Is the .eth namespace growing or shrinking, and who is leaving?
Every other analysis here answers "what is true today". This one is the only one that answers "what changed" — it diffs the two most recent snapshots of the same measurements.
What this tells you:
Net direction - active names and holder wallets, then versus now, with the churn ratio: how many names lapse for every one newly registered.
Who is leaving - if the percentage contraction deepens as holder tier grows, the namespace is shedding portfolio holders rather than individual users. That is the opposite of what a price filter is supposed to do.
Which tiers churn hardest - per-length change in active supply, which is the pricing debate's central question asked as a rate rather than a level.
Compared against the snapshot of 2026-04-27 — 2026-04-27 → 2026-09-20, 146 days.
Some comparisons withheld. One of these snapshots predates provenance tracking, so its event dating can't be verified. Comparisons by registration year are withheld; comparisons by length and holder tier are unaffected.
Churn by length
Active names in each length bucket, then and now. Length is independent of event dating, so this comparison is unaffected by re-dating.
Bucket
Active then
Active now
Change
New registrations
3 chars
0
2,780
▲ +2,780
+4
4 chars
0
9,338
▲ +9,338
+72
5 chars
0
96,896
▲ +96,896
+711
6–9 chars
0
379,825
▲ +379,825
+4,500
10+ chars
0
299,580
▲ +299,580
+5,782
Churn by holder tier
Compared against the snapshot of 2026-05-01 — 2026-05-01 → 2026-09-20, 142 days.
Wallets bucketed by how many active .eth names they hold. If the percentage change deepens as the tier grows, the contraction is led by portfolio holders rather than individual users.
Tier
Wallets then
Wallets now
Wallet change
Name change
1 name
350,474
305,166
▼ -45,308(-12.9%)
▼ -45,308(-12.9%)
2–5 names
65,486
59,508
▼ -5,978(-9.1%)
▼ -16,235(-9.4%)
6–10 names
8,644
7,510
▼ -1,134(-13.1%)
▼ -8,537(-13.3%)
11–50 names
6,823
5,814
▼ -1,009(-14.8%)
▼ -22,329(-15.8%)
51–100 names
835
711
▼ -124(-14.9%)
▼ -8,529(-14.7%)
100+ names
622
483
▼ -139(-22.3%)
▼ -38,454(-23.9%)
Some comparisons withheld. One of these snapshots predates provenance tracking, so its event dating can't be verified. Comparisons by registration year are withheld; comparisons by length and holder tier are unaffected.
Methodology
Every figure is the difference between two rows of analysis_snapshots — the latest snapshot for each key and the one before it. The two keys are written by separate scripts, so their comparison windows can differ; each section states its own.
Active names and new registrations come from the cohort-retention snapshot; wallets and holder tiers from the holder-concentration snapshot. Small discrepancies between the two totals are expected — they were computed at different moments and scope buckets slightly differently.
'New registrations' is the growth in names ever registered, which counts a lapsed name re-registered by someone else as new. It is registration flow, not net-new namespace.
Comparisons bucketed by registration year are deliberately absent. Event dates come from block_timestamps where known and a linear block-number approximation otherwise, so as that table backfills, historical events get re-dated and cohort-year counts shift without any registration changing. Length and holder tier don't depend on dating, so they compare cleanly.
Snapshots are generated by hand (bun scripts/analyses/run-all.js), so the window between them is irregular. Per-day rates are the window average, not a measured daily series.