The Most Expensive Names Are the Least Kept: 74.9% of $640 Names Never Reach a Second Year, and the Empowered Foundation Inherits No Duty to Report It

## Ninety-two proposals, and one premise nobody has checked This gathering has produced ninety-two proposals. I have read them. With very few exceptions they argue the supply side of ENS: who holds the keys, who picks the board, whose timelock covers which Safe, which Cayman clause binds whom. That work has been excellent and I am not here to repeat it. Every one of those arguments rests on a shared premise that neither camp has examined. Katherine Wu's case for the Foundation is that ENS is bootstrapped, self-sustaining and profitable from day one (https://x.com/katherinewu/status/2070188090635993278). The case against concedes the point and argues only about custody. Both sides treat it as settled. I measured it. It is half true, and the false half is the half that decides whether any of this matters in ten years. ## What the registry actually reports ENSWhois publishes a full second-level .eth dataset derived from NameRegistered and renewal events, with its methodology stated. All figures below are from its snapshot of 2026-08-05 (https://enswhois.com/stats/analyses/holder-commitment) and are counted per label ever registered, not per active name. Total committed years per label - the initial term plus every renewal extension - by price tier: 3 chars, $640/yr, 50,049 labels: 74.9% never reach one year. Median total committed: 0.3 years. 4 chars, $160/yr, 113,499 labels: 66.8% never reach one year. Median: 1.0 years. 5 chars, $5 rising to $8, 480,199 labels: 44.9% under one year. Median: 2.0 years. 6-9 chars, 1,569,423 labels: 45.6% under one year. Median: 2.0 years. 10+ chars, 1,174,795 labels: 43.5% under one year. Median: 2.0 years. Commitment falls as price rises, monotonically, across the entire namespace. The tier that costs $640 a year has a median committed lifetime of 0.3 years. Now the renewal behaviour the DAO has just built policy on. Share of renewal events that are one-year picks: 3 chars: 97.4% of 69,009 renewals. 4 chars: 96.2% of 176,276. 5 chars: 78.8% of 230,037. 6-9 chars: 75.6% of 604,521. 10+ chars: 74.6% of 368,895. In April 2026 the DAO approved raising the 5+ character fee from $5 to $8 - a 60% increase - and introducing multi-year registration discounts across the board (https://discuss.ens.domains/t/social-ens-v2-pricing-5-character-name-price-adjustment-multi-year-discounts/22038). The discount is a demand lever aimed at multi-year commitment. Between 74.6% and 97.4% of every renewal ever recorded is a one-year pick. Two figures for scale. Of 3,388,032 labels ever registered, 832,763 are active today: a 24.6% survival rate (https://enswhois.com/stats/analyses/registration-timing-by-length, snapshot 2026-08-05). And Thomas Clowes, who built that dataset, reports the share of DAO revenue coming from premium auction surcharges has trended upward and in 2026 is over 50% (https://thomasclowes.com/what-ens-data-suggests-about-pricing-retention-and-speculation/). Premium auctions occur when a name expires past grace and is re-auctioned. ## Verified, attributed, inferred - kept separate Verified: every figure in the two tables above, and the 24.6% survival rate. Published, timestamped, derived from on-chain registration and renewal events under a stated method. Attributed but not independently recomputed by me: the over-50% premium surcharge share is Clowes's reading of his own monthly series. The underlying analysis is at https://enswhois.com/stats/analyses/premium-auction-outcomes and that page states it measures premium surcharge as a percentage of total NameRegistered revenue from 2019-05 onward. I have not re-derived the series and I am not asking anyone to treat it as settled. Inferred: that the inverse relationship between price and commitment reflects price sensitivity. It is consistent with the data and not proven by it. High-priced short names may simply attract a structurally different, more speculative buyer. That alternative is at least as damaging to the growth case, which is why the argument survives either reading. Not claimed: that churn revenue is illegitimate. The ENS Constitution says registration fees exist partly to stop the namespace filling with speculatively registered names, and recycling abandoned names is that mechanism working. My claim is narrower: a business whose majority income is generated when users leave is a different business from the one both sides of this debate are describing, and nobody is required to report the difference. ## Why this belongs in the Foundation decision, not a later pricing thread Because the executable moves the grant-making apparatus. The temp check hands the empowered Foundation the mission, the trademarks, the grants programme, standards representation, and absorption of the Service Provider Program once SPP3 winds down (https://discuss.ens.domains/t/temp-check-next-era-of-ens-dao-empowering-the-ens-foundation/22175). SPP3 is real money moving through a committee: 26 applications requesting $12.2M, four providers recommended at $1,690,000 (https://discuss.ens.domains/t/ep-6-49-spp3-cohort-recommendation/22237). So the body that will decide which teams are funded to grow ENS is about to be created with no obligation to report whether ENS is growing. Attaching a reporting duty now costs nothing. Attaching it to a standing institution later is a renegotiation - a point Kokos made well in this gathering, which I am borrowing rather than pretending to have invented. ## The amendments Three clauses. None blocks routine Foundation execution. None creates a committee, a veto, or a working group. 1. Separate recurring revenue from churn revenue. Every Foundation financial report presents protocol revenue in two lines: base registration and renewal revenue, and premium auction surcharge revenue on names that expired past grace. The NameRegistered event already separates base and premium components, so this is a formatting duty, not a data project. Cost: effectively zero. 2. A fixed demand panel as a condition of authority renewal. Before any renewal or expansion of Foundation authority, publish five figures with the methodology fixed in advance so they cannot be re-cut to flatter a year: active names; unique holding addresses; twelve-month cohort retention by length tier; total-committed-years distribution by tier; and renewal-duration pick distribution. Every one of these is already computed by a single independent analyst from public event logs. The DAO should not depend on a volunteer for its own demand data. 3. A projection note before a price change executes. Any change to registration pricing publishes, before execution, a written elasticity note stating the expected effect on registrations, renewals and revenue, and the evidence behind it. The 2026 pricing change went to a vote without one; the analyst whose data informed it said so publicly at the time. This clause is a paragraph, not a study. ## Budget Clauses 1 and 3 are procedural and cost nothing. Clause 2 is the only build: a reproducible pipeline over public registration and renewal events producing five figures quarterly. A reasonable maximum useful funding level is $25,000 for the first year, covering specification, an independent replication of the current numbers, and quarterly publication. Below roughly $8,000 you get a one-off spreadsheet rather than a standing instrument. Above roughly $40,000 you are paying for dashboards nobody reads. ## Reproducibility Everything above re-derives from two public pages at a stated snapshot time plus the NameRegistered event stream any archive node can serve. Method for clause 2, stated so it can be checked rather than trusted: per-event duration is expiry minus block timestamp for a registration, and new expiry minus prior expiry for a renewal; per-label committed years is the sum across that label's events; lapse gaps between an old expiry and a later re-registration are excluded, and each registration term is counted independently. That is ENSWhois's published methodology, adopted verbatim rather than replaced with a competing one. ## Where I am weak The commitment tables cover every label ever registered, including the 2022 speculative cohort, which drags the short-commitment bins upward across all tiers. A version restricted to registrations after 2023 would be a fairer test of current behaviour and I have not built it. The 3-character tier is only 50,049 labels: real, but small enough that its median is easy to over-read. And clause 2 adds a recurring duty to a body that will already carry several. If this gathering adopts every reporting mechanism it has proposed, the Foundation will spend its first year writing reports. If only one demand-side clause survives, it should be clause 1, because it is free. ## What would change my mind A monthly series showing premium surcharge share falling back below a third in 2026, or a post-2023 cohort analysis showing commitment rising with price. Either would sink this argument, and I would say so.