Errata

Known model errors

A public log of model bugs we found and fixed. Each entry has a date, a severity, the rough blast radius, a root-cause note, and the fix reference.

Legend

Severity bands

P0Wrong number on a bellwether, or a wide ticker-count impact.
P1Wrong number on a small set, or a verdict-band wobble.
P2Cosmetic, single-ticker, or a brief window with limited blast radius.

Log

Entries (most recent first)

2026-05-02P0294 mid-cap tickers

Standalone+consolidated XBRL rows produced inflated CAGR (e.g. FOSECOIND 55% vs real 16%).

Two filing variants for the same year were being summed before the CAGR pass. Reclassifying the synthesised annual row prevents the duplicate from entering the growth calculation.

Fix:period_type='annual_synth' reclassification

2026-05-02P1~10 bellwethers x 30 min

NESTLEIND, TITAN, HDFCBANK showed FV=0.0 verdict=data_limited for ~30 minutes during the null-CAGR gate v1 rollout.

The first cut of the null-CAGR gate was over-aggressive and tripped on legitimate names. Reverted within the 30-minute window; v2 of the gate ships behind a validator with explicit per-sector thresholds.

Fix:Reverted in bccb69c

2026-05-01P11 ticker (peer cohort impact)

KOTAKBANK shares_outstanding stored 5x off (99,464 lakhs vs real 19,900). Caused PE 3.9x in peer cohorts.

Unit-jump corruption in the upstream feed slipped past the validator suite. The direct fix restored the correct count; the broader question of why the validator did not catch a 5x jump is still open.

Fix:Direct UPDATE; upstream ingest investigation ongoing

2026-04-29P11 ticker x ~6 hours

INFY DCF rendered Rs 16.85 (USD-tagged financials being treated as INR).

USD-denominated line items were entering the INR DCF. The canonical price cascade now resolves currency before any per-share computation runs.

Fix:Canonical price cascade

We publish bugs we know about. If you find one, write to hello@yieldiq.in.

Methodology Team Status About