For many CIOs in manufacturing, physical inventory remains one of the most persistent and least modernized risks to their enterprises. It is operationally critical, financially material and foundational to trust across ERP (enterprise resource planning), SCP (supply chain planning) and MES (manufacturing execution systems). Yet it is still managed using assumptions that no longer reflect how factories actually operate.
Despite decades of ERP investment, most manufacturers continue to rely on inventory practices that assume static systems, frozen transactions and post-count manual reconciliation. In live, high-mix manufacturing environments, where production runs continuously, configurations change frequently and execution data moves in real time, those assumptions no longer hold true.
The result is a pattern I have seen repeatedly across industries: even after physical counting is officially completed, factories remain shut down for extended hours. Inventory reconciliation stalls because finance and operations leadership often the CFO or VP of Operations are unwilling to release systems due to unresolved discrepancies and “missing” counts. What was intended to be a controlled inventory exercise turns into prolonged operational downtime, driven not by counting itself but by the absence of a deterministic, continuous system-level reconciliation mechanism.
What I observed across manufacturing environments
Early in my career, I assumed inventory inaccuracies were primarily execution issues—training gaps, missed system transactions or incomplete procedures. Over time, after working across multiple plants and industries, including aviation, automotive and warehouse automation, I realized those assumptions were incorrect. In most cases, operators were sufficiently trained, processes were defined and intent was not the problem. The failure was structural: the underlying system architecture consistently relied on perfect human execution to maintain accuracy in environments that were inherently dynamic.
What I observed instead was consistent across organizations:
- ERP systems record transactions accurately, but only when humans execute them perfectly.
- MES systems reflect production reality, but not always financial truth.
- Inventory accuracy depends on slowing or stopping production rather than continuous verification.
- Work-in-process (WIP) and job material remain largely invisible at the enterprise level.
These issues manifest in different ways, but the root cause is the same. Enterprise systems designed decades ago were optimized for batch processing, periodic controls and human-mediated reconciliation. They were never architected for live, continuous manufacturing execution
The question CIOs/CFOs are always asked and why it misses the point
Almost every inventory discussion begins with the same operational debate:
Should we rely on full physical inventory or cycle counting?
Cycle counting is attractive because it minimizes disruption. In theory, it allows organizations to maintain inventory control without shutting down production. In practice, I have seen it introduce a subtle but dangerous failure mode: system inventory gradually drifts away from physical reality.
The problem is not that cycle counting is ineffective. The problem is that it operates at the wrong layer. It assumes the underlying inventory model is valid and attempts to correct discrepancies after they occur.
For CIOs, that distinction matters. You can count more often, apply analytics and tune thresholds, but if the system model itself cannot reconcile live execution deterministically, accuracy will always degrade over time and requires a full physical inventory at some time.
A scenario that a CIO/VP of operations will recognize immediately
Consider a simple scenario I have encountered many times.
A factory starts with 100 units of a part, both physically and in the system:
- Stockroom: 50
- WIP: 25
- QA: 25
An operator physically moves 10 units from the stockroom to WIP, but the corresponding system transaction is never executed or is delayed due to interface, recognition or batch-processing latency, leaving the ERP record temporarily or permanently out of sync with physical reality.
Physical reality becomes:
- Stockroom: 40
- WIP: 35
- QA: 25
The system still shows:
- Stockroom: 50
- WIP: 25
- QA: 25
Later, assuming the cycle count program runs and selects the WIP location, which is typical because most ERP cycle counting freezes a specific location, not the entire factory. The operator counts 35 units and posts a +10 adjustment.
The system now shows 110 units, even though only 100 physically exist unless all locations are counted at the same time.
Unless cycle counting is executed across all locations simultaneously, which effectively requires production disruption, this discrepancy compounds over time. I have seen organizations unknowingly inflate or deflate inventory simply by trying to improve accuracy.
Why WIP is the hardest problem and the biggest blind spot
If finished goods and raw materials are challenging, WIP is exponentially harder.
Modern manufacturing depends on:
- Multi-level bills of material
- Phantom assemblies
- Operation-level consumption
- Backflushed and non-backflushed components
As a result, production lines contain material that does not exist as clean, countable part numbers in ERP systems. Traditional physical inventory cannot capture this reality accurately. Cycle counting also fails because it treats WIP as static inventory when it is constantly changing state.
In many organizations, WIP becomes an accepted blind spot, estimated rather than measured. From a financial, audit and planning perspective, that gap is no longer acceptable.
The insight that changed my approach
After seeing the same failures repeatedly, I reached a clear conclusion:
Inventory accuracy was being treated as an operational event when it should have been treated as a system control function.
Enterprise ERP and MES platforms excel at recording transactions, but they were never architected to perform continuous physical verification during live execution. This gap is consistent with findings from my published research on ResearchGate, which shows that while ERP systems improve inventory visibility, real-time accuracy and reconciliation break down without continuous control logic and integrated execution-level validation
The research highlights that traditional ERP inventory models rely on periodic reconciliation mechanisms, cycle counts, physical inventories and post-facto adjustments, rather than embedded, always-on verification. Even when augmented with scanning or MES integration, accuracy remains reactive unless physical validation is treated as a system-level control, not a periodic activity.
For CIOs and VPs of Operations, this creates systemic enterprise risk:
- Accuracy depends on production slowdowns or freezes
- ERP, MES and financial systems drift out of alignment
- WIP and job-level material remain opaque
- Audit confidence relies on manual reconciliation
These are not shop-floor execution problems. They are enterprise data integrity and system architecture problems, validated both by industry practice and by applied research into ERP, IoT and warehouse execution systems.
Why time-based inventory was a step forward, but not enough
The industry has tried to address this gap. One example is the time-based physical inventory approach disclosed in US20080255968A1, which proposed reconciling inventory using timestamps rather than freezing transactions.
This approach was directionally correct. It acknowledged a reality every CIO understands: Manufacturing does not slow down simply because accounting needs certainty.
But in practice, time-based reconciliation alone does not solve the problem. It focuses on when inventory moved, not what that inventory represents in execution context. Secondly, it was still not implemented in most industries, as well as ERP software, due to the complexity needed.
Inventory accuracy is not just a time problem.
It is a context problem.
The architecture I designed and later filed for patent
The question that ultimately changed my thinking was simple:
How can physical inventory be reconciled deterministically against everything the system already knows, without stopping production and without losing execution context?
That question led me to design an ERP-agnostic execution architecture, later formalized in my patent filings, including:
- UK Patent Application GB2521650.8 – AI-enabled real-time physical inventory and automated job material reconciliation system
- UK Patent Application GB2521655.7 – Predictive horizon-based safety stock optimization system and method
The architecture rests on four core principles.
1. Snapshot-based verification without stopping production
Instead of freezing transactions, the system captures a reference inventory snapshot at a known time. All ERP- and MES-recorded movements – issues, receipts, transfers and completions are tracked continuously from that point.
When a physical count is captured later, the system mathematically reconciles it against those movements.
Inventory becomes a calculable system state, not an operational disruption.
2. Single-screen inventory capture for all inventory states
Traditional inventory processes force operators to choose between multiple ERP screens depending on inventory type. In live manufacturing, this fragmentation creates immediate errors.
I designed a single, unified inventory capture interface that allows operators to count any inventory they physically see, including WIP, without pre-classification. Operators enter part number, quantity, location or line and optionally job or operation.
The system, not the operator, determines inventory meaning based on execution context.
3. BOM explosion to reconcile what cannot be counted
Counting WIP visually is only half the problem. The harder challenge is reconciling what that WIP implies in terms of component consumption.
By dynamically exploding the Bill of Materials during reconciliation, the system:
- Determines expected consumption
- Compares it to actual issues, scrap and returns
- Computes deterministic variances
- Generates system-ready adjustments automatically
This replaces estimation with computation.
4. Deterministic logic with controlled AI augmentation
Core reconciliation remains deterministic and auditable. AI is applied after reconciliation to detect anomalies, identify confidence gaps and prioritize verification—not to replace core logic.
This preserves compliance while enabling intelligence.
What changed when inventory became a system capability
When inventory verification shifted from an operational event to a system function, the impact was measurable:
- Production shutdowns for inventory were significantly reduced
- ERP, MES and finance data stayed aligned
- Close cycles shortened
- Audit effort decreased
- Trust in inventory data increased
Inventory stopped being a periodic crisis. It became a controlled digital asset.
A CIO’s takeaway
Manufacturing is becoming more automated, distributed and real-time. Robotics, IoT and AI increase execution speed, but they also amplify the cost of weak data foundations.
Inventory accuracy cannot remain dependent on periodic freezes in an always-on environment.
For CIOs, the choice is clear:
- Continue layering controls on legacy inventory models
- Or redesign inventory as a continuous execution control system
One approach detects errors after they accumulate.
The other prevents them from accumulating at all.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?
Read More from This Article: Why your ERP still can’t solve inventory drift — and the architecture that will
Source: News