FlowLens FlowLens

Research library

On-chain research notes built for analysts

This library is where FlowLens writes down how a conclusion was reached, not only what it was. Every note carries the wallet labels used, the clustering decisions made, and the point where the evidence ran thin.

Read a note to learn a method you can rerun, or to check someone else's reasoning against yours.

Go to Guides Wallet intelligence

Reading rail

  • Method notes formats
  • Case notes flows
  • Cluster review second read
  • Corrections in the note
Notes with an open trail stated, not guessed

What belongs in the research library

Not every observation deserves a note. A piece gets published when it introduces a method, applies an existing method to a case worth documenting, or corrects something earlier. Routine readings of a single address belong in the Wallet Tracker, not here.

The library skews toward wallets, clusters, and token flows because those are the questions analysts bring to us most often. Market-structure notes appear when on-chain activity needs outside context to make sense. If you want to see how a single address is read before it earns a note, the Wallet Tracker on the Tools page is the shorter route.

Introduced

A method that did not exist in the library before.

Applied

An established method on a case worth documenting.

Corrected

A published conclusion that later reading did not support.

Close view of a drafting grid surface and a pencil tip on a dark research desk

Two formats

Method notes versus case notes

Documented first

Method notes explain how a signal is built

How addresses get labeled, how a wallet clustering threshold is chosen, what a funding trace can and cannot show, which window a query runs over. A method note is written so another analyst can rerun it and land near the same reading, even if they disagree with the conclusion.

When a note leans on a technique the library has not covered yet, the method note gets written first. That order is not ceremony; it is what keeps a case note from borrowing authority it has not earned.

Applied to a case

Case notes stop where the data stops

A case note traces a real flow of funds through wallets and venues, then ends at the last supportable step. It names what the trace shows and where the evidence thinned out, instead of closing the gap with a confident sentence.

Analyst working at a dark desk with monitors glowing behind them

How wallet labeling enters a note

Labeling is the fastest layer to work with and the easiest to over-trust. A labeled exchange address tells you funds reached a venue, not why. A labeled fund address tells you an entity is involved, not that a position changed direction. Wallet behavior analysis starts where the label stops being enough on its own.

Notes state the label source and the date it was applied in plain terms. Labels drift. An address associated with one service today can be reassigned if the entity changes custody, and the note flags when the label is older than the activity being discussed. If a reader cannot see when a label was set, they cannot judge how much weight it deserves.

Exchange label shows a venue was reached
Fund label shows an entity is involved
Label date always printed in the note

Clustering as a working hypothesis

Where wallet clustering does real work is in reframing a question. "Five wallets bought this token" becomes "these five wallets may be one actor" and the reading changes entirely.

Wallet Intelligence

Clustering groups addresses that probably share control. The signals are behavioral: common funding sources, coordinated timing, repeated interaction with the same contracts, gas patterns that line up. None of those is proof, and a good note says which signals were present and which were absent.

Common funding source

Addresses funded from the same origin within a narrow window.

Coordinated timing

Transactions that land close enough together to suggest one operator.

Shared contracts

Repeated interaction with the same contracts or the same counterparties.

Gas patterns

Fee settings and timing habits that coincide across addresses.

Signal absent

Named explicitly when the evidence does not point one way.

Flow readings

Reading exchange inflows and outflows in context

A research note rarely treats exchange flow as a standalone metric, because the same number can mean opposite things depending on venue and timing. Large inflows to a spot exchange wallet have historically been read as potential sell pressure, while sustained outflows get read as accumulation or movement to self-custody. Both readings are conditional.

Notes pair flow data with the wallets' history and, when it matters, derivatives positioning. CryptoQuant's exchange-flow work and CoinGlass's derivatives dashboards occupy adjacent territory here, which is why a note that touches momentum often points to market-structure context rather than claiming to resolve it from on-chain data alone.

The practical test is repetition. One inflow is a data point; the same direction of flow from wallets with shared history, over a window the note states plainly, is a pattern worth writing about. Follow the reading further on the Token Flows pages, where inflow and outflow rows are set side by side.

Token Flows
Sustained inflow to a spot venue conditional read
Sustained outflow to self-custody conditional read
Wallet history and timing always paired
Resolved from on-chain alone not claimed
Stacked machined metal plates lit from one edge on a dark surface

Cross-chain tracing and bridge hops

Bridges break trails. Funds that leave one chain and arrive on another often land in a fresh address with no direct transaction link, so a note has to reconstruct the path through bridge contracts and timing. That reconstruction is where labeling quality matters most, and it is the part of blockchain transaction analysis readers question first.

We describe the bridge, the deposit and withdrawal windows, and how confident the link is. When a hop cannot be confirmed, the note says the trail is open rather than presenting a guess as a connection. A trail left open is still useful: it tells the next analyst where to look and which assumption not to inherit.

What a bridge note records

  • Which bridge contract carried the funds between chains.
  • The deposit and withdrawal windows observed on each side.
  • How confident the link between the two addresses is, stated plainly.
  • Whether the trail stays open and what would close it.

Notes that involve query-driven data

Some questions need custom queries rather than dashboards. Flipside sits in that query-native space, grouped with SQL workflows and data-team tooling, and it is useful when a hypothesis requires counting something no interface exposes. Method notes built on that kind of query include the filters and the window used so the result can be checked.

Query-native Custom SQL-style counts, with filters and windows printed so the result can be rerun. Flipside
Protocol comparison Broad first questions about where activity and value sit across protocols. DeFiLlama
Position visibility A fast read on where a wallet holds positions across chains. Zapper
Cross-wallet holdings Holdings spread across wallets and exchange accounts in one view. CoinStats

Dashboard-first tools have the opposite strength: speed and consistency. A note usually starts broad with dashboards like these and narrows to a query only when the count it needs does not exist yet. No tool on this row is a partner of FlowLens; each is named as a data reference the note used.

Open the tools

Review before publication

Every note is read by a second analyst before it goes up. The reviewer checks the labeling claims, reruns a sample of the traces, and pushes back on conclusions that outrun the evidence. That review is internal and documented, not a certification of accuracy.

Second-read checklist

  • Labeling claims are checked against the source and the date they were applied.
  • A sample of the traces is rerun to confirm the note reproduces.
  • Conclusions that outrun their evidence are cut or marked as open.
  • When a published note turns out wrong, the correction appears in the note itself rather than as a quiet edit, so readers who cite it can see what changed.

Using the library without over-reading it

A research note is a snapshot of a question at a point in time. It is not a prediction, and it is not a recommendation about any token or wallet. The value is in the method and the reasoning, both of which survive long after the specific flows are irrelevant.

If you want to run the methods yourself, Guides walks through the same steps at a beginner pace. If a note contains a claim you can test and it does not hold, send it to us and we will look at the working.

Wide night view of a research desk with glowing monitors and a desk lamp

Questions readers bring to the library

Is a research note a recommendation about a token or a wallet?

No. A note documents how a question was approached and where the evidence held. It is a method and a reasoning trail, not advice about buying, selling, or holding anything. If you want the practical steps behind a method, that is what the Guides pages are for.

Why do some notes stop before reaching a conclusion?

Because the data stopped there. Bridge hops, thin labeling, and short history all limit what a trace can show. Saying the trail is open is more useful than closing it with a confident sentence, and it tells the next analyst which assumption not to inherit.

Can I rerun the methods from a note myself?

Yes. Method notes are written so the steps can be repeated with the same labeling sources, filters, and windows the note names. Where a step depends on a query, the filters and the window are printed with the result so you can check it against your own run.

Where does GMGN fit into the research?

GMGN is one of several third-party analytical data sources a note may draw on, alongside tools such as CryptoQuant, Nansen, CoinGlass, DeFiLlama, Flipside, Zapper, CoinStats, and Elliptic. Naming a source is a reference to where a number came from. It does not imply a partnership, an endorsement, or shared control over the analysis, and notes credit sources the same way regardless of which one produced the reading.

How does the library grow as new data sources appear?

The library is built to hold more than one source per method. New data sources and analytical modules get documented as they are added, with the same rules: state the source, the window, and the limits. When a new module changes how a signal is built, the method note is revised first and case notes follow.

Read a note, then check the reasoning against your own

The library covers method notes that explain how a signal is built, and case notes that apply a method to a real flow of funds. If a note contains a claim you can test and it does not hold, send us the working and we will look at it.

Read a method note Read a case note

Current notes

What is in the library right now

Notes are grouped by the question they answer. Each entry states its format, so you know before opening it whether you are reading a method or a case.

Method note

Choosing a wallet clustering threshold

Wallet Intelligence
Method note

Reading a funding trace without over-claiming

Research methods
Case note

A bridge hop that leaves the trail open

Token Flows
Case note

Exchange inflow read against wallet history

Token Flows
Method note

Building a query for a count no dashboard shows

Research methods
Correction

A label reassignment and what it changed

Research methods
Wallet Intelligence Request a correction