Calculator

Edge-Case Coverage Gap Estimator

Published May 10, 2026

Production defects in wealth-tech features cluster in code paths that no synthetic household exercised. This calculator turns that intuition into a number: given the corpus you have and the feature you're shipping, expected defect rate per release. And the delta to close it.

What you walk away with

~40s · 5 inputs
  • An expected production-defect rate for the feature given the current corpus.
  • The number of additional archetypes required to drive the rate to target.
  • A linked Edge-Case Coverage Score assessment for the systemic view.

Inputs

How many distinct synthetic households exercise this feature's code paths today.

archetypes

Wash-sale, AMT, RMD-aggregation, IRMAA-cliff, ITIN-filer, etc. Count distinct classes.

classes

Default source: Per Edge-Case Coverage Audit Checklist; typical wealth-tech feature: 4-8

paths
releases
%
Expected defect rate per release
44.4 %

Estimated probability that a given release ships a material edge-case-driven defect. Above 20% is structurally weak; below 5% is audit-grade.

Required archetype count to hit target
103 archetypes

Total archetypes needed to drive expected defect rate to the target. Add this many minus the current count.

Archetypes to add
43 archetypes
Expected material defects per year (annualized)
21.3 defects
Sensitivity

Expected defect rate under different edge-case class counts (e.g. as feature scope expands).

  • Edge cases halved (smaller scope)0.0 %
  • Edge cases as input (default)44.4 %
  • Edge cases doubled (larger scope)72.2 %
Talk to us
Calibration source: Wealth-tech production-defect patterns 2022-2025Calibration is observational — the 1.5-archetypes-per-intersection ratio is the median we've seen at customer firms that have driven their edge-case-driven defect rate below 5% per release. The model is intentionally simple; for a richer view, run the Edge-Case Coverage Score assessment.

FAQ

Doesn't this overstate defect rate at high edge-case counts?

It can. The model is linear in (classes × paths) but in practice some intersections are structurally empty (e.g. 'IRMAA' × 'young accumulator' has no realistic case). For features where many intersections are empty, the worksheet's coverage matrix gives a more accurate read.

Is the 1.5-per-intersection ratio universal?

Roughly. Compliance-touching features need more variants per intersection (closer to 2). Pure-internal features tolerate fewer (closer to 1). The default is the wealth-tech-feature median.