Inventory Management Is Purchasing Governance
Inventory problems get treated as warehouse problems and are almost always buying problems. The stock sitting on a shelf is the visible residue of decisions taken weeks earlier by people who never saw the shelf. Governing those decisions is what changes the balance.
The commitment is made at the specification
A purchase order is the moment the money becomes visible and not the moment it was committed. The real commitment happened when somebody chose a part, wrote a specification, or approved a design that only one supplier can satisfy.
Everything downstream of that choice is administration. The buyer negotiating the order has almost no room to move. Cost, lead time, and availability were all determined by an engineer or a product manager who was not thinking about any of the three.
Move the review upstream if the intent is to change outcomes. A specification review that asks whether a common part would serve does more for inventory than any amount of attention applied at the order stage.
Approval authority is per transaction and exposure accumulates
Approval limits are almost always written against a single order. Anyone below a threshold can commit, and each individual commitment is defensible on its own terms.
Nobody holds the total. The organization can therefore accumulate a large open commitment through a series of small approved decisions, and no control anywhere in the process was violated. The exposure is real and the process report is clean.
Report open commitment rather than spend. Spend describes what has already been invoiced, while open commitment describes what the business has promised to pay and cannot easily withdraw from. The second number is the one that constrains next quarter.
Carrying cost is understated wherever nobody calculates it
The cost of holding stock is routinely estimated at whatever figure the finance team last mentioned. The real components are specific and they add up faster than the estimate suggests.
Capital tied up is the obvious one. Storage, handling, insurance, and shrinkage follow. Obsolescence sits underneath all of them and is usually the largest. It is the only component able to consume the entire value of an item rather than a percentage of it.
Calculate the figure once for the business rather than borrowing a rule of thumb. Every stocking decision downstream depends on it, and a wrong input produces confidently wrong answers at scale.
Obsolescence is a deferred decision rather than an incurred cost
Dead stock sits on the balance sheet at cost until somebody books the write-off. Nobody wants to book it, because the write-off makes visible a decision that was already wrong and identifies the period rather than the person.
So the stock stays. It occupies space, absorbs counting effort, and misrepresents the asset position to everyone reading the accounts, including the people making the next stocking decision.
Set an aging rule and apply it without argument. Anything past a stated threshold gets a decision to sell, use, or dispose, and the decision is scheduled rather than triggered by discomfort. Composure at that moment is worth more than optimism about a future order.
Two axes matter and most companies use one
Classifying items by annual value is standard practice and only half the picture. Value tells you how much money an error costs. It says nothing about how predictable the item is.
Adding a second axis for demand variability changes the policy substantially. A high value item with steady predictable demand needs tight ordering and very little buffer. A low value item with erratic demand needs a generous buffer and almost no attention.
Managing both on value alone produces the opposite of both. The expensive predictable item accumulates buffer it never needed, and the cheap erratic one runs out and stops the work.
Safety stock is arithmetic rather than temperament
Buffers get set by whoever was most recently embarrassed by a stockout. That method produces buffers correlated with memory rather than with variability.
Safety stock is a function of two things. How much demand varies during the replenishment window, and how much the replenishment window itself varies. Where lead time is unreliable, the second term dominates and no amount of demand forecasting compensates for it.
The arithmetic is not difficult and the inputs are already in the systems. Applying it consistently across a few hundred items releases capital and improves availability at the same time, which is the combination that convinces a sceptical finance team.
Record accuracy is the precondition for every rule above
None of this works on records that do not match the shelf. Where the system shows stock that is not there, the reorder logic fires late and the operation runs short while reporting healthy coverage.
Annual counting does not fix this. It corrects the ledger once a year and leaves eleven months during which every automated decision runs on drifting data.
Cycle counting fixes it as an operating property rather than an event. A small number of items counted every week, weighted toward the ones that matter, keeps accuracy high continuously and surfaces the process defect causing the drift.
Order quantity models optimize a variable that stopped mattering
The classic order quantity model balances the cost of placing an order against the cost of holding what it delivers. It assumes steady demand and a fixed ordering cost, and it was built when placing an order carried real administrative expense.
Ordering is now nearly free in most businesses, which collapses one side of the balance. The model consequently recommends quantities larger than the situation warrants and pushes the operation back toward the batching that generates its own problems.
Treat it as a teaching device rather than a policy. The useful residue is the idea that order size is a tradeoff, and the tradeoff should be re-derived with current numbers rather than inherited.
Naming the models and what each one assumes
Three models carry most of the analysis, and each is safe only where its assumption holds. Applying any framework outside the conditions it was built for is the most common way a sound methodology produces an unsound policy.
Classification models assume the population is heterogeneous enough that differentiated policy pays, which is nearly always true above a few hundred items. Reorder point models assume lead time is known within a range, which is where a supplier with wide variance breaks them.
Activity-based costing supplies the discipline for the specification review, by attributing the handling, counting, and expediting effort a proliferating part number actually consumes. Without that evidence, adding a variant looks free at the moment it is proposed.
Where the stock is held changes who pays to hold it
Ownership of inventory is negotiable and rarely negotiated. Consignment leaves title with the supplier until the item is consumed, which moves the capital cost without moving the availability.
Vendor managed replenishment goes further by handing the supplier responsibility for maintaining agreed levels against shared consumption data. Both arrangements suit items where the supplier has better visibility of the pattern than the buyer does.
Neither is free. The supplier prices the carrying cost into the unit, so the question is whether they can carry it more cheaply than the business can. That depends on whether they pool the same item across several customers, and where they do the transfer is genuinely cheaper for both parties.
Coverage is a better instrument than value
Total inventory value is the number that reaches the board and the least useful one available. It moves with price, with mix, and with the calendar, so a change in the total explains nothing about whether the position improved.
Days of coverage by item class is the instrument that answers the question. It expresses stock in the only unit that matters operationally, which is how long the business can continue serving demand from what it currently holds.
Tracking coverage also exposes the two failure states separately. A rising total with falling coverage means the money moved into items nobody wants, and that combination is invisible in any valuation report.
Conditional rules for the buy decision
Where a part exists in several near-identical variants, the inventory problem is a specification problem. Consolidating variants releases more capital than tightening any ordering rule.
Where availability is poor while total stock is high, the mix is wrong rather than the quantity. Buying more will worsen both conditions simultaneously.
Where an item has not moved in a defined period, stop reordering it automatically and force a decision. Automatic replenishment against dead demand is the most reliable way to convert a small write-off into a large one.
Where a supplier's lead time varies widely, raise the buffer for that item specifically rather than across the category. Uniform buffers spread capital evenly across an uneven problem.
What the chosen position requires
Inventory policy follows from what the business promises rather than from an efficiency target. A distributor competing on next day availability must hold stock and should stop apologizing for the working capital that produces its advantage.
A make to order operation carries the opposite obligation and should resist the drift toward stocking finished goods that begins the first time a customer complains about a lead time. Strategic fit is what settles that argument before it becomes a habit.
Operational excellence here is the alignment between the promise and the stocking rules built to keep it. Stakeholder value erodes where the two drift apart, and the drift is gradual enough that nobody notices until the write-off arrives.
Structure protects the people holding the stock
The argument for governing this is not a tidier warehouse. Where policy is absent, the people closest to the stock absorb the consequences of decisions made elsewhere. That absorption arrives as expediting, apologies, and blame for shortages they did not cause.
That pattern erodes human capital quickly, because the work consists of managing the results of choices the person had no part in. Servant leadership expressed operationally means fixing the specification and approval structure rather than asking the warehouse to compensate for it.
Trust between functions recovers once the rules are explicit and shared. Coherence between what purchasing commits to and what operations can actually hold is a design output, and it is the thing that stops the same argument recurring every quarter.
Watch the full explainer
Related
Further material on management consulting and operational structure from World Consulting Group: [www.worldconsultinggroup.com](https://www.worldconsultinggroup.com)