Exchange flows is the on-chain measurement of Bitcoin moving onto and off exchange-custodial wallets. Net inflow signals coins moving into the actively-traded float (potential sell-side pressure); net outflow signals coins moving to self-custody or longer-term holding (accumulation). Exchange-balance-as-supply-proxy was one of the most-cited single signals for cycle positioning across 2017-2024. The post-2024 ETF era has complicated the framework: Bitcoin moving onto Coinbase's ETF-backing wallets reflects institutional demand creation, while Bitcoin moving onto its spot-trading wallets reflects potential sell pressure — disentangling the two requires wallet-attribution work that is increasingly important and increasingly difficult. Core metrics include exchange balance, net flow, stablecoin-flow counterparts as a demand proxy, and the derived "speculative supply" framework. The framework remains operational with appropriate ETF-aware adjustment but cannot be deployed naively in the post-2024 regime.
Why this note matters
Exchange flows is load-bearing in three respects:
- It partitions supply by venue. Where Realized price and HODL waves characterize the full UTXO set, exchange-flow analysis isolates the at-the-margin liquid float — the most widely-cited proxy for “what fraction of supply is actually for sale?” across 2017-2024.
- It captures post-2024 ETF flows directly. Net Bitcoin moving into ETF custodial wallets is the single most important demand-side signal of the current cycle; without a flow-and-custody framework, ETF-era market structure is unreadable.
- It bridges cohort frameworks and macro positioning. Exchange-flow signals integrate with Long-term vs short-term holder behavior and Whale behavior — LTH coins moving to exchanges signals distribution; whales accumulating from exchanges signals positioning.
The framework requires more care than Cluster 1 valuation metrics because wallet attribution is imperfect and custodial-cohort identification (spot vs ETF-backing vs corporate custody) is now the dominant analytical question.
What this metric measures
The conceptual claim. Exchange flows measures the net movement of Bitcoin between exchange-custodial wallets and the broader UTXO set. The framework rests on the proposition that exchange-held Bitcoin is closer to being sold than self-custodied Bitcoin — exchange custody is the structural staging ground for sell-side liquidity.
The basic metrics.
- Exchange balance — total Bitcoin currently held in identified exchange wallets, in BTC and as a fraction of circulating supply
- Exchange net flow — inflow (BTC moving onto exchanges) minus outflow (BTC moving off) over a given period (typically daily, weekly, or 30-day)
- Exchange inflow / outflow rates — gross movement, useful for distinguishing high-velocity exchange usage from net positioning
- Exchange-specific flows — per-exchange breakdowns (Coinbase vs Binance vs Kraken etc.) which can reveal venue-specific demand or regulatory dynamics
The post-2024 ETF refinement.
- ETF-custodial wallet flows — specifically the custodial wallets backing spot Bitcoin ETFs (Coinbase Custody for IBIT and GBTC; Fidelity for FBTC; others). These wallets typically grow on net ETF inflow days and shrink on net redemption days. The flow signal here is demand-side (institutional ETF demand) rather than the historical exchange-flow signal (potential sell-side pressure).
- Spot exchange flows — the traditional metric, ideally restricted to exchange wallets not serving as ETF custody. This preserves the historical sell-side-pressure interpretation.
The crucial methodological point: pre-2024 exchange-flow analysis implicitly assumed all exchange wallets served similar economic functions (trade liquidity, custodial holding for traders). Post-2024 requires explicit separation because ETF-backing wallets have categorically different economic function.
What exchange flows is not. Exchange-flow analysis does not directly measure spot-market trading volume or order-book depth. It measures movement of Bitcoin between exchange custody and off-exchange custody. A Bitcoin on an exchange that gets traded internally many times produces no exchange-flow signal until it’s withdrawn. The framework captures the staging-and-positioning layer, not the trading layer.
How it’s calculated
Wallet attribution as the foundation. The framework depends on identifying which Bitcoin addresses are controlled by exchanges. Identification methods:
- Direct attribution — exchanges publicly disclose hot/cold wallet addresses (some do, e.g., for proof-of-reserves)
- Heuristic clustering — transaction-pattern heuristics (common-spend, change-detection, deposit-address-clustering) that infer exchange ownership
- External attribution data — Chainalysis, Glassnode, and other commercial providers maintain large databases of attributed wallets
- Public-reserves disclosure — proof-of-reserves audits identify cold wallets and provide attestations
Inflow/outflow construction. For each block on the chain:
- For each transaction, classify each input UTXO (is it from an exchange wallet?) and each output UTXO (is it to an exchange wallet?)
- Sum the BTC value of inputs from exchanges (gross outflow); sum BTC value of outputs to exchanges (gross inflow)
- Net flow = gross inflow − gross outflow
Aggregation periods. Daily, weekly, and 30-day rolling windows are most common. Multi-day smoothing is operationally important because daily volatility is high.
Exchange-balance dynamics. Exchange balance changes by net flow each day. Track over time as a percentage of circulating supply for cycle-positioning context.
The stablecoin counterpart. USDT and USDC exchange flows serve as a complementary demand-side signal. Stablecoin inflow to exchanges typically signals incoming buying pressure (stablecoins are deployed for purchases); stablecoin outflow signals capital exiting the trading sphere. The cross-validation of Bitcoin outflow + stablecoin inflow has historically been a structurally bullish setup signal.
Specific ETF-flow construction. Post-2024, specific tracking of:
- Coinbase Custody wallet balances (backing for IBIT, GBTC, FBTC custodied by Coinbase)
- Fidelity, Bitwise, and other custodian wallet balances
- Net ETF inflows computed as the daily change in ETF-custodian holdings (cross-validated against the ETFs’ published creation/redemption data)
- Non-ETF exchange balance computed as total exchange balance minus identified ETF-custody balances
Data-provider variants. Glassnode, CryptoQuant, Coin Metrics, and Checkonchain each have proprietary attribution databases producing slightly different exchange-flow time series. Differences in attribution coverage, especially for newer exchanges and custodial entities, produce material differences across providers.
What it tells you
Net flow as a cycle-context signal.
| Net exchange flow regime | Cycle context | Operational reading |
|---|---|---|
| Sustained negative net flow (multi-month outflow) | Accumulation; bull-market setup | Coins moving to self-custody; structural supply tightening |
| Sustained positive net flow (multi-month inflow) | Late-cycle or pre-correction | Coins staging for sale; distribution-phase setup |
| Mixed flows with high gross volume | Active trading regime | Coins churning on exchanges; no clear net positioning |
| Sustained negative flow at cycle bottoms | Post-capitulation accumulation | Strong holders absorbing supply |
| Spike inflow events | Specific distribution events | Often coincides with cycle-top or capitulation triggers |
Exchange balance as supply-proxy. Historical pattern: exchange balance peaked in early 2020 at approximately 18-19% of circulating supply and declined steadily through 2020-2022 to approximately 11-12% — interpreted at the time as massive self-custody adoption. Post-2024 the balance dynamics are dominated by ETF-related custodial-cohort migration, which complicates the historical interpretation.
Specific positioning signals.
- Negative net flow + LTH-supply growth + sub-1 SOPR: structurally bullish accumulation setup; recurring near cycle bottoms
- Positive net flow + LTH-supply decline + LTH-SOPR > 2: late-cycle distribution setup; recurring near cycle tops
- Sudden large inflow event: potential distribution trigger or specific large-holder event; worth investigating
- Sudden large outflow event: potential accumulation or custody migration; often follows ETF creation cycles or institutional buy events
The ETF-flow demand signal. Post-2024, ETF-custodial wallet growth provides direct demand-side signal:
- Net positive ETF inflow (custodial-wallet balance growing) = institutional demand exceeding supply
- Net negative ETF outflow (custodial-wallet balance shrinking) = institutional demand softening or ETF redemptions exceeding creations
- Sustained ETF inflow alongside flat or growing total exchange balance = structurally bullish demand setup
Stablecoin-flow cross-validation. USDT/USDC inflow to exchanges accompanying Bitcoin outflow has historically been a robust bullish setup signal. The cross-validation is part of why systematic frameworks integrate stablecoin flows alongside Bitcoin exchange flows.
Custodial migration vs trader behavior. Large flows can reflect:
- Trader behavior — actual trading positioning by exchange users
- Operational moves — exchange cold-wallet reorganization, security upgrades
- Custodial-cohort migration — Bitcoin moving from exchange custody to ETF custody (or vice versa)
- Specific exchange events — exchange hacks, regulatory actions, exchange failures (FTX collapse in late 2022)
Distinguishing these requires analyst judgment plus external context. The framework can flag anomalous flows; classification often requires additional information.
Empirical track record
Historical exchange-balance peak and decline (2019-2024). The exchange-balance-as-percent-of-supply curve has been one of the more cited single-chart on-chain frameworks:
| Period | Exchange balance (approx) | Context |
|---|---|---|
| Early 2020 (peak) | ~18-19% of supply | Pre-COVID; large active-trading sphere |
| Late 2021 (cycle peak) | ~13-14% of supply | Substantial self-custody migration through 2020-2021 |
| Late 2022 (bear market) | ~12% of supply | Post-FTX self-custody acceleration |
| Late 2023 - early 2024 | ~11-12% of supply | Pre-ETF approval; multi-year low |
| Mid 2024+ | Composite; ETF-custodial growth dominates | Post-ETF regime requires reclassification |
The 2020-2024 decline was widely interpreted as structural supply tightening through self-custody migration; the post-2024 dynamics require ETF-aware reclassification.
Specific cycle-context historical episodes.
- 2020-2021 bull market: Sustained negative net flow alongside LTH accumulation. The “supply shock” framing (Willy Woo) captured this dynamic.
- 2021 April peak: Net inflow spike preceded the intra-cycle peak.
- 2021 November peak: Less extreme inflow spike than April; LTH distribution dominated the cycle-top signal.
- 2022 FTX collapse (November 2022): Massive negative net flow as users withdrew funds from exchanges; balance dropped substantially in weeks.
- 2024 January ETF approval: Coinbase Custody balance began rapid growth backing the newly approved spot ETFs; aggregate exchange-balance metrics shifted in composition.
- 2024-2025 cycle: ETF-custodial growth dominates aggregate exchange-balance dynamics; non-ETF exchange balance has continued declining.
ETF-flow-driven price dynamics (2024+). Post-ETF-approval, daily ETF inflows have been an unusually direct demand-side signal:
- High-inflow days have frequently coincided with substantial spot-price advances
- Net redemption days have frequently coincided with corrections
- Multi-week trends in ETF flows have led multi-week trends in spot price
The directness of the relationship is partly because ETF flows reflect a specific institutional-demand channel that operates at scale. Whether the relationship persists as ETF-flow patterns mature is an open question.
Stablecoin-flow signals. Multi-cycle pattern: USDT and USDC exchange-inflow surges have preceded bull-market initiations and consolidation breakouts; outflow regimes have characterized bear markets. The cross-correlation has been substantial enough that combined Bitcoin-outflow + stablecoin-inflow signals are operationally favored over either alone.
Limitations
Wallet-attribution is imperfect and proprietary. The framework depends entirely on identifying which addresses belong to exchanges. Different providers (Glassnode, Coin Metrics, Checkonchain, CryptoQuant, Chainalysis) have different attribution coverage, especially for newer or smaller exchanges. Cross-provider comparisons can produce different exchange-flow signals from the same underlying data.
ETF-era contamination. The most critical contemporary limitation. Pre-2024 exchange-flow analysis treated exchange wallets as roughly homogeneous (trader custody, sell-side staging). Post-2024 requires explicit separation: ETF-backing wallets serve demand-side institutional function, not sell-side trader function. Without separation, aggregate exchange-balance metrics are misleading.
Custodial migration produces noise. Exchange cold-wallet reorganizations, security upgrades, and operational restructuring produce large flow events that don’t reflect economic decisions. The framework cannot distinguish operational events from positioning events without external context.
Pre-trade staging is not the only path to sale. Bitcoin can be sold via OTC desks, futures markets (cash-settled), and other venues without ever appearing in exchange-flow signals. As institutional adoption deepens, off-exchange sale channels become more important; the framework captures less of total sell-side activity over time.
Stablecoin signals shared with broader crypto. Stablecoin exchange flows reflect demand for all crypto assets, not specifically Bitcoin. The cross-validation with Bitcoin flows is useful but the stablecoin signal alone has degraded specificity as the broader cryptocurrency market has grown.
Daily-frequency noise. Daily exchange-flow readings are noisy. Multi-day smoothing (7-day or 14-day) is operationally required; single-day readings should rarely be treated as definitive.
Geographic and regulatory blind spots. Non-US exchanges have varying attribution quality; many Asian exchanges have limited coverage. Geographic shifts in trading activity can produce flow signals that reflect attribution coverage rather than underlying behavior changes.
Cycle attenuation affects calibration. Like all on-chain extremes, exchange-flow extremes have attenuated cycle-over-cycle. The historical “exchange balance hits new multi-year low triggers bull market” signal is harder to deploy when balance dynamics are dominated by ETF-cohort migration.
The decentralization-narrative confound. Some pre-2024 exchange-balance decline reflected genuine self-custody adoption; some reflected exchange-cold-wallet reorganization that didn’t change ultimate ownership; some reflected aggregation into structures that subsequently became ETF custody. The “structural supply tightening” framing was partially overstated.
Exchange-hack and failure events. Major exchange events (Mt. Gox 2014, FTX 2022, various others) produce extreme flow signatures that aren’t representative of ordinary holder behavior. The framework’s signals during such events require careful contextual reading.
Counter-arguments and tensions
”ETF-era has broken aggregate exchange-flow analysis”
The argument: Post-2024 exchange-balance and net-flow metrics aggregate ETF-custodial dynamics with sell-side trader staging dynamics. The two have opposite economic interpretations (demand vs supply). Aggregate metrics are increasingly meaningless without explicit ETF separation.
Response: Substantially right. Aggregate metrics without ETF separation are misleading in the post-2024 regime. The mitigation: explicit separation of ETF-custodial wallets from spot-exchange wallets recovers the historical interpretation for the non-ETF subset. The framework needs adaptation, not abandonment, but practitioners must do the methodological work. The honest reading: pre-2024 exchange-balance time-series should be treated with caution when compared to post-2024 readings.
Wallet-attribution quality affects everything
The argument: The framework’s signals depend on wallet-attribution quality. Glassnode’s attribution is proprietary and opaque; CryptoQuant and Coin Metrics have different attribution coverage. For newer exchanges, attribution coverage is often weak. Cross-provider differences can be substantial; users without ground-truth attribution have no way to validate signals.
Response: Substantively right and worth taking seriously. Multi-provider cross-validation reduces but doesn’t eliminate the dependence. The honest reading: exchange-flow signals are useful directional indicators with structural uncertainty in absolute magnitudes. Users should treat signals as approximate and confirm via multiple metrics.
”Off-exchange sale channels matter more now”
The argument: As institutional adoption deepens, OTC desks, futures-settled positioning, derivative-market hedging, and other off-exchange channels handle increasing fractions of total sell-side activity. The exchange-flow framework captures a smaller fraction of total positioning each year. Specific signals (large LTH wallet moves coins to OTC desk for institutional buyer) don’t show up as exchange flows at all.
Response: Right. The framework’s coverage has degraded over time. The mitigation: combining exchange-flow signals with other layers (Coin Days Destroyed for old-supply movement, Whale behavior for entity-level positioning, derivative-market metrics for off-exchange positioning) recovers some of the lost signal. No single framework captures all positioning channels; integration is essential.
Stablecoin-flow signals lack Bitcoin-specificity
The argument: USDT and USDC exchange flows reflect demand for all crypto assets, not specifically Bitcoin. A surge in stablecoin inflows could mean Bitcoin demand, altcoin demand, or DeFi-protocol activity. The cross-validation with Bitcoin flows is useful but the stablecoin signal alone is non-specific.
Response: Right. Stablecoin flows are a broader-market signal. The framework’s value is in the combined signal (Bitcoin-outflow + stablecoin-inflow), not in stablecoin-flow alone. The honest reading: stablecoin flows are a contextual indicator, not a Bitcoin-specific demand metric.
”Exchange-flow analysis is theatrical for retail audiences”
The argument: The framework’s accessibility (intuitive “coins going to exchanges = bearish; coins leaving = bullish” framing) makes it popular with retail audiences but the underlying signals are noisy and the methodological caveats are typically ignored in popular discussion. Headlines like “exchange balance hits new low” produce overconfident retail interpretations.
Response: Right as critique of unsophisticated use. The framework requires methodological care that popular communication typically omits. The systematic frameworks (Check, Ryan) embed exchange-flow analysis in broader analytical contexts that support proper interpretation. Standalone exchange-flow charts shared on social media can mislead. The honest reading: the framework is operationally useful with proper context; it can be misleading without.
The “decentralization-narrative” confound
The argument: Pre-2024 exchange-balance decline was widely interpreted as massive self-custody adoption (“decentralization!”). Some of the decline was genuine self-custody migration; some was exchange-cold-wallet reorganization; some was institutional aggregation that subsequently became ETF custody. The framework’s signals were partially over-interpreted in service of a popular narrative.
Response: Substantively right. The 2020-2024 exchange-balance decline reflected multiple distinct phenomena that the popular “self-custody adoption” narrative simplified. Some balance went to genuine self-custody; some moved through cold-wallet reorganization without ownership change; some accumulated in custodial structures that subsequently became ETF-backing. Honest analysis of the decline requires disentangling these threads, which is methodologically harder than the popular framing suggests.
Daily-noise and operational distortion
The argument: Daily exchange flows are dominated by operational events (cold-wallet movements, security rebalancing, batch transactions) that don’t reflect economic decisions. Multi-day smoothing helps but doesn’t fully address the operational-noise contamination. Single-day readings can be misleading even with smoothing.
Response: Right. The framework’s operational use requires multi-day smoothing plus context. The systematic frameworks deploy exchange flows with appropriate aggregation periods (7-day, 14-day, 30-day) and integrate with multiple complementary metrics. Users who treat single-day spikes as definitive will be misled.
”Geographic gaps mean the framework misses regime shifts”
The argument: Asian exchange coverage is patchy. As trading-activity centers shift geographically (Asia-Pacific gains share, US ETF era shifts US activity), the exchange-flow framework may miss structurally important shifts that show up only in poorly-covered exchanges. Apparent “exchange balance decline” could partially reflect activity migration to uncovered venues.
Response: Real concern. The framework’s geographic coverage is a known weakness. Multi-provider integration partially mitigates by combining different attribution coverages. The honest reading: exchange-flow signals are most reliable for well-covered exchanges and less reliable as global pictures.
Open questions for further development
- How should the framework systematically handle ETF-custodial vs spot-exchange wallet separation? Standardized reporting that explicitly partitions ETF custody from spot exchange would strengthen contemporary use. Some providers (Glassnode, Checkonchain) have begun publishing such breakdowns; codifying the convention across providers is an active area.
- What is the appropriate role of stablecoin-flow analysis as Bitcoin’s institutional adoption deepens? Stablecoin flows may become less Bitcoin-specific over time; alternative demand-side proxies (futures basis, ETF flow data, options-market metrics) may complement or replace stablecoin flows.
- How can off-exchange sale channels be incorporated into the framework? OTC desks, dark pools, and derivative-market positioning produce sell-side pressure that doesn’t appear in exchange flows. Integration with broader market-structure data is an open direction.
- What is the appropriate way to handle exchange-failure events (FTX-style)? Major exchange failures produce extreme flow signatures that aren’t representative of ordinary positioning. Frameworks for analytically excluding such events would strengthen baseline signal quality.
- How does exchange-flow analysis interact with cohort frameworks systematically? The intersection of Long-term vs short-term holder behavior and exchange flows is operationally valuable but not fully codified. LTH coins moving to exchanges is a particularly informative subset signal.
- Can wallet-attribution quality be improved through open-source frameworks? Currently dominated by proprietary commercial providers; open-source attribution would democratize the analytical framework.
- What is the appropriate way to engage hyperinflation or major-fiat-regime-change scenarios? Exchange flow dynamics may shift fundamentally if denomination assumptions change; framework adaptation for such regimes is unspecified.
Canonical sources for this note
Primary framework sources
- Glassnode research, various pieces on exchange flows and exchange balance — most-cited contemporary framework source
- CryptoQuant platform — alternative attribution database; substantial exchange-flow analytical infrastructure
- Coin Metrics State of the Network reports — adjacent exchange-flow treatment
- Checkonchain platform — James Check’s analytical framework integrating exchange flows with cohort dynamics
- Chainalysis Crypto Crime Reports and related publications — alternative attribution and flow-analysis perspective
Practitioner literature
- James Check, extensive Glassnode Week On-Chain newsletters during the 2020-2023 tenure — applied exchange-flow analysis across multiple cycles and the FTX episode
- James Check, ongoing Checkonchain platform analysis 2024+ — refined ETF-aware framework
- Ryan (On-Chain Mind), various video analyses applying exchange-flow framework
- Willy Woo, various pieces on supply-shock and exchange-balance dynamics
- Various BitMEX Research pieces on exchange-flow analytical patterns
ETF-era specific literature
- Various Glassnode and Checkonchain analyses of Coinbase Custody and other ETF-backing wallet dynamics 2024+
- ETF issuer disclosure and creation/redemption data (BlackRock IBIT, Fidelity FBTC, Bitwise BITB, others)
- Various practitioner analyses of net ETF flows as Bitcoin demand signals
- Sosovalue and similar dashboards aggregating ETF flow data
Adjacent on-chain literature
- Various analyses of stablecoin exchange flows (USDT, USDC) as Bitcoin-demand cross-validation
- Specific analyses of major exchange events (FTX collapse, Binance regulatory pressure, various others)
- Coinbase proof-of-reserves disclosures and other exchange transparency reports
Critical perspectives
- Engagements with wallet-attribution quality limitations
- Critiques of pre-2024 “self-custody adoption” narrative overinterpretation
- Within-Bitcoin debates about ETF-custodial contamination of historical exchange-flow signals
Related notes
- On-chain analytics and market psychology — sub-MOC parent
- Realized price — definitional foundation; exchange-flow cohort variants build on cost-basis machinery
- MVRV ratio — cycle-positioning valuation metric
- NUPL — cycle-positioning valuation metric
- SOPR — spending-dynamics metric; complementary to flow analysis
- Long-term vs short-term holder behavior — age-based cohort framework; LTH→exchange moves are particularly informative
- HODL waves — age-distribution framework
- Coin Days Destroyed — velocity-and-age metric; exchange-flow events often produce CDD spikes
- Whale behavior — entity-size cohort framework; large exchange flows often associated with whale activity
- Miner flows — adjacent flow framework; miner-cohort spending dynamics
- Sentiment indicators — off-chain sentiment proxies; complementary to flow analysis
- Psychological phases of the market cycle — synthesis where exchange-flow extremes mark phase transitions
- Using on-chain data for macro positioning — operational bridge to macro frameworks; ETF flows are a primary input
- The Power Law model — longer-horizon trajectory framework
- Four-year halving cycles — cycle structure exchange-flow extremes anchor
- Diminishing returns thesis — cycle-over-cycle attenuation framework
- Bitcoin and global liquidity — macro framework; ETF flows interact with broader liquidity dynamics
- Bitcoin and the ISM PMI cycle — macro framework
- Monetization S-curve — adoption framework
- Bitcoin fixed supply and issuance schedule — supply foundation
- The halving - Mechanism — schedule affecting cohort dynamics
- Portfolio approaches to Bitcoin — practical allocation framework exchange-flow signals inform
- James Check — primary contemporary anchor; exchange-flow framework developer
- Ryan - On-Chain Mind — adjacent contemporary anchor
- Dylan LeClair — adjacent on-chain voice
- Giovanni Santostasi — Power Law modeler; adjacent
- Plan B — S2F framework (engaged critically)
- Michael Saylor — Strategy corporate-treasury whale; ETF-era custodial-cohort context