speculative window as written. An integrity refund - returning 100% when a launch does not proceed - is a separate mechanism, not a speculative window.
16 launchpad policies read from official terms and docs · 21 sales reviewed, 16 with verifiable price data from CoinGecko, CoinCodex, CoinPaprika, CoinMarketCap and GeckoTerminal · 6 documented refund or cancellation cases · 20+ private founder disclosures, 2024–2026, anonymized. All prices and policies verified 23–24 July 2026.
Named third-party statements are those parties' own public claims and are not independently verified by AlphaMind. Figures marked illustrative are scenarios, not measured cases. Settled capital is not published by any launchpad or aggregator, so it is never estimated here. Report refreshed quarterly; corrections to research@alphamind.co.
Robustness: counting tokens rather than sales (Helios ran on three launchpads) gives 9 of 14 below sale price in-window — 64%; the median in-window low across the 11 was −44%, so the finding is not carried by outliers. The 21 sales are recent sales on the four platforms where refund badges appeared most frequently, as of 23–24 July 2026; the five sales without verifiable price data are listed, with reasons, in the full report. Our founder disclosures skew toward founders who chose to talk to a non-refundable launchpad — the range is what was disclosed to us, not a market average. Launchpads' own documents treat refunds as material: Seedify's revenue documentation models a 50% refund case, and Kommunitas has announced triggered refunds of 80–100%.
AlphaMind provides launch infrastructure and campaign support and operates a non-refundable, committed-capital model. Project teams remain responsible for token terms, legal compliance, user eligibility, liquidity and post-sale execution. We do not guarantee profits, listings or token performance.
AlphaMind Research, The Refund Gap Report, July 2026, alphamind.co/crypto-launchpad-refunds/. Data verified 23–24 July 2026; refreshed quarterly.