Docs

Keep-rate.

The share of what was converted that is still held, measured on chain at 7 and 30 days. It is the difference between a rule and a farm, and it cannot be faked.

The method

Recorded at settlement, measured at 7 and 30 days by anyone who calls measure_receipt, computed from raw units so a rebase never looks like a sale. Per recipient and asset, over the receipts whose window has been measured:

held      = min( balance_raw_at_measure , Σ amount_raw over those receipts )
keep-rate = Σ held × price_at_settle  /  Σ amount_raw × price_at_settle

price_at_settle = paid_usdc ÷ amount_raw, on the receipt

A person’s balance is one number that has to cover every receipt they have in that asset, so the cap applies to the group, never per receipt: two receipts of 0.065 each against a balance of 0.065 read as 50%, not 100%.

Why it cannot be faked

• The denominator is the receipt’s own amount_raw, written by the program when the tokens landed.
• The numerator is a balance the program read from the recipient’s own token account, on a date anyone can check.
• Neither is ours to choose, and both are reproducible by a stranger with an RPC endpoint.

What it refuses to say

A window that has not matured is not a zero; the page states the date. A receipt that matured but was never measured is excluded and reported, never counted as spent: “they sold it” and “nobody looked” are opposite claims. The 7-day figure exists from the first measurement; the 30-day figure appears when the first receipt is that old.