DEX volume: the most abused number in the sector
Reported trading volume is easy to manufacture, easy to double count, and mostly generated by machines rather than by people wanting to trade.
Decentralized exchange volume is the notional value of trades executed through a venue's contracts over a period. Every one of those trades is real in the sense that it settled on a public ledger. Almost none of the interpretation usually attached to the number survives contact with how the number is produced.
Why it is so easy to manufacture
On a centralized exchange, inflating volume requires the operator to print trades in its own database. On a decentralized exchange anyone can inflate volume without the operator's involvement: open two addresses, trade between them through the pool, and repeat. The trades are genuine settlements, so they appear in the on-chain record and in every downstream aggregate. The only real constraint is cost — the swap fee and the gas — and both can be small enough to make sustained wash trading affordable when there is a reason to do it.
There is usually a reason. Programs that distribute tokens in proportion to volume traded make wash trading directly profitable whenever the reward exceeds the round-trip cost. Anticipated airdrops produce the same behavior speculatively, with thousands of addresses generating activity in the hope of qualifying. A venue seeking a ranking, a listing or a partnership has an incentive of its own. None of this requires deceit by the protocol's developers; the incentive is available to anybody with an address.
Machines account for most of what remains
Filtering out wash trades still leaves a total dominated by automated activity. Arbitrage bots trade constantly to keep pool prices aligned with other venues, and each correction is volume. MEV extraction adds more: a sandwich attack around a user's swap creates two additional trades for every one the user made, tripling recorded volume for that interaction while the user receives a worse price.
Arbitrage volume is economically real — it is what keeps a pool's price honest, and it pays fees to liquidity providers. But it is a function of price volatility and of the gap between venues, not of anyone wanting to acquire an asset. A quiet market with wide spreads can generate more of it than a busy one with tight spreads. Reading total volume as user demand mistakes the plumbing for the water.
Routing multiplies the same trade
A user submitting one swap through an aggregator may have it split across three pools and routed through an intermediate asset. Depending on the counting rule, that single intention can register as several trades across several venues, each of which reports its leg. Where a route passes through a stablecoin as a bridging asset, the notional of the intermediate hop is counted as volume too, even though nobody chose to keep it for a moment.
Aggregation across venues then repeats the issue at a higher level. Summing per-venue volume for an ecosystem total counts multi-hop routes multiple times, and the effect grows as routing becomes more sophisticated. A rising ecosystem total can therefore reflect better routing rather than more trading.
Notional is the wrong unit for part of what gets counted. A swap between two dollar-referenced stablecoins registers its full size as volume although almost no economic exposure changed hands, and pools built for such pairs charge very little precisely because the risk of quoting them is small. A venue whose activity concentrates in those pairs reports a large volume figure and a modest fee figure, and both are accurate.
Dividing fees by volume turns this into a usable check. The implied rate that results should sit within the range of the fee tiers the venue actually offers; a rate far below the cheapest tier means the mix has shifted toward the pools where manufactured activity is least expensive, and a rate that moves sharply without any fee change means the composition of trading moved rather than its size.
Checks that are actually diagnostic
No single test proves manipulation, but several readings are hard to explain innocently, and they can be run from the data on this site.
- Volume against deposits. DEX volume to TVL far above the level seen at comparable venues implies the same capital was turned over implausibly often.
- Fees against volume. Large volume with negligible fees points to trading concentrated in near-zero-fee pools, which is where manufactured volume goes because it is cheapest.
- Concentration. Volume dominated by one pair, one pool or a handful of addresses is a different fact from the same total spread across many.
- Persistence. Volume that collapses the week an incentive program ends was the incentive, not the venue.
- Price impact. Real size moves a pool price; volume that leaves price impact and depth unchanged is worth a second look.
Market cap to DEX volume is a useful cross-check for the same reason, though it inherits every distortion above and adds the market's own valuation to the mix.
What the number is still good for
Volume remains the best available measure of whether a venue's contracts are being used at scale, and it is the input to fee income: on most designs, fees are a fixed fraction of volume, so DEX volume 24h and fees 24h should move together. A divergence between them is itself informative, because it means the mix of pools changed.
It is also more auditable than its centralized equivalent. Every trade has a transaction hash, a block, an address and a pool, and anyone can reconstruct the total from on-chain data. The problem is not that the record is hidden; the problem is that the record faithfully includes activity created for the purpose of being recorded.
The DEX section shows volume alongside deposits, fees and pool concentration so the checks above can be made without leaving the page, and methodology states how routed trades are counted here. The next lesson looks at a demand measure that is much harder to fake because its supply is controlled by third parties: stablecoins hosted on a chain.