DeFi analytics.

Coverage is Aave v3 on Ethereum and nothing else. This is not the DeFi lending market, not all Aave versions, and not all chains — a liquidation on another protocol or another chain is not missing from these numbers, it was never in scope for them.

The brief asks for a protocol and version breakdown. The source contains one protocol and one version, so that breakdown would be a single-row table implying a comparison the data cannot support. It is stated as the coverage line above instead.

• Count — number of liquidation events whose block_time falls inside the selected window. No deduplication is applied beyond the source's own; one on-chain event is one row.
• By asset — events grouped by collateral symbol and by debt symbol, summing the raw token amounts within each symbol. Amounts are only ever summed across identical tokens, never across different ones.
• Largest events — the source's top five by raw collateral seized. Because the ranking compares raw token amounts, entries in different collateral assets are not directly comparable; read each row on its own terms.
• Cascade — the window is split into 24 equal buckets and events are counted per bucket. A bucket is marked as clustered when it holds at least three events and at least three times the median of the non-empty buckets. Below eight non-empty buckets no bucket is marked, because a median over fewer is not a baseline.
• Missing values — a row without a usable amount is counted as an event but excluded from that asset's total, and the exclusion is stated on the affected row. Nothing missing is ever replaced with zero.
• Freshness — the provider does not publish a refresh interval for this dataset, so freshness is reported as the age of the newest record rather than judged against an invented threshold. The stale marker is set by our ingestion, not inferred in the browser.

All windows and timestamps are UTC.

Coverage is USDC mint events on Ethereum. Not USDT, not other chains, not burns, not bridges — those are absent from the source, not filtered out by us.

• Mint volume — the sum of minted USDC across every mint event whose timestamp falls inside the window. The unit is USDC, not dollars: it is a token quantity, and it is labelled USDC rather than carrying a dollar sign.
• Series — the same events grouped into equal-width buckets. The bucket width is read from the spacing of the series the server returns and stated under the chart, so a chart never implies a resolution it does not have.
• Mint-spike alert — the 95th percentile of the bucket amounts in the current window, nearest-rank, computed over non-empty buckets. A bucket at or above that threshold is flagged. Below twenty non-empty buckets no threshold is computed and no alert can fire, because a 95th percentile of a handful of points is noise wearing a statistic's name.
• Missing values — a bucket the server did not report is left out of the series and out of the percentile. It is never plotted as zero, because zero means "no USDC was minted", which is a claim about the chain rather than about our data.
• Freshness — as with liquidations, the provider publishes no refresh interval, so the newest record's age is reported directly and the stale marker comes from our ingestion.

All windows and timestamps are UTC.