More inventory isn’t automatically more availability. It can be the evidence that your replenishment machine is obeying bad instructions with tremendous discipline. The warehouse gets fuller, the cash position gets heavier, and the customer still hears that the line they need is unavailable. Very efficient misery.
Call the control point replenishment signal integrity. It’s the condition where the demand signal, the physical count, the replenishment rule, and the allowed substitute all describe the same operating reality. Vendor-managed inventory delegates the replenishment decision. It doesn’t delegate ownership of bad data, customer service, or working capital.
That matters because your stock has several states at once. Physical quantity, owned inventory, committed quantity, and available-to-promise quantity aren’t interchangeable.
A supplier can retain title to stock on your shelf. An accepted order can commit stock without removing it physically. A system can show available stock that a count can’t find.
One number on a dashboard has never been enough. It just has better manners than a stockout.
The Boundary
Treat your VMI as a controlled delegation, not as a magic shelf that refills itself. Before a supplier replenishes for you, the arrangement needs a shared definition of:
- Stock boundary → item identifiers, physical locations, title-transfer point, minimum and maximum stock, and obsolete-stock treatment.
- Signal boundary → demand history, forecast visibility, promotions, returns, lead-time assumptions, and the count-reconciliation interval.
- Authority boundary → replenishment frequency, authority to replenish, authority to exceed the maximum, discrepancy handling, and the exit treatment for remaining stock.
Those are operating controls, not contract decoration. A supplier who receives only an on-hand balance can be perfectly punctual and still feed the wrong bin. If you haven’t defined whether ownership transfers on delivery, withdrawal, or invoice, you can’t say clearly whose inventory cash is growing.
And a substitute that’s merely available isn’t necessarily approved for the customer’s application.
Keep the four inventory quantities separate in your record. Physical on hand answers what exists in the warehouse. Owned inventory answers whose balance sheet carries it. Committed stock answers what accepted orders have already claimed. Available to promise answers what you can offer a new customer.
The broader setting changes how much buffer the system needs. The World Bank’s 2023 Logistics Performance Index found an average interval of 44 days from entering an export port to leaving the destination port, with wide variation across potential routes.
A local warehouse can be a timing buffer in that world. It still isn’t an excuse to replenish a phantom balance.
The Scorecard
Don’t score your arrangement with a single fill-rate percentage. Use a compact vendor-specific scorecard, since NIST’s small-manufacturer guidance favours two or three key measures mixed with qualitative evidence rather than an indiscriminate KPI catalogue.
The six fields below are the minimum diagnostic set, because each one exposes a different way that more stock can coexist with worse service.
| Scorecard field | Literal definition | Evidence to retain | What a bad reading means |
|---|---|---|---|
| Demand signal | Consumption, forecast, promotion, return, and lead-time inputs used for replenishment | Source system, revision, date, and exception code | The rule may be acting on demand that no longer exists or missing demand that does. |
| Physical count variance | Difference between a physical count and the shared inventory record | Count sheet, location, item identifier, adjustment owner, and reason code | The supplier may replenish an inventory figure rather than inventory that exists. |
| Fill definition | The stated numerator, denominator, event boundary, exclusions, and observation unit | Order, order line, unit, delivery, or customer-demand record | A high percentage may omit the demand, line, or time interval that actually failed. |
| Stockout duration | Hours or days from unavailability to restored availability | Stockout start, restoration event, affected line, and customer impact | A short count of stockouts can hide one long interruption. |
| Substitution rate | Frequency and value of substitutions, with approval status | Original line, substitute, approver, reason, and customer outcome | Shipped volume can be protected by sending the supplier’s available stock instead of the needed item. |
| Inventory days | Inventory held relative to the demand it is meant to cover | Ownership, location, aging, and replenishment decision | Working capital may be increasing in slow or obsolete lines while the high-runner remains short. |
Your denominator isn’t a footnote. On time asks whether receipt lands in the agreed window. In full asks whether accepted quantity and assortment match the order. OTIF requires both. A late but complete delivery can’t offset an early shortage, and an order-level measure can hide the line that stops your customer’s job.
Score at order-line level when lines are independently needed, and keep an order-level view where one missing line makes the whole order unusable.
Preserve your original and revised promise dates. The revised date can govern recovery. The original date reveals how often a supplier asked to move the commitment.
Preserve missing data as missing, too. A blank receipt date isn’t an on-time delivery wearing an invisibility cloak.
Before You Buy a Pallet
Use the shape of the failure before you change a reorder point or buy another pallet. The same symptom can belong to three different stages.
Inventory rises while the same high-runner stocks out → the inventory signal is broken
Compare system quantity with a physical count. Inspect returns recorded as consumption, late emergency withdrawals, missing receipts, location transfers, and adjustments without a retained reason code.
The supplier ships to the agreed trigger, but the warehouse is still short → the demand or count input is broken
Check the consumption history, forecast revision, promotion feed, lead-time input, and count-reconciliation interval before you call the shipment late.
The supplier fills orders, but the wrong items accumulate → the incentive or substitution control is broken
Inspect whether shipped revenue, case fill, or a supplier-side fill rate is rewarded while availability, obsolescence, inventory cash, and substitution quality are not.
Buying reports a high fill rate while customers report unavailable lines → the measurement definition is broken
State whether the denominator is units, order lines, orders, deliveries, or actual customer demand, and show stockout duration and cancellations separately.
VMI looks healthy because the supplier owns the stock, but cash and space pressure rise → the ownership boundary is broken
Separate physical stock from buyer-owned inventory, then inspect title transfer, maximum stock, aging, and obsolete-stock treatment.
A manager calls the supplier unreliable after every shortage, but the evidence is mixed → the attribution rule is broken
Measure forecast error separately from replenishment execution. A valid rule acting on bad input isn’t the same failure as a supplier ignoring valid input.
That ladder keeps the conversation factual. It also stops procurement from turning every data problem into a supplier negotiation and every supplier problem into an emergency buy. Both are expensive forms of handwriting.
The Cash Link
Your inventory cash isn’t limited to the invoice price. Procurement cost includes freight, tariffs, longer lead times, inventory carrying requirements, and the overhead of managing distant suppliers.
Distribution adds receiving, storage, picking, packing, delivery, credit, returns, and information systems. Your gross-margin percentage is the opening pool that has to fund all of that. It isn’t a permission slip to fill every empty shelf.
The trade-off changes by line and by location. Fastenal held roughly 54% of its 2025 year-end inventory at customer-near selling locations and 46% at distribution and manufacturing locations.
Local stock can support same-day availability. It can also fragment the pool and strand slow movers. Central stock aggregates demand, but it adds transport time and a handling leg.
Your scorecard therefore needs inventory days and stockout duration together, not a universal turns target.
Supplier incentives can make your cash error predictable. A purchase made to cross a rebate tier can lower cost of goods sold while leaving you with inventory customers won’t absorb. Price protection, stock rotation, and obsolescence protection can change the loss, but the agreement decides whether they apply and how claims are handled. Stock that’s protected on paper can still consume space and cash while the claim waits.
Monday
Don’t begin with the whole warehouse. Begin with one high-runner that stocked out while inventory grew.
The point is to establish a defensible fact pattern before your next replenishment order makes the picture worse.
- Count the line physically at each location you use. Record the item identifier, lot if relevant, physical quantity, system quantity, ownership state, committed quantity, and available-to-promise quantity.
- Pull the recent demand signal and replenishment record. Keep the consumption, forecast revision, promotion, returns, trigger, supplier order, promised date, revised date, receipt, and any substitution together.
- Assign the failure: signal if data or demand was wrong, execution if valid instructions weren’t performed, substitution if service was claimed with an item the customer didn’t approve. Give the correction an owner and keep the reason code.
If the line is critical and your only alternate is a vendor record, don’t call the category resilient.
An executable alternative has current drawings, tooling access, approved samples, workable lead time, and order history. A name in a database isn’t immediate recovery capacity.
More inventory does not equal more availability. Better signals equal more availability. A supplier can execute perfectly and still replenish the wrong business. Measurement decides which happened.
Count the line. Name the signal. Define the denominator. Only then buy more stock.