Length Churn Heatmap | ENSWhois Analyses

Every monthly cohort of .eth registrations, scored on whether the names are still held 24 months later.

Each cell is one (registration month × label length) cohort, coloured by 24-month retention.

Green = the cohort mostly stuck
Red = it churned out.

24 months is the pricing-debate sweet spot - long enough that the first renewal decision has happened, short enough that most cohorts qualify for evaluation.

What this tells you:

The 2022 boom collapse - vivid red across mid-2022 cohorts indicates that registrations by NFT-era speculators lapsed two years in.
Pre-boom cohorts - early registrations across all lengths show high retention; the original holders self-selected for long-term commitment.
Recent cohorts read as low - the 24m threshold simply hasn't elapsed yet, not actual churn. Read older months as the most reliable signal.

Hatched cells: cohort younger than 24 months (after 2024-05). Value is a floor - no real measurement is possible yet.

Cohort →

2019

2020

2021

2022

2023

2024

2025

2026

3 chars

100

4 chars

100

5 chars

100

6–9 chars

10+ chars

Methodology

Each label is tracked individually. Cohort = first NameRegistered event per label on mainnet .eth controllers. A cell value is the share of that cohort's labels whose current domains.expiry_timestamp is at least first_registered_ts + 24×30 days.
A 0% cell therefore means: none of the labels registered that month have a current expiry that reaches the 24-month mark - i.e. they all either lapsed without re-registration, or were re-registered with expiries too short to clear the threshold.
Caveat: the check is on the current expiry, so a name that lapsed and was later re-registered by someone else with a long expiry would still count as 'retained'. Real continuous-hold rates are slightly lower than what's shown - the bias only goes one way.
Cohorts younger than 24 months (registered after 2024-05) are marked with diagonal hatching and a vertical boundary line. Their values are mathematically floor-bound — the threshold can't have been satisfied yet — and shouldn't be read as real attrition.