Spare Parts Planning — Overview
Most tools forecast what you used. Sparely prepares you for what's coming — and proves it.
My work
Will the parts for my planned jobs be there, and which critical parts are unprotected?
What do I reorder today, how much, and which Min–Max settings are wrong?
What must I buy, from whom, by when — and which open POs must I chase?
How much capital is tied up, how much risk do we carry, and what are the decisions worth?
Process
From SAP extracts to an approved purchase requisition
- 1Baseline data
6 files: 7,481 parts, 14,484 FuncLocs, 16,655 issues, 5,795 notifications, 10,153 plan items, 6,189 PO lines
- 2Clean & link
25 data-quality checks (duplicates, units, currencies, cycles); parts linked to equipment, failures, purchases and the maintenance plan
- 3Forecast & optimize
Hybrid demand forecast (bias 10% vs 115% static), ML early warning (AUC 0.89), cost–risk optimizer calibrated to 99 emergencies / yr
- 4Role workspaces
Maintenance requests job parts; the Planner uses identical stock first, then approves reorders and Min–Max; Procurement gets a PR draft
- 5Proven impact
Replaying May–Sep 2026: stockouts 952 → 209, demand days served 69% → 93%, inventory held -54%
Case requirements coverage
Each expected output of the brief, where it is in the system, and the live evidence
| Expected output / key question | Where | Evidence (live) |
|---|---|---|
| Demand forecasting (history, failures, maintenance, shutdowns, installed base, abnormal use) | Demand Forecast | 826 planned part lines in 12 months; turnaround split out; abnormal issues capped |
| Dynamic Min–Max, safety stock, reorder point, replenishment qty | Planner workspace | 1,160 reorders; Min–Max recomputed per part on every request |
| Inventory health: excess, slow-moving, non-moving, potentially obsolete | Inventory Health | 1,848 non-moving · 1,161 potentially obsolete |
| Reorder / maintain / reduce / monitor, balancing cost vs stockout & downtime risk | Material List | 1,160 reorder · 3,505 reduce · 1,494 monitor · 1,258 maintain |
| Stockout risk & overstock risk | Material List | 138 critical parts ≥ 50% stockout risk · excess Rp 573.3B at PO prices |
| Slow-moving alert | Inventory Health | 730 parts not issued for > 180 days |
| Cost–risk optimization scenarios (raise / lower inventory) | Cost–Risk Scenarios | Presets + live re-optimization of all parts + efficient frontier + replay under the scenario |
| Shutdown and turnaround activities | Maintenance | Last turnaround found in the data (360 parts); readiness for the next one with order-by dates |
| Right part available at the right time (the core problem) | Inventory Health | 77 needs covered by the same item held under another material number · Rp 2.2B not bought |
| Equipment and spare part criticality, incl. process safety | Material List | 860 safety-critical parts (safety and shutdown valves) costed as class A and never left without a spare · 544 long-lead critical parts flagged |
| Governed decisions | Management | 22 buys and 985 reductions need Management sign-off; every decision is logged |
| Evidence that the policy works | Management | Replay of 3,035 real demand days: 878 → 127 stockouts on critical parts |
| Key Q1 — predictability of irregular demand | Demand Forecast | Demand split by cause, installed-base pooling, ML early warning, out-of-time backtests |
| Key Q2 — balance inventory cost, availability, operational risk | Cost–Risk Scenarios | Per-part service level chosen by minimum holding + downtime-risk cost |