1. Establish a trustworthy item identity
A line must map to a canonical product or supported diagnostic test. Normalised units and relevant diagnostic attributes must agree. Unsupported or ambiguous lines are not forced into a cohort.
Comparison methodology
How Bill Comparisons Work
BillzWise uses recent, consented, privacy-protected bill observations to estimate a local median and expected range—then explains the evidence and its limitations.
Evidence window
Most recent 90 days
Minimum cohort
3 observations · 2 bills · 2 contributors
Main benchmark
Median, not the lowest price
2. Select relevant local evidence
The engine uses active, non-outlier observations from the last 90 days and matches the same provider or broad local area context. The current receipt is excluded from comparing against itself.
3. Require independent evidence
A ready cohort needs at least three qualifying observations from at least two distinct bills and two independent contributors. Provider diversity and recency influence confidence even after the minimum is met.
4. Calculate the verdict
The median is the fairness benchmark. The 25th–75th percentile interval is the expected range. The lowest observed price can suggest an alternative opportunity, but it never becomes the primary fairness benchmark.
5. Show uncertainty
Every result includes evidence counts, providers represented, recency, matching dimensions, confidence, and reasons confidence was reduced. Without enough relevant evidence, BillzWise says so instead of implying a reliable verdict.
A price outside the observed range is a prompt to ask questions, not proof of overcharging, fraud, or wrongdoing. Taxes, bundles, service levels, offers, and unrecorded circumstances can affect a bill.