Unit 2 · Level 2 · Labels & cohorts
Entity labeling
Raw addresses are anonymous strings, but behaviour betrays identity. Analytics firms label addresses as exchanges, funds, market makers or foundations by clustering wallets that move together, tracing known deposit addresses, and sometimes just reading announcements where entities doxx themselves. A labeled blockchain is a completely different dataset from a raw one: suddenly 'address 0x7a3...' becomes 'a market maker rebalancing'.
Free to play. No ads, no token, no account needed to start.
What you get asked
How do analysts typically figure out that a cluster of addresses belongs to one exchange?
Labeling is detective work: co-spending patterns, known deposit addresses, and public slip-ups. There's no official registry. Labels are inferred, not issued.
Match each labeled entity to its typical on-chain behaviour
Each entity type has a rhythm. Market makers churn, funds sit, treasuries drip. Behaviour is the fingerprint that makes labeling possible.
Why does a market maker's €50M inflow to an exchange mean something different from a fund's?
A market maker's job is shuttling inventory wherever they quote, so their flows are plumbing. A long-only fund depositing the same amount is far more likely to be selling. Same flow, different meaning: that's why labels matter.
Grouping addresses that repeatedly transact together and treating them as one owner is called ___, the core technique behind entity labels.
Heuristics like 'addresses spending in the same transaction share an owner' let analysts collapse millions of addresses into thousands of entities.
What's the right level of trust for an entity label?
Labels come from heuristics, and heuristics have error rates. Great analysts use labels constantly AND distrust them a little. You'll see why in lesson 4. 🐜
The rest of this unit
Carefully turning anonymous addresses into funds, foundations and holder tribes.