The case method behind every reading
Volume Forensics is a small editorial desk that reads Solana transaction records to work out how much of a pair's activity was produced deliberately and how much arrived on its own. This page states the method, the limits and the things the desk will not publish.
The desk exists because the gap between a turnover figure and what that figure means is wider on a permissionless chain than almost anywhere else. Anyone can generate trades, the tooling is sold openly, and the resulting number appears on the same screens as everything else. Reading that gap honestly is a specific skill, and it is the only thing this site is about.
What the desk covers
Three areas, matching the three sections of the site. Signatures describes the shapes produced activity leaves in a transaction record. Evidence covers how the underlying data is pulled so that a second analyst can reproduce or contradict a reading. Grey zone is where several honest explanations survive the evidence and judgement has to be declared rather than hidden inside a confident sentence.
Everything is written about Solana specifically, because the mechanics matter. Slot timing sets the resolution of any cadence claim. Fee payers can differ from traders. Aggregators split one user action into several legs. A method written for a different chain would quietly import the wrong assumptions.
What counts as evidence here
Only material that is in the transaction record and can be pulled again by somebody else: transaction signatures, slots, block times, signers, fee payers, pre and post token balances, program identifiers, inner instructions and pool reserve states. A reading is publishable when a reader could repeat the same pull over the same window and reach an independent conclusion, including a different one.
- Raw fields are stored, not derived ratios, because a ratio cannot be turned back into its inputs.
- Windows are written down before the data is examined, not chosen afterwards to fit a story.
- Routed legs are deduplicated so that one user action is not counted as three trades.
- Gaps in the data are recorded as gaps rather than smoothed over.
- Venue coverage is named, because a figure that names no venues cannot be checked.
Screenshots of dashboards, aggregate numbers with no stated window, and anything relayed second hand are treated as leads worth checking, never as evidence. The full collection procedure is written up in gathering evidence from chain data.
What does not count
Intent is not on chain. Neither is identity, beneficial ownership, nor the reason an account did anything. Every sentence that would need one of those to be true is either removed or rewritten as a hypothesis with its falsifier attached. That single rule removes most of what usually gets published about manipulated charts.
There are also no statistics here about how common produced volume is, how many pairs show a given signature, or what share of turnover on any venue is generated. The desk has no dataset that would support those claims, and inventing a percentage would be the exact failure this site exists to describe.
How confidence is stated
Every inference carries one of four bands: high, moderate, low or undecidable. The band is chosen by the evidence available, not by how interesting the conclusion would be, and each band comes with wording that the write-up has to use so a reader can tell a measurement from a judgement. Undecidable is a legitimate result and is published as often as any other. The standard itself is set out in stating confidence honestly.
What is out of scope
The desk does not name tokens, wallets, teams or projects as manipulated. It does not publish addresses. It does not apply regulated-market terms such as wash trading to any identifiable party, because those terms depend on intent and ownership that a transaction record does not contain. Patterns are described using anonymised or explicitly hypothetical examples throughout.
The desk also does not publish evasion guidance. Explaining how a detection method works is useful to readers judging a chart; explaining how to defeat that method is a different product and this is not it. Where a section naturally raises the question, the answer is a plain refusal rather than a hint.
Finally, nothing here is a forecast. Turnover is a record of trades that already happened, and no amount of processing turns a record into a prediction. There are no price targets, no calls, no win rates and no claim that any reading pays.
How numbers are used
Numbers appear in three forms only. Protocol facts are stated as facts, such as one SOL being 1,000,000,000 lamports, the base transaction fee of 5,000 lamports per signature, and the roughly 400 millisecond slot target. Illustrative arithmetic uses invented round figures and is always labelled as describing no real pair. Anything else is attributed to a named public source that is linked so the reader can check it.
The byline
Articles are signed by The Volume Forensics Desk. There is no invented analyst with a fabricated resume, no stock photograph and no biography, because a fake person is exactly the kind of unverifiable trust signal this site argues against. The desk is the editorial unit that reviews, publishes and corrects the work, and it takes responsibility for every page.
Sourcing and corrections
Sources are linked inline where a claim depends on them, using protocol documentation and public explorers rather than aggregated commentary. External links carry nofollow, which is a technical convention and not a judgement about the source.
Corrections are handled in the open. If a page contains an error of fact, a broken inference or a definition used inconsistently, the page is edited, the change is described in the text where it matters, and the modified date updates. Pages are not silently deleted to make a mistake disappear. Anything the desk gets wrong can be reported through the contact page, and a specific, checkable correction will always get a faster response than a general complaint.
Commercial position
This site describes tools that generate trading activity, and it links to a commercial console in a small number of places where the paragraph is genuinely about that category of tooling. Those links are marked as ordinary links and open in a new tab. They do not change what the desk writes: produced volume is treated as a normal, observable category of Solana market activity throughout, which is the same position the site would hold with or without them.
What the desk will not do in exchange for anything is invent evidence, soften a limitation, publish a rating it cannot support, or name a party it cannot name. If those constraints ever conflict with a commercial arrangement, the constraints win and the arrangement ends.