Communication choice
“2% defective” is usually clearest for acceptance. “One defect per 49 good units” can be more concrete for process teams, but should never replace the raw counts.
Probability
Defect Rate Odds Calculator
Translate inspected and defective counts into defect rate, defect-to-good odds, good-to-defect ratio, and finite log odds with explicit zero and all-defective boundaries.
BINARY QUALITY BALANCE
Express the observed defect balance without hiding boundary cases
Enter inspected and defective unit counts. The calculator preserves raw counts, converts them to probability and odds forms, and labels zero-denominator states rather than returning misleading infinities.
Observed defect rate-
Defect-to-good odds-
Good units per defect-
Conforming rate-
Log odds-
Boundary state-
LIVE DECISION RECORD
Count, rate, and odds reconciliation
The live register shows the raw binary counts and every derived representation of the same observed sample.
| Representation | Numerator | Denominator | Current value | Interpretation |
|---|
CURRENT CALCULATION PROCESS
Formula, current substitution, intermediate values, and reconciliation
p = d/n; defect odds = d/(n-d); good-to-defect ratio = (n-d)/d; log odds = ln(d/(n-d))
| Symbol | Meaning and unit | Current value |
|---|---|---|
| n | Inspected units, units | 600 |
| d | Defective units, units | 12 |
Waiting for valid inputs.
FIVE-STEP USE
Start with the inspection scope, not the percentage
- Freeze the lot, time window, product, and defect classification used for the count.
- Enter every inspected unit, including conforming and defective units.
- Enter defective units once; do not count multiple flaws on one unit as multiple defective units.
- Read probability and odds together, then inspect the named boundary state.
- Export raw counts with the derived values so another reviewer can reconstruct every ratio.
FIVE ODDS FUNDAMENTALS
One sample, four useful representations
- Binary unit
- Each inspected unit is counted once as defective or conforming.
- Defect probability
- `d/n` uses all inspected units as the denominator.
- Defect odds
- `d/(n-d)` compares defective directly with conforming units.
- Reciprocal ratio
- `(n-d)/d` states how many conforming units accompany one defect.
- Log odds
- The natural logarithm turns multiplicative odds changes into additive differences.
DEFAULT CALCULATION
Convert 12 defects in 600 inspected units
Conforming = 600 - 12 = 588; p = 12/600 = 0.02; odds = 12/588 = 1/49
The conforming rate is 98%. The good-to-defect ratio is 588/12 = 49, and the finite log odds are `ln(12/588) = -3.891820`. No display rounding is used in the reconciliation.
THREE DEEPER ANALYSES
Use odds only where their structure helps
Odds-ratio preparation
Finite odds can later compare two processes through an odds ratio. This page deliberately stops before claiming significance or causality.
Boundary governance
Adding arbitrary pseudo-counts to avoid zero may be valid in a declared statistical model, but silently doing so would change the observed evidence. This calculator does not.
TWO QUALITY CASES
Finite and zero-count evidence
Incoming inspection
Twelve defective units among 600 produce finite odds and an auditable 1:49 defect-to-good simplification. Procurement can retain both counts before comparing suppliers.
Zero defects in a pilot
Zero among 25 yields 0% observed and zero defect odds, but not proof of a perfect process. Record “no observed defects” and move to a confidence-limit decision if assurance matters.
ODDS GLOSSARY
Six terms in the exported record
- Inspected unit
- One item receiving a final binary classification.
- Defective unit
- An inspected item failing the stated acceptance definition.
- Conforming unit
- An inspected item not classified defective.
- Defect rate
- Defective units divided by inspected units.
- Odds
- Event count divided by non-event count.
- Log odds
- Natural log of finite defect odds.
LIMITS AND EVIDENCE
Preserve the unit definition
- The model handles defective units, not multiple defects per unit.
- It does not estimate confidence limits, certify a lot, or correct biased inspection.
- Zero and all-defective samples produce named non-finite ratio states.
- Counts from different products or inspection rules should not be pooled without review.
- Results support quality communication and are not a regulatory release decision.
Retain: lot identifier, inspection dates, unit and defect definitions, inspector or system, raw counts, and any reinspection rule.
RELIABLE SOURCES
Primary references for binary defect data
DEFECT ODDS FAQ
Questions about ratios and boundaries
How are odds different from defect rate?
Defect rate divides defects by all inspected units. Defect odds divide defects by conforming units, so 2% probability corresponds to about 1 defect for every 49 conforming units.
Why report both defect-to-good and good-to-defect forms?
The first is convenient for modeling and log odds; the second is often easier to communicate on the shop floor. They are reciprocals only when both counts are positive.
What happens when no defects are observed?
Defect odds are zero, but good-to-defect ratio and log odds require division by zero or log zero. The page reports a named boundary instead of a fake finite number.
Does zero observed defects prove a zero defect rate?
No. It describes this sample only. A confidence interval or acceptance-sampling rule is needed to quantify residual uncertainty.
Can I compare odds from differently sized samples?
You can compare observed odds, but sampling uncertainty differs. Retain raw counts and use interval or regression methods before claiming a stable difference.
Is every nonconformance a defective unit?
Not necessarily. This calculator assumes one binary unit classification. If one unit can carry multiple defects, use a defects-per-unit model instead.
IMPORTANT QUALITY NOTE
Observed balance is not process assurance
Odds faithfully restate the supplied binary counts. They cannot establish whether the sample was representative, the inspection system was capable, or the process will remain stable.