A signature is a property of a set of trades, never of one trade. Any single swap on Solana is unremarkable: an account authorised it, a program executed it, tokens moved. The information sits in how a few hundred of them relate to each other, which is why every file in this section works on a window rather than on an example.
Each signature below is paired with the explanation that defeats it. That pairing is not a disclaimer bolted on at the end, it is the whole method. A signature that cannot be defeated by an innocent explanation would be proof, and nothing in this section is proof.
The shapes produced activity leaves in a transaction record: spacing that is too regular, sizes drawn from a narrow set, round trips that return their own inventory, and counterparty graphs that collapse to a handful of funders.
The difference between spontaneous and produced turnover is a distribution question, not a label. What each side looks like in raw data, and the six checks that separate them without accusing anyone.
Open the case fileInter-arrival gaps and trade sizes carry more information than turnover ever does. How to read spacing, quantisation and slot alignment, and why each one has an innocent explanation.
Open the case fileA round trip returns inventory to where it started. How to detect the pattern from token balance deltas, how it differs from ordinary two-sided trading, and what it cannot prove.
Open the case fileReading a shape is the easy half. The harder half is gathering the data so that someone else can check you, and deciding how firmly to state a conclusion when several explanations survive. Those two problems have their own sections.
How to collect chain data so that a second analyst can reproduce your reading: which fields to store, which to derive, how wallet clusters are built, and where each method quietly stops working.
Open the sectionThe region where the same on-chain footprint has several honest explanations. Market making, incentive farming, treasury rebalancing and produced flow overlap here, and the correct output is a confidence level rather than a verdict.
Open the sectionEvery term this desk uses, defined once and used the same way everywhere: signatures, cluster edges, confidence bands and the words that are deliberately avoided.
Open the glossary