FI Aruanne Anomaly Detector

Detect discrepancies between any two Finantsinspektsioon reports — half-year vs full-year (I vs II), full-year vs full-year (II vs II), or half-year vs half-year (I vs I). Period is read from each report's timeid, not its filename.

Select exactly 2 XML report files to compare (e.g., "2025 II aruanne8220.xml" and "2026 I aruanne8220.xml")

Processing files...

Change flag thresholds % deviation from expected — drives flags & colors in both tables
What do the flags mean?

Deviation is symmetric. A ratio is measured against its expected baseline — 100% (unchanged) when comparing two different years, or 50% (half of full) for a same-year half-vs-full comparison. Halving and doubling count as equally big: at a 50% threshold, both a drop to 50% and a rise to 200% flag critical.

  • Flagged change — a metric moved past a threshold vs the expected baseline. Coloured INFO WARNING CRITICAL by size.
  • New / Dropped — a category went from nothing to something (new business) or to zero (dropped/omitted) between two years.
  • Impossible — a half-year figure exceeds the full-year figure (same-year only). Physically impossible; must be fixed.
  • Data issue — a negative amount or a non-integer policy count.
  • Derived ratios — avg premium/policy and commission rate. These should stay stable; a jump usually signals a data-entry error even when the headline number looks plausible. Shown in each row's expanded detail.
  • Impact € (Δ premiums) — euros of premium that moved (materiality). Rows are sorted by status first, then by this, so the most serious, biggest-money changes are on top.

Use Show to pick which severities appear (Normal is hidden by default). Low-volume rows (fewer than the policy threshold in both periods) are not flagged, and can be hidden.

Period Comparison Table ▼

Findings & export

Investigation View ▼
Show:
policies
LLM-Friendly Output ▼