Decision Rules Guidance

Inputs

Upper Tolerance T_u

Lower Tolerance T_l

Nominal Value (NV)

Measured Value x_m

Measurement Unc. (k=1)

Maximum Allowable Risk (PFA)

In-Tolerance Probability (ITP)

Active Decision Rule

Derived

Tolerance T (B9)

TUR / C_m (B17)

GB multiplier r (B15)

GB width w (B16)

Acceptance upper AL (B20)

Acceptance lower AL (B21)

Decision (B26)

p_c at AL (B28)

Process Yield, ITP (B37)

C_pk (B38)

Worst-case specific risk (B39)

Specific Risk Detail (at measured value)

P(true > T_u | x_m) (B71)

P(true < T_l | x_m) (B72)

Total Specific PFA (B73)

Category (B74)

Usable Spec Range % (B79)

Guard-Band Width by Method (k=2 U)

Decision Rule	GB Width (w)	Accept ±
1. Simple Acceptance	-	-
2. Specific Risk	-	-
3. ILAC G8 (95%)	-	-
4. Dobbert M6	-	-
5. RSS / M5	-	-
6. Fixed 0.98xU	-	-
7. k=2 Subtraction	-	-

Decision Rule

GB Width (w)

Accept ±

1. Simple Acceptance

2. Specific Risk

3. ILAC G8 (95%)

4. Dobbert M6

5. RSS / M5

6. Fixed 0.98xU

7. k=2 Subtraction

Worked Example - UKAS LAB 48: Why Simple Acceptance is Problematic

Measured Value

U (k=2)

Simple Accept?

Specific PFA

Risk Level

3-Outcome Verdict

PFA Sensitivity by TUR

TUR

PFA at x=0.9 T (Simple)

Max PFA

Dobbert M6 GB Width

GB vanishes?

Conformance Probability → Guard-Band Multiplier

r = NORM.S.INV(p_c) / 2, applied as AL = TL − r × U₉₅.

Conf. Probability (p_c)

Max Risk per Side

GB Multiplier (r)

Bias Impact Analysis

Illustrative scenario: known bias = |NV − x_m|, DUT reading = NV + 0.8 T. Same equipment and TUR, with vs without correcting the bias.

Quantity	Without Bias Correction	With Bias Correction
Known Bias	-	n/a
DUT Reading (uncorrected)	-	n/a
Measurement Error	-	-
TUR	-	-
LSL	-	-
USL	-	-
Corrected Measured Value	-	-
PFA (Lower Limit)	-	-
PFA (Upper Limit)	-	-
Total PFA	-	-
Decision (Simple Acceptance)	-	-
Decision (2.5 % Guard Banded)	-	-
PFA reduction from bias correction	-	n/a

Quantity

Without Bias Correction

With Bias Correction

Known Bias

n/a

DUT Reading (uncorrected)

n/a

Measurement Error

TUR

LSL

USL

Corrected Measured Value

PFA (Lower Limit)

PFA (Upper Limit)

Total PFA

Decision (Simple Acceptance)

Decision (2.5 % Guard Banded)

PFA reduction from bias correction

n/a

PFA Calculation for an Entire Loading Range

Instrument Capacity

Resolution UUT

Tolerance (± fraction)

Tolerance basis

Plus 1 Count

CMC (± fraction)

k factor for CMC

Instrument Readings (editable)

Applied

Reading 1

Reading 2

Reading 3

Reading 4

Reading 5

Avg

Method 5 (RSS / Z540.3) - per-reading PFA & acceptance

Applied

Avg

Req Tol

LSL

USL

U (k=2)

TUR

PFA Total

Accept ±

Pass/Fail

Method 6 (Dobbert managed guard band)

Applied

Avg

Dobbert M

Accept ±

Lower

Upper

Pass/Fail

Method 6 acceptance limit: A2% = L - U95% x [1.04 - e^(0.38 ln(TUR) - 0.54)]

Difference in Acceptance Limits - Method 5 vs Method 6

Applied

Avg

M6 Accept ±

M6 P/F

M5 Accept ±

M5 P/F

% Diff in AL

Metrology Cost Model - Production Inputs

UUT Specification (±)

Items Produced / Year

Cost per Item ($)

Cost to Retest Piece ($)

Max Allowable PFA (fraction)

In-Tolerance Probability (ITP)

Total Production Cost-

uUUT (production σ)-

Equipment Comparison (guard band sized to the PFA target)

Equipment

Price ($)

1σ Unc.

TUR

Global PFR

Global PFA

Falsely Rej.

Falsely Acc.

FR Cost ($)

FA Cost ($)

Total Risk ($)

Total Cost Analysis

Equipment

Equip Cost

Production Cost

False Reject Cost

False Accept Cost

Total Cost

Budget vs Baseline

Payback (days)

Cumulative Total Cost Over Time

DMM

Annual Op. Cost

Equip. Cost

Year 1

Year 2

Year 3

Year 4

Year 5

Year 7

Year 10

Sample Size Determination

Reliability Target

Confidence Level

Sample Size (c=0 failures)-

Calibration History (Clopper-Pearson exact)

Total Calibrations (n)

Failures / Out-of-Tol (c)

Upper Confidence Bound (EOPR)-

Lower Confidence Bound (EOPR)-

Observed EOPR-

Failure Rate-

Additional samples needed-

Sample Size for Zero Failures (n = ⌈ln(1−conf)/ln(reliability)⌉)

Target EOPR

80 % Conf.

90 % Conf.

95 % Conf.

99 % Conf.

EOPR Sensitivity (point estimate = 1 − c/n)

Failures ↓ / n →

100

Coverage Factor (k) Reference - Normal Distribution

Coverage Probability

Target Coverage

k for target

Decision Rules & Statements of Conformity

Simplified guidance for calibration laboratories - ISO/IEC 17025:2017, UKAS LAB 48, ILAC-G8:09/2019, JCGM 106:2012.

1. Key Definitions

Term	Definition	Source
Decision Rule	A rule describing how measurement uncertainty is accounted for when stating conformity with a specified requirement.	ISO 17025, Cl. 3.7
Guard Band (w)	An interval between a tolerance limit (TL) and a corresponding acceptance limit (AL), used as a buffer for measurement uncertainty. w = \|TL - AL\|.	ILAC-G8:09/2019
Tolerance Limit (TL)	The specified upper or lower bound of permissible values (the specification limit).	ILAC-G8:09/2019
Acceptance Limit (AL)	The specified bound of permissible measured values after applying a guard band.	ILAC-G8:09/2019
PFA	Probability of False Accept - a non-conforming item is accepted (consumer's risk).	JCGM 106:2012
PFR	Probability of False Reject - a conforming item is rejected (producer's risk).	JCGM 106:2012
TUR	Test Uncertainty Ratio = Tolerance / Expanded Measurement Uncertainty. TUR = T / (2u).	ANSI Z540.3
Simple Acceptance	AL = TL (no guard band). Carries significant PFA near limits; must be paired with a TUR or U constraint to satisfy Cl. 3.7.	ILAC-G8, Def. 1.9

Term

Definition

Source

Decision Rule

A rule describing how measurement uncertainty is accounted for when stating conformity with a specified requirement.

ISO 17025, Cl. 3.7

Guard Band (w)

An interval between a tolerance limit (TL) and a corresponding acceptance limit (AL), used as a buffer for measurement uncertainty. w = |TL - AL|.

ILAC-G8:09/2019

Tolerance Limit (TL)

The specified upper or lower bound of permissible values (the specification limit).

ILAC-G8:09/2019

Acceptance Limit (AL)

The specified bound of permissible measured values after applying a guard band.

ILAC-G8:09/2019

PFA

Probability of False Accept - a non-conforming item is accepted (consumer's risk).

JCGM 106:2012

PFR

Probability of False Reject - a conforming item is rejected (producer's risk).

JCGM 106:2012

TUR

Test Uncertainty Ratio = Tolerance / Expanded Measurement Uncertainty. TUR = T / (2u).

ANSI Z540.3

Simple Acceptance

AL = TL (no guard band). Carries significant PFA near limits; must be paired with a TUR or U constraint to satisfy Cl. 3.7.

ILAC-G8, Def. 1.9

2. ISO/IEC 17025:2017 Requirements

Clause	Requirement	Key Implication
3.7 - Definition	A decision rule describes how measurement uncertainty is accounted for; any conformity statement must be accompanied by a defined decision rule.	No uncertainty = no metrological traceability
7.1.3 - Customer Agreement	The specification, standard, and decision rule shall be clearly defined and, unless inherent in the specification, agreed with the customer.	Proactive communication before measurement
7.8.6.1 - Documentation & Risk	Document the decision rule employed, taking into account the level of risk (PFA, PFR, statistical assumptions).	Must document risk level, not just pass/fail
7.8.6.2 - Reporting	The certificate identifies: (a) which results the statement applies to, (b) which specs met/not met, (c) the decision rule applied.	All three elements on every certificate

Clause

Requirement

Key Implication

3.7 - Definition

A decision rule describes how measurement uncertainty is accounted for; any conformity statement must be accompanied by a defined decision rule.

No uncertainty = no metrological traceability

7.1.3 - Customer Agreement

The specification, standard, and decision rule shall be clearly defined and, unless inherent in the specification, agreed with the customer.

Proactive communication before measurement

7.8.6.1 - Documentation & Risk

Document the decision rule employed, taking into account the level of risk (PFA, PFR, statistical assumptions).

Must document risk level, not just pass/fail

7.8.6.2 - Reporting

The certificate identifies: (a) which results the statement applies to, (b) which specs met/not met, (c) the decision rule applied.

All three elements on every certificate

3. Decision Rule Comparison - Core Approaches + Variants

Decision Rule	Guard Band Formula	Risk Framework	Target Risk	Reference
1. Simple Acceptance (Shared Risk)	GB = 0 (AL = TL)	Neither - risk ignored	None (shared)	ASME B89.7.3.1; ILAC-G8 6.1
2. Specific Risk Guard Band	AL = TL - (GB_mult x U); GB_mult = NORM.S.INV(1-risk)/2	Specific (conditional)	≤ 2.5 % PFA per side	ILAC-G8; JCGM 106
3. ILAC G8 (95 %, bilateral)	w = u x 1.96	Specific (conditional)	Pc ≥ 98 %	ILAC-G8; ISO 14253-1
4. Dobbert Method 6 (Z540.3)	GB = U x M(TUR); M = max(0, 1.04-e^(0.38 ln(TUR)-0.54))	Global (unconditional)	PFA ≤ 2 %	ANSI/NCSL Z540.3
5. RSS / Method 5 (Z540.3)	GB = U x TUR x (1 - √(1 - 1/TUR²))	Global (unconditional)	PFA ≤ 2 %	ANSI/NCSL Z540.3
6. Fixed 0.98 x U Guard Band	GB = 0.98 x U	Specific-equivalent (fixed)	Conservative (pc ≈ 0.975)	ILAC-G8 (simplified)
7. k=2 Subtraction (AL = T - U)	GB = U	Producer-biased	Conservative	MIL-STD-45662A

Decision Rule

Guard Band Formula

Risk Framework

Target Risk

Reference

1. Simple Acceptance (Shared Risk)

GB = 0 (AL = TL)

Neither - risk ignored

None (shared)

ASME B89.7.3.1; ILAC-G8 6.1

2. Specific Risk Guard Band

AL = TL - (GB_mult x U); GB_mult = NORM.S.INV(1-risk)/2

Specific (conditional)

≤ 2.5 % PFA per side

ILAC-G8; JCGM 106

3. ILAC G8 (95 %, bilateral)

w = u x 1.96

Specific (conditional)

Pc ≥ 98 %

ILAC-G8; ISO 14253-1

4. Dobbert Method 6 (Z540.3)

GB = U x M(TUR); M = max(0, 1.04-e^(0.38 ln(TUR)-0.54))

Global (unconditional)

PFA ≤ 2 %

ANSI/NCSL Z540.3

5. RSS / Method 5 (Z540.3)

GB = U x TUR x (1 - √(1 - 1/TUR²))

Global (unconditional)

PFA ≤ 2 %

ANSI/NCSL Z540.3

6. Fixed 0.98 x U Guard Band

GB = 0.98 x U

Specific-equivalent (fixed)

Conservative (pc ≈ 0.975)

ILAC-G8 (simplified)

7. k=2 Subtraction (AL = T - U)

GB = U

Producer-biased

Conservative

MIL-STD-45662A

4. Non-Binary Decision Rules (3-Outcome Model)

Outcome	Condition	Specific PFA	Action
PASS	Measured value within acceptance limits (AL)	≤ agreed threshold (e.g. 2.5 %)	State conformity with confidence
POSSIBLE PASS	Within tolerance (TL) but outside acceptance (AL)	> threshold but < 50 %	Report MV, U, calculated PFA; customer decides
FAIL	Specific PFA ≥ 50 %, or outside TL by more than U	≥ 50 %	State non-conformity

Outcome

Condition

Specific PFA

Action

PASS

Measured value within acceptance limits (AL)

≤ agreed threshold (e.g. 2.5 %)

State conformity with confidence

POSSIBLE PASS

Within tolerance (TL) but outside acceptance (AL)

> threshold but < 50 %

Report MV, U, calculated PFA; customer decides

FAIL

Specific PFA ≥ 50 %, or outside TL by more than U

≥ 50 %

State non-conformity

9. Sizing Rules & Quick Math (DRG V1.42)

Rule of thumb: at shared risk (no guard band), TUR ≥ 4.6:1 keeps PFA ≤ 2 % (program-level). Required U = Tolerance / 4.6.

Device Tolerance (T)	Required U (k=2)	Resulting TUR	Global Risk
0.0001	-	4.6 : 1	PFA ≤ 2 %
0.0005	-	4.6 : 1	PFA ≤ 2 %
0.001	-	4.6 : 1	PFA ≤ 2 %
0.005	-	4.6 : 1	PFA ≤ 2 %
0.01	-	4.6 : 1	PFA ≤ 2 %

Custom Sizing - enter your tolerance: → Required U (k=2): -

10. Decision Rule Selection Framework

"Start With the End in Mind" - work backward from the risk target to the equipment:

Step	Action
A	Set the risk target and decision rule (e.g. PFA ≤ 2 %); state it on POs and calibration instructions.
B	Convert the risk target to required capability (TUR ≥ 4.6:1 at shared risk, or EOPR ≥ 89 %).
C	Translate capability to equipment and provider requirements (U_exp ≤ Tolerance / 4.6).
D	Validate with reliability data and control charts; see the Reliability tab for EOPR estimation.

Step

Action

Set the risk target and decision rule (e.g. PFA ≤ 2 %); state it on POs and calibration instructions.

Convert the risk target to required capability (TUR ≥ 4.6:1 at shared risk, or EOPR ≥ 89 %).

Translate capability to equipment and provider requirements (U_exp ≤ Tolerance / 4.6).

Validate with reliability data and control charts; see the Reliability tab for EOPR estimation.

12. Five Rules to Reduce Measurement Risk

#	Rule
1	Know the Right Requirements - engage metrology before purchase.
2	Use the Right Equipment - match CMC and traceability to tolerance; look beyond sticker price.
3	Follow the Right Processes - training, adapters, environment, SOPs.
4	Check Your Work - checklists, cross-checks, measurement assurance.
5	Stay Vigilant - SPC charts, interval review, reliability tracking.

References: ISO/IEC 17025:2017 - UKAS LAB 48 - ILAC-G8:09/2019 - JCGM 106:2012 - ANSI/NCSL Z540.3 - ASME B89.7.3.1 / B89.7.4.1 - ISO 14253-1 - Decision Rule Guidance (Morehouse) V1.42. Morehouse Instrument Company - www.mhforce.com

JCGM 106:2012 Formula Validation

The app's engine benchmarked against JCGM 106:2012 worked examples. The full 182k-cell workbook grid (incl. 40,000 Halton NORM.S.INV samples) is validated headlessly; this tab shows the headline checks.

Table 1 - Resistor Example (Global Consumer & Producer Risk)

JCGM 106 9.5.3: TU=1500.2, TL=1499.8, u=0.04, u₀=0.12, AU=1500.18, AL=1499.82.

Quantity	Engine (BVN)	Exact analytic	JCGM published	Δ (engine-exact)	Status
Global consumer risk RC (PFA)	-	-	1.0 %	-	-
Global producer risk RP (PFR)	-	-	6.9 %	-	-

Identity Checks (closed form)

Quantity	Computed	Δ	Status
Cm = (TU-TL)/(4u)	-	-	-
Cp = (TU-TL)/(6u₀)	-	-	-
Guard-band mult r = (TU-AU)/(2u)	-	-	-
Specific-risk Pc (eq.14, two-sided)	-	-	-

Quantity

Computed

Status

Cm = (TU-TL)/(4u)

Cp = (TU-TL)/(6u₀)

Guard-band mult r = (TU-AU)/(2u)

Specific-risk Pc (eq.14, two-sided)

Engine Identity vs Risk Calc & Curves

Quantity	Engine value	Recomputed	Δ	Status
Half-width T = (TU-TL)/2	-	-	-	-
TUR = T/(2u)	-	-	-	-
1-ITP at xm	-	-	-	-

Quantity

Engine value

Recomputed

Status

Half-width T = (TU-TL)/2

TUR = T/(2u)

1-ITP at xm

Guard-Band Width vs TUR (T=10 reference)

TUR

u = T/(2·TUR)

U = 2u

0.98·U

Acceptance frac.

Spec PFA @ AU

About This Tool

Standalone, single-file web port of the Decision Rules Guidance Worksheets For Web workbook. Every shown value matches the workbook cell to the IEEE-754 floor (HARD ≤ 1e-12 relative; exact for integer/rational; character-identical for TEXT). One math engine on window.DRG; the UI computes nothing.

Explore Further

Workbook Metadata

Reference workbook

Decision Rules Guidance Worksheets For Web .xlsx

App version

V1.3

Authors

Henry Zumbrun - Morehouse Instrument Company · Greg Cenker - IndySoft

Web port

Harrison Zumbrun - Morehouse Instrument Company

Standards

ISO/IEC 17025:2017 · ILAC-G8:09/2019 · JCGM 106:2012 · ANSI/NCSL Z540.3

Changes Log

Rev	Date	Author	Sections	Summary
V1.3	2026-06-10	Harrison Zumbrun	About; resources; guide banner	About This Tool: removed the EndInMind cross-reference and added an Explore Further section (Start With the End in Mind guide, the Quality Magazine decision-rules article, the free Force Calibration eBook, and the Morehouse technical blog). Added a site-wide banner in the toolbar linking the Start With the End in Mind guide. No calculation changes.
V1.2	2026-06-10	Harrison Zumbrun	UI; authorship	Moved the title and subtitle into the letterhead, centered between the Morehouse and IndySoft logos; the navigation is now a full-width green strip. Fixed table header/column misalignment on all results tables (the table class name collided with Tailwind's .grid utility, which set display:grid; display:table restored). Workbook authorship corrected: authors Henry Zumbrun (Morehouse Instrument Company) and Greg Cenker (IndySoft); web port by Harrison Zumbrun (Morehouse Instrument Company). No calculation changes.
V1.1	2026-06-10	Harrison Zumbrun	Risk Calc sheet complete; tab scenario gates	Ported the last Risk Calc regions: the specific-risk chart scaffolding (D2:E82 measurement-pdf grid, R/S product pdfs, TL/TU/NV/AL markers, P/Q TEXT labels) rendered as a new Specific Risk chart, and the UKAS LAB 48 worked-example table (exact strings incl. the workbook's verdict glyphs). Risk Calc Layer 2 now 957 cells, 593 bit-identical. Added Excel-COM alternate-input scenario gates for Method/Cost/Reliability (layer2_excel_scenarios_tabs.py): 11 scenarios, 2,588 cells, zero failures.
V1.0.1	2026-06-10	Harrison Zumbrun	Single-sourcing; verifier pass	Two-independent-verifier adversarial pass (both PASS; each tab's gate confirmed to catch a 1 % corruption). Routed the two UI Dobbert-M sites that match the engine formula through E.dobbertM() (bit-identical; Method 6 keeps its distinct clamped-TUR/no-max variant, documented). Tightened the check_engine.js smoke floor to 1e-12.
V1.0	2026-06-09	Harrison Zumbrun	IO layer; all tabs complete	Added the import/export/session layer: per-tab CSV export, CSV import for the Method reading grid and Cost equipment list, session JSON save/load, continuous localStorage autosave, and chart PNG export (graceful offline fallback). Validated by validators/test_io.py (12/12). All six workbook tabs now ported and exactly validated.
V0.9	2026-06-09	Harrison Zumbrun	JCGM 106 Validation	Ported the JCGM 106 Validation report tab: the resistor worked-example global risk (PFA_L/PFR_L vs exact analytic vs JCGM published), identity checks, specific-risk example, engine-vs-Risk-Calc identity, and the guard-band-width-vs-TUR risk-curve table. Layer 2 (layer2_jcgm.py): 81 cells, 70 numeric within 1e-12 + 11 string-exact, 69 bit-identical. The 40,000 Halton NORM.S.INV grid samples are covered by Layer 0.
V0.8	2026-06-09	Harrison Zumbrun	Guidance Summary	Ported the Guidance Summary reference tab: definitions, ISO 17025 requirements, the 7-method decision-rule comparison, the 3-outcome model, the sizing-rules table with the Required-U = Tolerance/4.6 formulas + custom-sizing input, the selection framework, and the five rules. Layer 2 (layer2_guidance.py): 6/6 sizing cells bit-exact.
V0.7	2026-06-09	Harrison Zumbrun	Reliability	Ported the Reliability tab: zero-failure sample size (CEILING(ln(1-conf)/ln(reliability))), Clopper-Pearson exact bounds (BETA.INV), observed EOPR / failure rate, additional-samples-needed (AdditionalSamplesNeeded), the zero-failure reference grid, the EOPR sensitivity grid, and the coverage-factor k table. Layer 2 (layer2_reliability.py): 74 cells, all within 1e-12, 65 bit-identical.
V0.6	2026-06-09	Harrison Zumbrun	Cost Model	Ported the Cost Model tab: production inputs, the 7-DMM equipment comparison (guard band sized by PFA_GB_SOLVER_ASYM to the PFA target, global PFA/PFR via PFA_L/PFR_L, ROUND-based false-reject/accept costs), total-cost analysis with budget-vs-baseline and payback, key-insight figures, and cumulative-cost projections. Layer 2 (layer2_cost.py): 192 cells, 173 numeric within 1e-12 + 19 string-exact, 161 bit-identical; the guard-band bisection solver reproduced exactly.
V0.5	2026-06-09	Harrison Zumbrun	Method 5 & 6 & TL	Ported the full Method 5 & 6 & TL tab: config + editable 10-point reading grid, per-reading Method 5 (RSS, NORM.INV acceptance) and Method 6 (Dobbert, TUR clamped to 4) tables with summaries, the M5-vs-M6 comparison, and the inlined Method-6 acceptance-limit diagram. Layer 2 (layer2_methods.py): 429 cells, 389 numeric within 1e-12 + 40 string-exact, 374 bit-identical.
V0.4	2026-06-09	Harrison Zumbrun	Risk Calc - methods/sensitivity/reference/bias	Added the 7-method Accept-± column, the PFA sensitivity-by-TUR table, the conformance→guard-band-multiplier reference table, and the bias-impact analysis block (all derived from the main inputs). Layer 2 extended to 592 cells, all within 1e-12, 353 bit-identical.
V0.3	2026-06-09	Harrison Zumbrun	Risk Calc - curves & charts	Wired the distribution chart-data arrays (A92:A171 product/measurement/biased pdfs) and the Global Risk Curves (PFA vs ITP at TUR 1:1-4.6:1 via PFA_SYMMETRIC); rendered both with Chart.js (graceful offline). Layer 2 extended to 460 cells (44 scalar + 416 array), all within 1e-12, 244 bit-identical.
V0.2	2026-06-09	Harrison Zumbrun	Branding	Added the Morehouse and IndySoft logos (extracted from the workbook) as an inlined base64 letterhead in the app header, matching the spreadsheet's top-of-sheet placement.
V0.1	2026-06-09	Harrison Zumbrun	Engine, Risk Calc core	v1 scaffold: single-sourced engine on window.DRG (Cephes-grade normal CDF/inverse, incomplete-beta inverse, both bivariate-normal quadratures, the 11 reachable LAMBDAs, Excel-faithful ROUND/INT/CEILING/TEXT); Risk Calc & Curves scalar calculator wired with cell-keyed instrumentation.

Rev

Date

Author

Sections

Summary

V1.3

2026-06-10

Harrison Zumbrun

About; resources; guide banner

About This Tool: removed the EndInMind cross-reference and added an Explore Further section (Start With the End in Mind guide, the Quality Magazine decision-rules article, the free Force Calibration eBook, and the Morehouse technical blog). Added a site-wide banner in the toolbar linking the Start With the End in Mind guide. No calculation changes.

V1.2

2026-06-10

Harrison Zumbrun

UI; authorship

Moved the title and subtitle into the letterhead, centered between the Morehouse and IndySoft logos; the navigation is now a full-width green strip. Fixed table header/column misalignment on all results tables (the table class name collided with Tailwind's .grid utility, which set display:grid; display:table restored). Workbook authorship corrected: authors Henry Zumbrun (Morehouse Instrument Company) and Greg Cenker (IndySoft); web port by Harrison Zumbrun (Morehouse Instrument Company). No calculation changes.

V1.1

2026-06-10

Harrison Zumbrun

Risk Calc sheet complete; tab scenario gates

Ported the last Risk Calc regions: the specific-risk chart scaffolding (D2:E82 measurement-pdf grid, R/S product pdfs, TL/TU/NV/AL markers, P/Q TEXT labels) rendered as a new Specific Risk chart, and the UKAS LAB 48 worked-example table (exact strings incl. the workbook's verdict glyphs). Risk Calc Layer 2 now 957 cells, 593 bit-identical. Added Excel-COM alternate-input scenario gates for Method/Cost/Reliability (layer2_excel_scenarios_tabs.py): 11 scenarios, 2,588 cells, zero failures.

V1.0.1

2026-06-10

Harrison Zumbrun

Single-sourcing; verifier pass

Two-independent-verifier adversarial pass (both PASS; each tab's gate confirmed to catch a 1 % corruption). Routed the two UI Dobbert-M sites that match the engine formula through E.dobbertM() (bit-identical; Method 6 keeps its distinct clamped-TUR/no-max variant, documented). Tightened the check_engine.js smoke floor to 1e-12.

V1.0

2026-06-09

Harrison Zumbrun

IO layer; all tabs complete

Added the import/export/session layer: per-tab CSV export, CSV import for the Method reading grid and Cost equipment list, session JSON save/load, continuous localStorage autosave, and chart PNG export (graceful offline fallback). Validated by validators/test_io.py (12/12). All six workbook tabs now ported and exactly validated.

V0.9

2026-06-09

Harrison Zumbrun

JCGM 106 Validation

Ported the JCGM 106 Validation report tab: the resistor worked-example global risk (PFA_L/PFR_L vs exact analytic vs JCGM published), identity checks, specific-risk example, engine-vs-Risk-Calc identity, and the guard-band-width-vs-TUR risk-curve table. Layer 2 (layer2_jcgm.py): 81 cells, 70 numeric within 1e-12 + 11 string-exact, 69 bit-identical. The 40,000 Halton NORM.S.INV grid samples are covered by Layer 0.

V0.8

2026-06-09

Harrison Zumbrun

Guidance Summary

Ported the Guidance Summary reference tab: definitions, ISO 17025 requirements, the 7-method decision-rule comparison, the 3-outcome model, the sizing-rules table with the Required-U = Tolerance/4.6 formulas + custom-sizing input, the selection framework, and the five rules. Layer 2 (layer2_guidance.py): 6/6 sizing cells bit-exact.

V0.7

2026-06-09

Harrison Zumbrun

Reliability

Ported the Reliability tab: zero-failure sample size (CEILING(ln(1-conf)/ln(reliability))), Clopper-Pearson exact bounds (BETA.INV), observed EOPR / failure rate, additional-samples-needed (AdditionalSamplesNeeded), the zero-failure reference grid, the EOPR sensitivity grid, and the coverage-factor k table. Layer 2 (layer2_reliability.py): 74 cells, all within 1e-12, 65 bit-identical.

V0.6

2026-06-09

Harrison Zumbrun

Cost Model

Ported the Cost Model tab: production inputs, the 7-DMM equipment comparison (guard band sized by PFA_GB_SOLVER_ASYM to the PFA target, global PFA/PFR via PFA_L/PFR_L, ROUND-based false-reject/accept costs), total-cost analysis with budget-vs-baseline and payback, key-insight figures, and cumulative-cost projections. Layer 2 (layer2_cost.py): 192 cells, 173 numeric within 1e-12 + 19 string-exact, 161 bit-identical; the guard-band bisection solver reproduced exactly.

V0.5

2026-06-09

Harrison Zumbrun

Method 5 & 6 & TL

Ported the full Method 5 & 6 & TL tab: config + editable 10-point reading grid, per-reading Method 5 (RSS, NORM.INV acceptance) and Method 6 (Dobbert, TUR clamped to 4) tables with summaries, the M5-vs-M6 comparison, and the inlined Method-6 acceptance-limit diagram. Layer 2 (layer2_methods.py): 429 cells, 389 numeric within 1e-12 + 40 string-exact, 374 bit-identical.

V0.4

2026-06-09

Harrison Zumbrun

Risk Calc - methods/sensitivity/reference/bias

Added the 7-method Accept-± column, the PFA sensitivity-by-TUR table, the conformance→guard-band-multiplier reference table, and the bias-impact analysis block (all derived from the main inputs). Layer 2 extended to 592 cells, all within 1e-12, 353 bit-identical.

V0.3

2026-06-09

Harrison Zumbrun

Risk Calc - curves & charts

Wired the distribution chart-data arrays (A92:A171 product/measurement/biased pdfs) and the Global Risk Curves (PFA vs ITP at TUR 1:1-4.6:1 via PFA_SYMMETRIC); rendered both with Chart.js (graceful offline). Layer 2 extended to 460 cells (44 scalar + 416 array), all within 1e-12, 244 bit-identical.

V0.2

2026-06-09

Harrison Zumbrun

Branding

Added the Morehouse and IndySoft logos (extracted from the workbook) as an inlined base64 letterhead in the app header, matching the spreadsheet's top-of-sheet placement.

V0.1

2026-06-09

Harrison Zumbrun

Engine, Risk Calc core

v1 scaffold: single-sourced engine on window.DRG (Cephes-grade normal CDF/inverse, incomplete-beta inverse, both bivariate-normal quadratures, the 11 reachable LAMBDAs, Excel-faithful ROUND/INT/CEILING/TEXT); Risk Calc & Curves scalar calculator wired with cell-keyed instrumentation.

Web Port Notes

Item	Web app	Workbook	Agreement
Standard normal CDF	Hart/West cumulative normal	NORM.S.DIST / NORM.DIST	≤ 1e-12 (Excel is true-accurate; Layer 0)
Inverse normal	Acklam + Halley refinement	NORM.S.INV / NORM.INV	bit-identical on tested cells (Layer 0)
Inverse incomplete beta	Numerical-Recipes incbi	BETA.INV	bit-identical on tested cells (Layer 0)
Bivariate normal CDF	same Genz 6-pt & 5-term split LAMBDAs, ported op-for-op	BIVARIATE_NORMCDF / BIVAR	algorithm-identical

Item

Web app

Workbook

Agreement

Standard normal CDF

Hart/West cumulative normal

NORM.S.DIST / NORM.DIST

≤ 1e-12 (Excel is true-accurate; Layer 0)

Inverse normal

Acklam + Halley refinement

NORM.S.INV / NORM.INV

bit-identical on tested cells (Layer 0)

Inverse incomplete beta

Numerical-Recipes incbi

BETA.INV

bit-identical on tested cells (Layer 0)

Bivariate normal CDF

same Genz 6-pt & 5-term split LAMBDAs, ported op-for-op

BIVARIATE_NORMCDF / BIVAR

algorithm-identical