Methodology & data
### What this research measures
This research describes observed crypto activity and its macroeconomic context. The starting point is a question that the available data can answer. A chart is an explanation of a defined measure, within a stated period and sample.
Comparisons use 7-, 30- or 90-day windows where the measure supports them. Some measures require a fixed window or a longer historical reference. Their pages state that choice. A historical percentile describes a position within the stated sample; it is not a probability of a future outcome.
### Sources and coverage
The existing research names DefiLlama, CoinGecko, GitHub and growthepie among its sources. The source used for a particular result matters more than this general list. Provider coverage and definitions can differ across protocols, chains and dates.
For any figure, read the source, measure, units, observation period and comparison group together. A result for a selected group does not describe the entire market. A missing observation is not zero. A published zero and a missing value require different treatment.
### Definitions before interpretation
Fees paid by users, protocol revenue and income received by token holders are different measures. A change in dollar-denominated value locked can reflect prices, quantities, or both. A price-adjusted estimate of deposits remains an estimate, subject to coverage and changes in asset composition.
Stablecoin supply is a stock measure. Its change does not, by itself, establish how much new investor capital entered crypto. Developer activity records activity in the covered repositories; it does not measure software quality or economic adoption on its own.
### Checks and limitations
The published Data check method flags missing days, isolated observations and revisions where the required history is available. It preserves the source series. A flag identifies something to examine; it does not establish that the source made an error.
The spike and drop checks need surrounding observations. Recent dates can lack the required context. Revision checks need earlier downloads of the same observation. A date with no flag is not necessarily a date on which every check was possible.
Sensitivity calculations show how a result changes when specified observations are excluded. They are alternatives to compare with the published series, not replacements for it. The relevant method page explains the calculation and its exclusions.
### Dates and revisions
The observation date and the publication date answer different questions. Historical notes describe the data available for the stated period. Source revisions can change an earlier observation after publication.
If a figure appears inconsistent, please send the page URL, the measure and the source observation to luca.valente.cv@gmail.com. Corrections belong beside the affected result, with the date and a description of the change.
### Code, data and reuse
The public methods repository states an MIT licence for its code. Its published CSV outputs are labelled CC BY-NC 4.0, which includes a noncommercial restriction and an attribution requirement. Public access does not mean unrestricted reuse.
Third-party data remains subject to its source terms. A code licence does not grant rights to upstream data. Check the licence attached to the specific file and the original provider terms before reuse. Contact me if the permitted use is unclear.
### What this research does not do
This site does not say what prices do next, and it gives no trading instructions or personalised investment advice. It does not identify exchange flows, large wallet holders or supposedly informed investors through unverified wallet labels. Correlation is not treated as causation. An unusual observation is not, by itself, evidence of manipulation, and an incomplete series is not filled with an invented value.
### Inspect the work
Read the method attached to the result. Inspect the linked code and available data. Check the comparison group and the dates before using the conclusion elsewhere.