Geographic intelligence

1. Sum the address’s hourly transaction counts over the trailing 90 complete UTC days, across every chain in its chain_family; n is the total. Addresses with n below 300 get a null region_confidence.
2. The trough is the 6-hour window with the smallest share of n; its centre hour is t. The UTC offset is round(3.5 - t), wrapped to -11..+12, taking 03:30 as the local sleep centre.
3. The band is the one of seven that contains the offset.
4. Depth d = 1 - (trough share / 0.25); night share is the share in the 12 hours centred on the trough.
5. high needs n >= 3000, d >= 0.70, night <= 0.25; medium needs n >= 1000, d >= 0.50, night <= 0.30; low needs n >= 300, d >= 0.30, night <= 0.35; otherwise region and region_confidence are null. A second trough within 10% of the first caps confidence at low.

​

Coverage

​

Tables

• dimensions.accounts region columns, on Accounts ▸ Registry.
• facts_accounts.activity_hours, the hourly evidence region is derived from.
• reports_accounts.activity_coverage, the share of each day’s active senders that are attributed and regioned.
• reports_accounts.region_distribution, a trailing 90-day snapshot of senders by region and confidence.
• reports_accounts.region_hit_rate, region accuracy for accounts with a known expected band.

​

Sample queries

select
    region,
    region_confidence,
    senders,
    sender_share
from `reports_accounts.region_distribution`
where chain_id = 'ethereum'
  and window_end = (
      select max(window_end)
      from `reports_accounts.region_distribution`
      where chain_id = 'ethereum'
  )
order by senders desc

select
    chain_id,
    address,
    region,
    region_confidence
from `dimensions.accounts`
where project_id = 'coinbase'
order by chain_id, region_confidence desc

select
    chain_id,
    attributed_share,
    regioned_share
from `reports_accounts.activity_coverage`
where activity_date = date_sub(current_date(), interval 1 day)
order by regioned_share desc