This is an adapted excerpt from the opening of my new book, The DeFi Data Quick-Start: Reading On-Chain Truth with Dune Analytics. If it makes you want to run the queries yourself, that is the whole idea.
The dashboard that lied
Picture a scenario that plays out somewhere in DeFi every quarter.
The foundation behind an L2 launches an incentive program: trade on our network, earn rewards proportional to your volume. Six weeks in, the growth dashboard looks spectacular. Daily DEX volume has tripled, unique trading wallets are up 40%, and someone drafts a blog post about "organic adoption."
Then someone on the data team pulls the underlying trades. A third of the new volume traces back to a few dozen wallets trading the same two tokens back and forth with each other, dozens of times an hour, at sizes calibrated to just clear the rewards threshold. Another slice is bot flow whose activity scales with volume mechanically. Much of what remains routes through aggregators, so nobody can actually say which frontend, if any, deserves credit for bringing the user.
The chart was not wrong. Every transaction on it really happened, and every one is publicly verifiable by anyone with a block explorer. The chart was still a lie, because "volume" was standing in for "adoption," and on-chain those are very different things.
The transparency paradox
Blockchain data is strange: it is at once the most transparent dataset in finance and one of the easiest to misread.
In traditional markets, data arrives pre-interpreted. An exchange reports volume through regulated channels, an auditor signs the books, a vendor cleans and labels the feed before you see it. You trust a chain of institutions, and in exchange the numbers come with meaning attached.
On-chain, the trust model is inverted. You do not have to trust anyone, because every transaction is public, permanent, and independently verifiable. But nothing arrives labeled. The chain records what happened with perfect fidelity and says almost nothing about why it happened or who was really behind it. It is a flawless court stenographer that transcribes every word and identifies none of the speakers.
The specific gap that matters: when a swap executes on a decentralized exchange, no standard field announces where the trade came from. Nothing says "this user came from the Uniswap web app" or "this was a wallet's built-in swap feature" or "this is a market-making bot talking directly to the contract." Everything except the raw movement of tokens has to be inferred.
Skip that inference and just count, and three kinds of flow masquerade as adoption.
Wash trading is a trade with yourself: one entity controls both sides and swaps an asset back and forth to be counted trading, not to trade. Anywhere volume is rewarded, and round-tripping costs a fraction of a cent, washing becomes a yield strategy and people industrialize it. A volume counter cannot tell it from real demand, because it counts every leg the same.
Proxy routing is subtler, because the trades are usually real. Most DEX flow passes through aggregators, routers, wallet swap buttons, and custom proxy contracts before it lands in a pool. A naive dashboard sees the router as the trader, credits the wrong layer, and can tally the same order two or three times as it moves through the stack. The interface that actually brought the user is often invisible on-chain.
MEV-farm flow is the perverse one. In a sandwich attack a bot buys right before a user's swap, lets the user execute at a worse price, and sells right after. Three transactions land where there was one real trade. A single instance of a user being exploited registers as a tripling of activity, so a venue with heavy sandwich activity shows inflated volume precisely because its users are getting worse execution.
What neutral measurement actually requires
If dashboards can be this wrong on data this transparent, honest measurement takes four commitments. They sound obvious. Few published DeFi metrics survive all four.
- Independence from the measured party. The methodology is not controlled by anyone with a stake in the result. That is the difference between an audit and a press release.
- Stated, reproducible methodology. Which trades counted, which were excluded and why, how origin was assigned, what the denominator is. On-chain, anyone can re-run the query and check your work, so faith is not required.
- The same ruler for everyone. One standard across every venue, even when the result is unflattering, measured against an external baseline rather than each venue's self-reported one.
- Honest denominators. State the base rate and pick the denominator before seeing which one tells the better story.
That is the real sense in which on-chain data is different. Not that it is harder to misread than other data, but that for once you can check. The book exists to make you the person who checks.
What the book does with it
The DeFi Data Quick-Start is short and hands-on. In about a hundred pages it goes from your first query against live DEX trades to a dashboard you can defend: navigating the data stack from raw nodes to curated tables, attributing volume to the frontend that actually originated it, detecting sandwich attacks from in-block ordering, and measuring execution quality against a neutral price baseline. Every query in it was run against live on-chain data. The methodology draws on years building data infrastructure at 0x Labs and on ClearTrace, the execution-intelligence system we built to attribute DEX flow across Ethereum, Base, Arbitrum, and Optimism.
If you want to stop taking on-chain numbers at face value, that is what it is for.
The DeFi Data Quick-Start: Reading On-Chain Truth with Dune Analytics is on Amazon, and more is on the author page.
Andrew Maury is the founder of Rantum, a senior data science and ML studio that turns messy, fragmented, and adversarial data into models, APIs, and products that ship.
Top comments (0)