Critical Spares Engine
A quantitative + qualitative work engine for Program Managers and Supply-Chain leads responsible for data-center M&E spare-parts readiness. Built for hyperscale operators, facilities teams, and sourcing professionals who need rigorous, transparent math — not spreadsheet guesswork. 20 modules: 9 quantitative analytical engines + 10 operating-engine generators + 1 methodology reference.
ⓘ How criticality is computed
Simplified FMECA Criticality Number: Cm = (S_eff/10) × (D/10) × λ × N × (oh/8760) where S_eff is severity reduced by redundancy buffer, D is detectability, λ is annual failure rate, N is installed base, and oh/8760 normalises to fraction of year (annual basis → oh=8760 → factor=1).
Risk Priority Number (RPN): RPN = S_eff × D × (λ×10) — analogous to FMEA RPN = Severity × Occurrence × Detection.
Tier thresholds: VITAL Cm ≥ 0.5; ESSENTIAL 0.1–0.5; DESIRABLE <0.1. Decision: VITAL → STOCK + DUAL-SOURCE; ESSENTIAL → STOCK; DESIRABLE → DON'T STOCK (review).
Ref: MIL-STD-1629A FMECA; Quality-One FMECA guide.
ⓘ How readiness is computed
Readiness% = min(100, confirmedSupply / required × 100)
confirmedSupply = onHand + (PO qty if commit date ≤ need date AND commit is confirmed). If no confirmed commit, PO qty counts as 50% confidence. Status: RED if <80% or commit after need; YELLOW if 80–99%; GREEN if ≥100% with slack ≥7 days.
ⓘ Newsvendor + Fill-Rate methodology — units & formulas
Unit conversion: All inputs use annual units (μ in units/yr, σD in units/yr, lead time L in weeks). Internally: L_yr = L_wk / 52; σ_L_yr = σ_L_wk / 52.
Demand during lead time: μ_LT = μ_annual × (L/52) | σ_LT = √((L/52)×σ_D² + μ_annual²×(σ_L/52)²) — combines demand variability and lead-time variability.
Critical ratio: CR = Cu / (Cu + Co) where Cu = under-stock cost per stockout event; Co = carrying cost over part life = carryRate × unitCost × partLife. For critical DC spares Cu ≫ Co, pushing CR close to 1 and Q* higher.
Newsvendor Q*: Q* = max(0, ⌈μ_LT + Φ⁻¹(CR) × σ_LT⌉) where Φ⁻¹ is the inverse normal CDF (Beasley-Springer-Moro rational approximation, verified: Φ⁻¹(0.975)=1.96, Φ⁻¹(0.99)=2.326).
Safety stock (fill-rate target): SS = max(0, ⌈Φ⁻¹(FR) × σ_LT⌉). Reorder Point: ROP = ⌈μ_LT + SS⌉.
Poisson mode (slow movers, demand <5 units/yr): λ_LT = λ_annual × (L/52); P(stockout at level S) = 1 − Σ_{k=0}^{S−1} e^{−λ_LT} λ_LT^k/k!. Falls back to normal approx for λ_LT > 200. Ref: Sherbrooke 1985 METRIC — newsvendor model (critical-fractile).
ⓘ Simplified 2-echelon MEIO / VARI-METRIC approach
This is a transparent heuristic approximation (not a full VARI-METRIC solver). Hub units are sized to cover demand during hub-to-site lead time at the target fill rate. Site-level stock covers demand during the OEM lead time minus hub coverage. Remaining budget fills the central depot.
Fleet readiness is estimated as min(100, totalStock / (demandDuringMaxLT × safetyFactor)). Hub delta = readiness with hub minus readiness without hub.
Ref: METRIC/VARI-METRIC review (UMD); MEIO framework (Umbrex).
ⓘ Composite risk score weights
Weights: Financial Health 15%, Single-Source 20%, Geographic Concentration 12%, Lead-Time Volatility 15%, OTIF 18%, Capacity Headroom 10%, Geopolitical 10%. Contract status adjusts ±5 pts. Score 0–100. Bands: <30 LOW; 30–59 MEDIUM; 60–79 HIGH; ≥80 CRITICAL.
Kraljic quadrant: supply-risk dimension = composite score / 10; spend-impact from direct input. Ref: Kraljic, P. (1983) "Purchasing must become supply management", HBR.
ⓘ LTB / DMSMS methodology — formulas & NPV interpretation
LTB quantity: LTB_Q = max(0, ⌈annualDemand × supportYears × 1.15 − onHand − openPO⌉). Safety factor = 1.15 (15% buffer for demand uncertainty + scrap risk).
NPV Option A (LTB Stock): t=0: −LTB_Q × unitCost (buy upfront). t=1…n: −carryingCost_t (average remaining inventory × carryRate × unitCost). End of life: ±scrap value (unused units × unitCost × (1 − scrapRisk) − unused × unitCost × scrapRisk).
NPV Option B (Requalify): t=0: −altQualCost. t=1…n: −demandYr × unitCost × 1.20 (spot premium during qual period) or × 0.90 (qualified alternate at discount) after qualification completes.
Decision rule: Both NPVs are costs (negative). Higher NPV = less negative = lower total cost = better option. If NPV_B > NPV_A → Requalify is cheaper. Ref: DCF: NPV = Σ CF_t / (1+r)^t.
EOL Exposure Score: (installedUnits × criticality × supportYrsRemaining) / max(1, qualifiedAlternates). Bands: <50 LOW; 50–149 MEDIUM; 150–299 HIGH; ≥300 CRITICAL.
ⓘ How priority logic works
Status: RED if any critical shortage or >3 late POs; YELLOW if 1–3 late POs or unconfirmed suppliers; GREEN otherwise. Priority: critical shortages → P1; late POs >3 or supply severity ≥4 → P2 for supply workstream; finance/exec ask → P3. Each Follow-Up row is generated from your inputs with a recommended message and consequence. The EOD draft is a status email skeleton populated from your inputs.
ⓘ RAG logic and cadence derivation
Each metric is rated GREEN (meets target), YELLOW (within 10% of target), or RED (fails). Overall RAG = worst of the 4 "critical" dimensions (OTIF, Commit Accuracy, Responsiveness, CA Closure). Review cadence: Critical supplier with any RED → Weekly Operational Review; Preferred supplier or any YELLOW → Monthly Business Review; Tactical/Replaceable with all GREEN → Quarterly Executive Business Review.
ⓘ Leverage & BATNA logic
Leverage strength is auto-derived: volume commitment is Strong if annual spend >$500K; dual-source threat is Strong if alternates >0, Weak if 0; multi-year is Moderate; forecast visibility Moderate; standardisation Moderate. BATNA: 0 alternates + critical/preferred → "Limited BATNA — pivot to non-price terms"; ≥2 alternates + tactical/replaceable → "Credible BATNA — credibly threaten competitive bid". Counterproposal is templated per scenario type and raw-material justification.
ⓘ How requirements are built
All 14 standard MSA/SOW areas (Scope, Pricing, Lead Time, Forecast, Capacity, Delivery/Incoterms, Warranty, Quality, Documentation, EOL Notice, Last-Time-Buy, Change Notice, SLA, Inventory, Termination) are always shown. Rows where you toggled a requirement are highlighted. The "Proposed Contract Language Concept" column shows the plain-English clause intent — a brief to hand to legal, not a legal clause.
ⓘ How the proposal is built
The Problem Statement reframes your text with frequency and annualised impact (frequency multiplied per year). Future-State Process is a step-list (intake → triage → assignment → execution → control → review) tailored to ticked root causes. RACI maps the selected stakeholders to each step. Controls & KPIs are generated from ticked root causes — e.g., "no SLA" → SLA control + SLA-adherence KPI. The 30/60/90 plan is generated from a standard PM-ops template with milestones adapted to the problem.
ⓘ How agendas are generated
The prep brief agenda is auto-generated from the meeting type using the canonical review cadence templates from the master sourcing engine: Supplier Operational Review → PO review, expedite, shortage, commit validation, quality, recovery; Monthly Business Review → KPI trend, cost, lead time, capacity, improvement, contract, upcoming demand; Quarterly Executive Review → strategic alignment, supply roadmap, commercial partnership, long-term capacity, risk & resiliency. The notes template is a live editable structured form — click "Add row" to expand any table.
ⓘ How the plan is generated
Stakeholder map attributes (What They Care About, Communication Style) are drawn from the master PM sourcing engine Module 9 stakeholder registry. Channel and cadence are auto-derived from urgency: Critical/Urgent → phone call + written follow-up same day; Elevated → email + meeting within 48h; Routine → email/async. Message drafts are register-adjusted: executive = concise + decision-oriented; supplier = specific ask + written commitment + deadline; finance = numbers + scenario; legal = risk framing + clause intent.
ⓘ How EOL quantities are computed
LTB quantity: LTB_Q = ceil(installedUnits × failureRate × supportYears × 1.20 − onHand − openPO). Safety multiplier 1.20 covers demand uncertainty and scrap risk. Fleet exposure score: EOL_Score = (installedUnits × criticality × supportYears) / max(1, qualAlts).
Options viability: "Qualify Alternate" shown as viable if alternates > 0 or alt-qual lead time < support years. "LTB Stock" always viable as fallback. "Redesign" viable only if redesign input = Yes. "Refurb / Harvest Pool" viable if criticality ≤ 6 or installed base is large. "Do Nothing" viable only if criticality < 4.
For full NPV comparison of LTB vs Requalify, open the Last-Time-Buy tab (Tab 6) with the same inputs.
ⓘ How interpretations are generated
The solver scans the ask for supply-chain signal words (readiness, inventory, supplier, cost, lead time, EOL, forecast, risk, process, visibility, alternate, expedite, shortage) and maps each to a candidate interpretation from the master PM sourcing engine Module 14 framework. The sharpened problem statement uses the SMART structure: what to improve, from what baseline, to what target, across what scope, by when. The 30/60/90-day plan follows the Discover → Stabilise → Systematise pattern standard for hyperscale PM roles.
ⓘ How the story is built
The builder combines your inputs with competency-specific narrative scaffolding for the selected PM dimension (e.g. Supplier Negotiation → frame the constraint, name your leverage, show the process you built). The register targets hyperscale program management interviews: structured → leads with the headline result, names cross-functional stakeholders, shows the system built (not just the firefight), ends with scale/lesson. Coaching notes are generated per competency based on common interviewer scoring rubrics.
Browse the seed catalog of data-center M&E spare parts — legacy raised-floor through AI-factory liquid-cooled. Loading catalog… (Full DB scales to 100k+ rows — see
data/spares-parts.sqlite.)
| Part ID | Description | OEM | System / Sub | DC Gen | Crit | MTBF (yr) | LT typ (wk) | Cost (typ $) | Lifecycle | EOL risk | #Alts | Use |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading catalog data… | ||||||||||||
| OEM Name | HQ | Market Position | Fin. Health | Typ. Lead (wk) | Typ. OTIF (%) | Single-Source Risk | Contract Models |
|---|---|---|---|---|---|---|---|
| Loading OEM data… | |||||||
| Framework | What it does | Module | Citation |
|---|---|---|---|
| FMECA (MIL-STD-1629A) | Failure Modes, Effects & Criticality Analysis — computes Criticality Number Cm = β · α · λp · t; assigns Category I/II effect severity → stock decision | 1 · Criticality | Quality-One FMECA guide |
| RCM Criticality Ranking | Reliability-Centered Maintenance — ranks assets by consequence × likelihood × detectability → VITAL / ESSENTIAL / DESIRABLE tiers | 1 · Criticality | rgbwaves.com |
| VED Analysis | Vital / Essential / Desirable — rapid 3-bucket criticality sort for early triage before full FMECA data is available | 1 · Criticality | Supply-chain standard; no single canonical reference |
| ABC-XYZ Matrix | ABC = value/usage Pareto (A=top 70% cost); XYZ = demand variability (X stable, Z erratic) → 9-cell stocking policy matrix | 3 · Optimal Stock | Standard inventory management practice |
| Newsvendor / Critical-Fractile | Finds optimal stock Q* where CR = Cu/(Cu+Co) = P(D ≤ Q*). For DC spares Cu ≫ Co → stock generously. | 3 · Optimal Stock | Sherbrooke 1985 (INFORMS) |
| Fill-Rate Safety Stock | SS = z(FR) × σLT where σLT = √(L·σD² + μD²·σL²) — combines demand and lead-time variability | 3 · Optimal Stock | Silver, Pyke & Thomas: Inventory and Production Management |
| Poisson / Compound-Poisson | Demand model for slow-moving critical spares (λ = installed base × AFR). P(stockout at S) from Poisson CDF — more accurate than Normal for <5 units/yr demand | 3 · Optimal Stock | Sherbrooke 1985 |
| METRIC (Sherbrooke 1968) | Multi-Echelon Technique for Recoverable Item Control — minimises expected backorders across sites + depot for a given budget | 4 · Hub Positioning | UMD DRUM review |
| VARI-METRIC (Slay 1984) | Adds variance correction to METRIC → within 1% of optimal vs ~11% for base METRIC; supports multi-echelon repairable item planning at scale | 4 · Hub Positioning | Scialert ITJ 2014 |
| MEIO | Multi-Echelon Inventory Optimisation — generalises METRIC/VARI-METRIC to full network: multiple products, echelons, fill-rate and budget constraints simultaneously | 4 · Hub Positioning | Umbrex MEIO |
| Supplier Risk Index | Composite 0–100 score: Financial 15%, Single-Source 20%, Geo Concentration 12%, LT Volatility 15%, OTIF 18%, Capacity 10%, Geopolitical 10%. Bands: LOW/MEDIUM/HIGH/CRITICAL | 5 · Supplier Risk | Adapted from procurement risk literature; weights configurable |
| Kraljic Matrix | Supply risk × profit/spend impact → Strategic / Bottleneck / Leverage / Non-Critical quadrants → per-quadrant sourcing strategy | 5, 7 · Supplier Risk, Kraljic | Kraljic, P. (1983) "Purchasing must become supply management", HBR |
| DMSMS | Diminishing Manufacturing Sources & Material Shortages — monitors lifecycle: Active → NRND → Last-Time-Buy → Obsolete; triggers proactive EOL action | 6 · Last-Time-Buy | Wikipedia DMSMS |
| Last-Time-Buy (LTB) | LTB_Q = max(0, ⌈annualDemand × supportYrs × 1.15 − onHand − openPO⌉). NPV comparison: Option A (buy-stock) vs Option B (requalify alternate) via DCF | 6 · Last-Time-Buy | Lifetime buy estimations (UMD) |
| Monte-Carlo Simulation | 1,000+ scenarios with variable lead time, demand, supplier reliability → P(stockout), P10/P50/P90 readiness, expected downtime cost, tornado chart of variance drivers | 8 · Monte-Carlo | Box-Muller normal variate; standard Monte-Carlo methodology |