Methodology

1. Overview

This site monitors BTC-inbound cross-chain swaps across four protocols — Chainflip, Garden Finance, NEAR Intents, and THORChain — and compares quoted swap prices to actual settlement prices, tracks affiliate fee mark-ups, and maps the frontends, wallets, and aggregators driving each protocol's volume. Data is sourced directly from each protocol's public APIs and on-chain records; nothing here is self-reported by the protocols.

Out of scope: EVM-only DEX flow (tracked separately on the DEX page via Oku/Velora quotes), the Relay aggregator (which has ~13.9M settlements but low per-swap affiliate signal — see section 8), BOB Gateway quote comparisons (tracked on the Monitor page alongside other protocols), and any protocol-native token swaps (RUNE, FLIP, NEAR, SEED/GARDEN) which are excluded from cross-protocol volume comparisons to avoid inflating figures.

2. Protocols Tracked

Chainflip

Chainflip is a native cross-chain AMM using a decentralised validator set and a just-in-time liquidity system. Every swap is routed through a registered broker, making affiliate attribution structurally complete.

• Settlement data: Chainflip explorer GraphQL at explorer-service-processor.chainflip.io/graphql — polled every 5 minutes for new swap completions.
• Affiliate model: Mandatory broker + optional affiliate. Every swap has exactly one broker (SS58 address). A separate affiliate address may be registered on top of the broker. This means every Chainflip settlement has at least one affiliate row, and 0% of settlements are "unattributed".
• Database coverage (all-time): 219,866 settlements, $1.17B USD attributed. USD coverage: 93.7% of settlements priced.

NEAR Intents

NEAR Intents is a solver-auction marketplace on the NEAR blockchain. Users post intents and competing solvers fill them. There is no single canonical frontend — dozens of wallets and applications use the same underlying protocol.

• Settlement data: NEAR Intents explorer API at explorer.near-intents.org/api/v0/transactions-pages — polled every 5 minutes.
• Affiliate model: Two mechanisms. (1) appFees — an array of {recipient, fee} pairs attached to the intent, where the recipient NEAR address identifies the frontend or aggregator. (2) referral — a plain string code. Many frontends use Foundation-controlled fee-collector addresses (e.g. 1csfundsadmin.sputnik-dao.near) as their appFees recipient; in those cases the referral or label field carries the real frontend identity.
• Database coverage (all-time): 975,777 settlements, $1.16B USD attributed. USD coverage: 55.5% of settlements priced (improves to 85.3% when only counting settlements with a price record, out of the full all-time set).

THORChain

THORChain is a decentralised cross-chain liquidity protocol using a UTXO-compatible vault system. Swaps are initiated via on-chain memo fields. No account registration is required to submit a swap — anyone can call the protocol directly.

• Settlement data: Midgard API (vanaheimex.com) — polled every 5 minutes for completed actions of type swap.
• Affiliate model: Optional memo-field code. Memo format: =:CHAIN.ASSET:ADDRESS:LIMIT:AFFILIATE:FEE. The AFFILIATE field is a short alphanumeric code (e.g. t for THORSwap, -_/tps for SwapKit-routed Trust Wallet). This field is absent on the majority of swaps.
• Database coverage (all-time): 3,880,432 settlements, $1.88B USD attributed. USD coverage: 77.4% of settlements priced.

Garden Finance

Garden Finance uses atomic HTLCs (Hash Time-Locked Contracts) to enable trustless BTC swaps. Settlement is atomic but depends on Bitcoin confirmation time. Primarily serves BTC-to-stablecoin and BTC-to-wrapped-BTC routes.

• Settlement data: Garden API at api.garden.finance/v2/orders — polled every 5 minutes for completed orders.
• Affiliate model: An integrator field on quotes and a separate affiliate_fees array on completed orders. Main integrators observed: Phantom and SatsTerminal. Garden Finance itself is a major direct-use frontend.
• Database coverage (all-time): 47,736 settlements, $159M USD attributed. USD coverage: 74.9% of settlements priced.

3. Why Per-Protocol Comparisons Aren't Apples-to-Apples

Chainflip, NEAR Intents, Garden and THORChain all expose a backend protocol that wallets, aggregators and DEX frontends integrate against. The structural difference is not whether there's a marketplace — it's how visible the per-frontend attribution is, and how much volume bypasses affiliate codes entirely.

Chainflip and NEAR Intents — marketplace protocols with comparable concentration

Both expose a public SDK / broker API that any wallet or aggregator can integrate. Both also ship an official web UI (swap.chainflip.io; near-intents.near / new.intents-referral.near) which captures a meaningful share but does not dominate. The top affiliates by USD volume have roughly equivalent shares across protocols:

The compositional mix differs in interesting ways: Chainflip volume is more concentrated in aggregators making direct API calls (Rango direct, Jupiter direct, LI.FI direct add up to ~35% of CF volume), while NEAR volume is more concentrated in wallets shipping their own swap UX (Ledger, Trust Wallet, etc., via SwapKit's SDK). But neither is "single-frontend-dominant" — both are real marketplaces. Chainflip's affiliate model is also mandatory: every settlement carries at least a broker, so 0% of Chainflip USD is "unattributed".

THORChain is direct-protocol-heavy

~85% of all-time THORChain USD volume has no affiliate code at all. The THORChain memo schema makes affiliate tagging optional — users hitting THORNode directly via scripts, custom apps, market-maker bots, or arbitrage systems leave the affiliate field empty. The /affiliates leaderboard's "Unattributed (THORChain)" reference row makes this visible. When reading THORChain affiliates, you are looking at the ~15% slice that chose to tag itself; the larger flow is anonymous-by-default.

Garden uses an integrator field on quotes

Garden's affiliate model is smaller in scope: a few named integrators (Phantom, SatsTerminal) plus Garden's own frontend. The HTLC settlement model means Garden's data set is also smaller (~47K all-time vs ~3.9M for THORChain). Garden is best read as a specialist BTC-to-stablecoin venue with two dominant retail access points.

4. Two-Tier Attribution Model

Each settlement with at least one affiliate row gets exactly one row in settlement_primary_affiliate. This table separates the end-user-facing surface (primary_label) from the routing infrastructure (aggregator_label). Four per-protocol rule sets (in src/enrichment/primary-affiliate.ts) determine how raw rows are resolved.

The Chainflip rule picks the top-fee affiliate row as primary_label and checks whether the broker is a known aggregator. The NEAR rule picks the first wallet/frontend in the resolved rows, then looks for an aggregator row. The THORChain rule strips the -_/ SwapKit prefix and resolves the sub-code. Garden uses the integrator field directly. See src/enrichment/primary-affiliate.ts for the full logic.

5. Canonical Label Consolidation

A single affiliate often appears under multiple raw codes across protocols. Trust Wallet, for example, is trustswapkit.near on NEAR Intents and memo code tps on THORChain (via SwapKit prefix -_/tps). The src/affiliates/labels.ts module maps every known raw code to a single canonical name so cross-protocol analytics work.

Lookup is address-first, with label as fallback. The exception is NEAR Intents Foundation fee-collector addresses (1csfundsadmin.sputnik-dao.near, fefundsadmin.sputnik-dao.near, buybacks.multisignature.near, app-fee.near) — these are shared fee sinks, not wallet identities, so when they appear as the address, the label or referral field carries the real frontend name.

The map currently covers ~80 canonical names across all four protocols. Anything not in the map passes through as-is (raw address or short code). The map is human-maintained and will have gaps for newly onboarded affiliates that have not yet been identified.

6. Coverage Limits

The table below summarises all-time settlement and USD coverage as of the May 2026 audit. The 30-day figures are lower due to a price-collection lag (see note below table).

Quote rate accuracy (Monitor page)

The Monitor page tracks how closely quoted swap rates match actual settled rates (in basis points). Matching uses a ±15-minute window: for each completed settlement, any quote from the same protocol within 15 minutes before settlement becomes a candidate match. The primary delta is (settled_rate - quoted_rate) / quoted_rate × 10,000. Negative values mean the user received less than quoted; positive means they received more. Quotes are polled every 1 minute for BTC input sizes 0.001, 0.01, 0.1, and 1.0 BTC.

7. Data Refresh Schedule

All data collection runs as background cron jobs on the same server as the dashboard. The schedule (from src/index.ts):

API response caching: live endpoints are cached for 10 minutes; heavy pre-computed views (e.g. aggregator comparison matrices) are cached for up to 1 hour.

8. Excluded Data

• Relay: ~13.9M settlements in the database. Relay was excluded from the affiliate analytics primarily because its volume is dominated by very small swaps with low per-swap affiliate signal, and including it would overwhelm the other protocols visually. Relay settlements are still collected and stored but not surfaced in the /affiliates leaderboard.
• Protocol-native token swaps: RUNE↔other on THORChain, FLIP-denominated swaps on Chainflip, NEAR as output on NEAR Intents, and SEED/GARDEN on Garden Finance are excluded from all cross-protocol USD volume comparisons to avoid inflating figures with tokens that have no BTC analogue.
• Failed and cancelled swaps: Settlements where settled_out = 0 are not stored. Only successful completions appear in the database.
• THORChain streaming swaps (partial): Large BTC swaps (>0.1 BTC) may execute over multiple blocks as streaming swaps. The settlement record uses the final outbound timestamp, not the first inbound deposit time. Intermediate partial fills are not tracked separately.

9. Open Questions / Known Gaps

• Unmapped Chainflip SS58 addresses: Some broker and affiliate addresses visible in the raw data have not yet been identified. They appear as bare cF... addresses in the /affiliates leaderboard. Identification requires cross-referencing Chainflip explorer or direct protocol team attribution.
• THORChain direct-user deduplication: The ~85% of unattributed THORChain volume cannot be broken down to individual entities without recipient-address heuristics. Two swaps from the same on-chain address have no shared identifier in the affiliate data. This is a structural property of the THORChain memo system.
• Garden historical unattributed data: 104 unattributed Garden settlements in the 30-day window have non-null raw_json. It is possible the parser misses some older Garden order format variants. A manual spot-check of those 104 records would confirm whether the gap is structural or incidental.
• USD coverage in recent windows: 30-day coverage runs 15–34% across all protocols due to price-backfill lag. Volume figures for recent windows should be treated as lower bounds until backfill catches up. The dashboard displays raw counts alongside USD totals for transparency.
• NEAR multi-label over-count for specific affiliates: blackmarket.tg and a small set of other NEAR affiliates appear with multiple distinct addresses per settlement (4 addresses, same canonical label). The settlement_primary_affiliate table resolves this to one row per settlement, but the raw affiliate leaderboard tab can show inflated per-affiliate USD for those names.
• Chainflip structural multi-label over-count: ~31.8% of Chainflip settlements have more than one distinct identity even after canonical-label collapse (broker + affiliate from separate entities). The de-duped primary-affiliate table is the correct source for volume attribution; the raw per-address breakdown overestimates total attributed Chainflip USD by ~25%.