About this Chart
Annualised protocol fee as a share of PT notional, smoothed over a rolling 7 calendar days. Protocol fee = AMM reserve slice (80%; the 20% LP share is excluded) + full limit-order taker fee + YT interest fee + post-maturity sweep — $45.08M all-time. Partner rewards, PT looping, sPENDLE unstake fees and Boros are excluded, having no market or a flow basis.
The 7-day figure is a ratio of rolling sums — 7 days of fee over 7 days of notional — not an average of the seven daily ratios. Averaging would equal-weight a thin day against a heavy one; summing numerator and denominator separately weights by size, which is the correct smoothing for a rate. Raw daily swung 0.25%–0.40% between adjacent days; the smoothed series reads 0.46%–0.77%.
Each market's first 2 days are excluded, along with any market-day below $100K of PT notional. Both guard the denominator rather than the numerator: a margin is a ratio, and a market still filling produces a meaningless one. sUSDe-27NOV2025 read 96.8% annualised on $0.068 of notional, falling to 2.6% once it held $70M; USUALX-27MAR2025 read 397.8% on its first day.
The first-2-days rule is not free: it removes $768,934 of fee, 2.23% of the numerator, and 69% of that is USDe-25JUL2024 — a market that launched at $107.7M and did $210,529 of fee on day 0. That was genuine points-farming volume, not an artifact. Day-of-life cannot tell the two apart when a market opens large.
PT is valued at par — the accounting-asset price — not at its own discounted market price. Pendle's PT price is exp(−implied_rate × years), so a market-priced denominator would inflate the margin by e^(r·t), biasing it along the very maturity axis this metric exists to compare (PT trades at 0.999 of par within a week of expiry but 0.931 beyond a year). Par is also the base 48% of the numerator accrues on, since yield is charged on the full underlying.
The denominator is PT, not LP TVL — PT is the (asset, expiry) yield contract, so it keeps maturities apart, where a market-grain denominator would repeat across duplicate sibling pools.
Because it is a ratio, series are drawn as lines and average across a timeframe rollup rather than summing — the per-chain, per-market and per-asset margins are independent ratios and do not sum to the total.
Market series are the top 5 per quarter by fee (59 markets), ranked on the same filtered population the chart draws from so every named series plots.
Pre-expiry only; see Pendle: Market Net Contribution for the full population, which deliberately keeps sub-$100K markets because their emissions are real spend.
Caveat: no single denominator matches the whole numerator. Par is right for yield fees (48%), defensible for limit-order fees (27%), and wrong for AMM swap fees (25%), where pool depth earns the fee.