Premium auction outcomes | ENSWhois Analyses

When a .eth name expires past grace, it enters a 21-day dutch auction where price decays from $100M to base. Do these auctions clear, and if so at what premium?

Distribution of premium paid (above base price) for cleared auctions, per length bucket:

Median (P50) - typical clearing premium for that bucket
P90 / P99 / Max - tail behaviour: where the truly-coveted names land in the price decay
Total auction premium - aggregate value extracted by the auction mechanism per tier

What this tells you: Indicates if and where premium pricing creates a meaningful auction-revenue stream, or whether most of the auction volume is cheap longer names cycling through.

Premium-auction clearing prices by length

Names that cleared through the 21-day dutch auction. Percentile columns are the distribution of premium paid per length bucket (in ETH).

Premium share of registration revenue (monthly)

Length	Auctions cleared	Median premium (ETH)	P90 (ETH)	P99 (ETH)	Max (ETH)	Total premium (ETH)
3 chars	944	0.0977	1.66	8.59	21.46	595.77
4 chars	4,951	0.0337	0.1889	1.22	11.00	528.15
5 chars	5,081	0.0040	0.0541	0.4646	7.22	171.28
6–9 chars	10,823	0.0066	0.0834	0.6294	16.50	501.23
10+ chars	4,434	0.0062	0.0636	0.3859	7.76	144.72

Premium-auction surcharge as a percentage of total NameRegistered revenue (base + premium). Highlights how much of any given month's registrar revenue came from auction-cleared names vs. fresh base-price registrations.

Chart starts at 2023-03-31 - that's when the ETHRegistrarController was updated to one that separated the base and premium components of registration cost in NameRegistered events.

Methodology

Premium auction = 21-day dutch auction after grace period ends. Price decays from $100M to 0 over the 21-day window.
All values shown are the premium component only - i.e. the dutch-auction surcharge above the pricing-tier base price. They do NOT include the base registration fee (e.g. $640/yr for 3-char × N years), which is separately determined by the USD price oracle at registration time.
Each row in the percentile table is one NameRegistered event on mainnet .eth controllers where premium > 0. Names that cleared after the auction window ended (premium = 0) are excluded from the percentile distribution but included in the monthly share chart's denominator.
Percentiles are computed across wei values per length bucket.
Monthly share chart pulls from the revenue-by-length snapshot. premium_pct = sum(premium_wei) / sum(base + premium_wei) across all NameRegistered events in that month, all length buckets combined. NameRenewed events never carry premium and are excluded.
The chart only starts when the V2 ETHRegistrarController went live (block ~16,925,617, around April 2023). V1-era controllers emit a single 'cost' field on NameRegistered with no base/premium split, so we have nothing to plot for the V1 era. Months with zero base+premium signal are dropped from the line entirely (rather than rendered as 0%, which would imply 'no premium happened' rather than 'no signal').