Even the best aggregators miss more than 1 in 1,000 swaps. We grade them against your goals to reach 99.9% reliability.
Enso's measured values, unchanged. Only the colour bands and score weights move with your setup. More redundancy relaxes no-quote and sim-failure. Without simulation, sim-failure grades as your revert rate. Click rows to build your set.
| Aggregator | Score1–10 · weighted | No-quoteavailability | Sim failurereverts | Median lat.speed | p95 lat.tail | Overquoteprice holds | Quote→fill gapavg drift | Costbps |
|---|
Your selected set, traced gate by gate. Thin lines: each source draining as quotes fail to return, arrive late, or fail simulation. Bold line: the chance at least one survives. Redundancy is the gap between them.
Six dimensions at once; further out is better on every axis. Hover an axis label for its exact definition. Selected sources in colour, the rest of the field faint behind them.
Two views of one failure: a quoted price that does not hold. Horizontal: how often quotes are inflated. Vertical: how much fills drift. Bottom-left is honest and tight.
A little drift is normal: prices move in the milliseconds between quote and fill, so a few hundredths of a bp reflects latency, not padding. It becomes a cost signal past the 0.05 bps noise band.
A quote only counts if it arrives in time. One source: its p95 must fit your budget. Several: the median is enough, because the fastest one covers. Chips show the share arriving in time.
Quality is half the picture; an aggregator must also support your chains. There is no single registry. Counts come from live provider APIs where available, otherwise official docs, linked per row.
Quality is one axis; price is the other. Cost stacks in layers: subscription, platform fee in bps, and hidden layers like gas treatment. This compares the published layers at your volume. Undisclosed stays undisclosed, not assumed zero.
All measurements are produced by Enso Shield — an independent benchmark that quotes every aggregator on the identical swap and re-simulates against live chain state (methodology ↗). DFC Research changes none of the numbers; we only re-grade and re-plot them for a given integration setup.
The reliability model. Each source yields a usable quote if it arrives in time and — when simulating — passes simulation. Arrival is estimated from Enso's median and p95 via a log-normal fit. With independent sources, the request fails only if all miss, so reliability = 1 − ∏(1 − uᵢ). The green band for redundancy-covered failures (no-quote; sim-failure when simulating) is 0.001^(1/N): the per-source rate at which N sources still clear 99.9%. Without simulation, sim-failure becomes revert rate directly and redundancy can't mask it, so it hard-pins to 0.1%.
The honest limit. This assumes source failures are independent. Correlated failures — honeypots, transfer-tax tokens, genuinely illiquid routes — fail every source at once and set a floor no amount of redundancy clears. Reaching sub-0.1% in practice is two jobs: enough fast, honest, independent quotes (modelled here) and pre-trade filtering to remove intrinsically unexecutable routes (not in this dataset).
Independence: route99 is built by DFC Research, the research arm of the DeFi Founders Club; some of the measured teams are club members. That is why no number here is ours: all figures come unchanged from Enso Shield, every ranking follows the same published rules for every aggregator, and the source is linked throughout so any figure can be verified directly.