Complaint clusters

Recurring themes mined from public review text and counted — grouped by the aspect they affect, over a rolling window.

What it means

A complaint cluster is a recurring theme that appears across multiple independent reviews in the same 90-day window. Clustering counts how often a theme recurs and tags it to the affected aspect — it is not a record of complaints upheld, disputes resolved, or any legal finding.

Inputs & data fields

FieldTypeMeaning

cluster_idstringStable identifier for a recognised complaint theme (e.g. "kyc_delay").

aspect_tagenumThe aspect the theme affects: kyc, withdrawal, bonus, support…

mention_countintDe-duplicated mentions in the rolling 90-day window.

trendenumDirection vs. prior window: rising · stable · falling.

Process & formula

Cluster review text by theme — clusterer v1.3.0.
De-duplicate, then count mentions per 90-day window.
Tag the affected aspect; compute trend vs the prior window.

Not an allegation. A cluster is a count of how often a theme recurs. It is never a complaint upheld, a finding of wrongdoing, or a legal conclusion.

How it affects public pages

KYCVerification delays↑ rising42

Affects /complaints + the dashboard rollup.

Worked example

Lumen Exchange · fake data · 90-day window ending 2026-06

KYCVerification delays↑42

WithdrawalSlow processing→28

SupportSlow response↓17

Review filtering →Publication states →