Eight oil and gas facilities — refineries, gas processing plants, offshore platforms, pipeline terminals, and drilling support yards — ran their spare parts inventory on eight separate ERP instances, each with its own reorder logic, its own criticality rules, and no visibility into what the other seven facilities were holding. The combined inventory sat at $180 million, and stockouts on critical parts were hitting 8% of the time — a number high enough that emergency freight and expedited orders had become a routine budget line rather than an exception. Twelve months of AI-driven inventory optimization across all eight sites released $12 million in working capital, bringing total inventory value down to $168 million, while the stockout rate on critical parts dropped under 2%, and the two numbers moving in the same direction at once is exactly the result a single-facility reorder-point system was never built to deliver, as a walkthrough with our team can show against your own facility footprint.
Case Study · Multi-Facility Spare Parts AI
$12M Working Capital Released Across 8 O&G Facilities Without a Single Stockout Tradeoff
Eight facilities, eight separate ERPs, one unified AI inventory layer. Total spare parts value dropped from $180M to $168M while the critical-parts stockout rate fell from 8% to under 2% — proving overstock and stockout are not opposite ends of a tradeoff when the forecast is built from the right data.
Before AI Optimization
Total Inventory Value$180M
Critical Stockout Rate8%
Reorder Logic8 separate ERPs
→
After 12 Months
Total Inventory Value$168M
Critical Stockout Rate<2%
Reorder Logic1 unified AI layer
The Starting Condition
Why Eight Facilities Were Simultaneously Overstocked and Understocked
This is a pattern that shows up almost anywhere a company operates more than a handful of sites on separate inventory systems, and it rarely gets caught by a single facility's own metrics, because each site's dashboard looked reasonable in isolation — it was only visible once all eight were compared against each other at the same time.
No Fleet-Wide Visibility
A critical bearing sat on the shelf at the gas processing plant while the offshore platform three hundred miles away placed an emergency order for the same part number, because neither site's ERP could see the other's stock, and no planner at either facility had a reason to think to check.
Reorder Points Set Per Site, Never Reconciled
Each facility's planner set thresholds based on that site's own history and gut feel, so identical equipment across two facilities carried wildly different safety stock for no operational reason.
Master Data Fragmented Across ERPs
The same physical part carried different part numbers, different descriptions, and different unit-of-measure conventions across the eight systems, making a simple question like "how much of this do we actually have" unanswerable without a manual reconciliation project.
Criticality Treated as Uniform
A control valve that stops production if it fails was stocked with the same safety margin logic as a gasket that causes an inconvenience, because the reorder formula had no way to weight consequence, only historical usage.
Facility-by-Facility Results
How the $12M Release Broke Down Across the Eight Sites
No two facilities released the same amount of capital or improved their stockout rate by the same margin — the AI layer optimized each site against its own criticality mix, lead-time exposure, and consumption pattern rather than applying a single fleet-wide target. A uniform percentage cut applied evenly across all eight sites would have looked simpler on a slide, but it would have ignored the fact that a pipeline terminal with predictable, low-variance demand carries a fundamentally different risk profile than an offshore platform dependent on scheduled supply vessels, and treating them identically would have either left too much capital tied up at the terminal or pushed the platform's stockout risk in the wrong direction.
| Facility | Inventory Before | Inventory After | Stockout Before | Stockout After |
|---|---|---|---|---|
| Gulf Coast Refinery | $42M | $39M | 9% | 2% |
| Permian Gathering Hub | $18M | $17M | 7% | 1% |
| Offshore Platform Complex | $31M | $29M | 11% | 3% |
| Gas Processing Plant | $22M | $20M | 8% | 2% |
| Pipeline Terminal Network | $15M | $14M | 6% | 1% |
| Midstream Storage Facility | $19M | $18M | 8% | 2% |
| Onshore Drilling Support Yard | $17M | $16M | 7% | 2% |
| Distribution & Blending Terminal | $16M | $15M | 6% | 1% |
Fleet Total: $180M → $168M · $12M Working Capital Released
The offshore platform complex carried the highest starting stockout rate of any facility in the fleet at 11%, driven by the combination of long supply-vessel lead times and a parts catalog that had grown organically over a decade of platform modifications without a corresponding review of what was actually still needed on board. It also saw the largest proportional stockout improvement, dropping to 3%, because criticality scoring caught a category of parts the platform's own planners had never had time to fully re-evaluate — spares tied to decommissioned or modified equipment still carrying the safety stock levels set when that equipment was originally installed. The Gulf Coast refinery, by contrast, started with a comparatively disciplined single-facility inventory program and still released $3M once fleet-wide visibility surfaced cross-facility transfer opportunities its own site-level optimization had no way to see on its own.
See This Broken Down Against Your Own Facility List
Every fleet has a different mix of criticality, lead times, and consumption history. A short session can show what a facility-by-facility breakdown looks like against your own ERP and CMMS data.
How the Reconciliation Actually Worked
Three Steps From Eight Siloed ERPs to One Unified Inventory Picture
None of the three steps below required replacing an ERP, migrating a facility to new software, or pausing procurement while the transition happened. Each facility kept running its existing system throughout, and the unification happened in a layer that read from all eight rather than asking any one site to change how it worked day to day.
1
Master data unification across all eight systems
Every part record from every facility's ERP was matched and normalized against a single master catalog, resolving duplicate part numbers, inconsistent descriptions, and mismatched units of measure without requiring a manual data-cleanse project before optimization could begin. This step alone is where most multi-facility inventory initiatives stall, because a data-cleanse project run by hand across eight separate systems and hundreds of thousands of line items can take longer than the optimization it's supposed to enable, and by the time it finishes the underlying consumption patterns have often shifted again. Running the matching algorithmically against existing data, rather than treating clean data as a prerequisite, is what let recommendations start generating within the first few weeks instead of after a multi-quarter cleanup phase.
2
Criticality scoring by consequence, not history
Every part was scored against the failure consequence of the equipment it serves — production-stopping, safety-related, or low-impact — so a control valve and a gasket stopped being treated by the same generic reorder formula. This is the step that let the fleet reduce total inventory value and improve stockout performance at the same time, because it stopped treating "reduce inventory" as a single lever applied uniformly and instead applied two different corrections in two different directions depending on what each part actually was.
3
Fleet-wide demand forecasting and rebalancing
Once the catalog was unified, the model could see that a part flagged as understocked at one facility was sitting excess at another, and recommend a transfer or a shared safety-stock strategy instead of two separate purchase orders for the same item. Over the twelve-month deployment, this became less about one-off transfers like the compressor seal example below and more about an ongoing rebalancing rhythm, where the fleet's total safety stock for shared part categories settled at a lower combined level than the sum of what eight independent facilities would have carried on their own.
Applied Example
The Compressor Seal That Was Sitting 340 Miles From Where It Was Needed
Six weeks into the deployment, the model flagged a dry gas seal cartridge for a critical compressor at the offshore platform complex as trending toward a stockout inside the next maintenance window — the platform's own consumption history showed no clear signal, since the part had only failed once in the facility's operating record. Cross-facility visibility showed a different picture: the gas processing plant onshore was carrying two of the same seal cartridges, both flagged by the new criticality scoring as excess relative to that site's own failure probability, sitting in a bin that hadn't been touched in fourteen months. Instead of the offshore platform placing an emergency order at expedited freight rates for a part with a multi-week lead time from the manufacturer, one of the two onshore units was transferred and staged ahead of the maintenance window at standard logistics cost. That single transfer captured in miniature what the fleet-wide optimization delivered at scale across all eight facilities — the part the plant needed already existed somewhere in the fleet, and the only thing missing had been a system that could see it. Before the unified catalog existed, this kind of match was theoretically possible but practically almost never happened, because finding it required a planner at one facility to somehow know to call a counterpart at another and ask about a part number that might not even be recorded the same way in both systems. The offshore platform's own planner had no visibility into onshore inventory, no standing process for checking, and no reason to suspect the part existed anywhere but a supplier's warehouse — which is exactly the condition that had been producing the fleet's emergency freight spend for years before the AI layer made the match automatically and flagged it as a recommended action rather than something a planner had to go looking for.
Fleet-Wide Metrics
Three Numbers That Moved Together, Not Against Each Other
Traditional inventory management treats stockout risk and carrying cost as a tradeoff — pushing one down usually pushes the other up. Across the full deployment, all three core metrics improved at the same time, which is the clearest evidence that the release came from correcting a mismatch between where capital sat and where risk actually lived, rather than from simply cutting stock levels and hoping for the best.
Emergency freight and expedited-order spend fell in direct proportion to the stockout rate, since the two are mechanically linked — a part that isn't on the shelf when it's needed almost always gets replaced through the most expensive available channel rather than the standard procurement process, and every stockout avoided is an expedited order that never gets placed.
How the Twelve Months Were Structured
A Four-Phase Rollout Across Eight Facilities
Phase 1
Data Unification
Master catalog built across all eight ERPs; duplicate and mismatched part records resolved without a manual cleanse project.
Phase 2
Criticality Scoring
Every part scored by failure consequence against the equipment it serves, replacing uniform safety-stock rules facility by facility.
Phase 3
Fleet-Wide Forecasting
Demand modeling ran across the unified catalog, surfacing cross-facility rebalancing opportunities the siloed ERPs could never see.
Phase 4
Continuous Rebalancing
Recommendations moved from advisory review to routine planner workflow as trust built, sustaining the release through the full twelve months rather than a one-time cleanup that would have drifted back toward the old baseline within a year.
Common Questions
Multi-Facility Spare Parts AI — Frequently Asked
These are the questions supply chain and reliability leadership ask most often before extending AI inventory optimization across more than one facility.
Do all eight facilities need to run the same ERP for this to work?
No — this deployment unified inventory data across eight separate ERP instances without requiring any facility to migrate or replace its existing system, since the AI layer connects to each source system and reconciles the catalog centrally rather than forcing a single platform standard, which is what made it possible to start the deployment without a multi-year platform consolidation project as a prerequisite. Book a demo to see how this connects to the specific systems each of your facilities runs today.
How long does master data cleanup take before optimization can start?
The unification step works on the data as it currently exists in each ERP rather than waiting for a separate data-cleanse project to finish first, so recommendations can start generating within weeks while data quality continues improving as more consumption history accumulates. Contact support to talk through what your current data quality supports as a starting point.
Does reducing inventory value increase the risk of a stockout?
Not when the reduction is criticality-weighted — this deployment cut total inventory value by $12M while the critical-parts stockout rate fell from 8% to under 2%, because the capital released came from excess non-critical stock, not from the safety margin on parts that actually stop production. Book a session to review how criticality scoring is applied to your own parts catalog.
Can parts actually be transferred between facilities, or is this just a reporting layer?
The model surfaces specific transfer recommendations, like the compressor seal example above, when a part is excess at one facility and trending toward a stockout at another, and a planner reviews and approves each recommended transfer before it moves, keeping a human decision in the loop on every physical movement. Ask our team about how transfer recommendations fit into your existing logistics process.
How is this different from a fleet-wide inventory dashboard?
A dashboard shows what each facility currently holds; this platform recommends what each facility should hold, continuously recalculating reorder points and transfer opportunities from unified criticality and consumption data rather than leaving a planner to manually interpret a report across eight separate views. Book a call to see the difference applied to your own fleet.
Find the Capital Trapped Across Your Own Facility Fleet
iFactory unifies spare parts data across every facility's ERP, scores every part by real failure consequence, and forecasts demand fleet-wide — releasing working capital while cutting critical stockouts, not trading one for the other, and without requiring any facility to change the system it already runs.
