Binary Data’s Bot-Follower Audit

Field notes

The proposed system starts with five account-level fields and scales to a stated 3–4.5 million rows, using proportions and representative samples instead of inspecting every record. Its taxonomy separates active bots, inactive ghosts, purchased batches, and broader suspicious accounts, while creation-date cohorts help identify coordinated groups. The page also specifies that Phase 3 SQL scripts must mirror the written method exactly, and compares final estimates against a historical 2017 benchmark of 9%–15% active bots.