The books reconcile. The accountant signs off. The same variance is back next month.
For most CFOs at mid-market distributors and manufacturers, the inventory close is not a question of whether the books will reconcile. They will. The variance will be explained.
The question is why the same variance shows up again the next month. The explanation is almost never an accounting error. It is structural, and it sits in 3 places that monthly inventory reconciliation is not designed to surface.
Reconciled is not the same as defensible
A reconciled book balances. A defensible book holds up under scrutiny that is not the accountant's.
A lender reviewing a facility renewal does not care that the variance was explained. They care whether the inventory number on the balance sheet is supported by traceable transactions. A buyer running diligence gives no benefit of the doubt to a recurring adjustment. They apply a haircut.
A reconciled number is good enough to close the month. It is not always good enough to bet the next 12 months on.
What is found: 3 failure points
1. Cost logic that no longer matches operations
Cost logic gets set up once, often years ago. Operations then makes small changes: a new supplier with different freight terms, a handling charge absorbed differently, a line item moved from one cost category to another. Each change is reasonable. None is flagged to finance.
Finance and operations end up looking at the same data and producing different numbers from it. The reconciliation closes the gap mechanically every month. The drift underneath is never addressed. This is what produces a stable-looking gross margin that nobody can fully explain when the auditor asks.
2. Definitions that drift across systems
Most mid-market companies run more than one system: one for finance, one for the warehouse, one for purchasing. Each has its own definition of what counts as inventory, when, and at what value. Those definitions were aligned at implementation. They rarely stay aligned.
The classic example is timing. A purchase order is recorded when it is issued in one system, when it is acknowledged in another, and when goods are received in a third. Each is consistent on its own. Together they put cost in the wrong period. The numbers balance. They are also wrong.
3. Nobody owns the reconciliation logic
When two systems disagree, someone has to decide which one is right. In most mid-market companies, nobody is formally responsible for that decision. Finance assumes operations owns the data. Operations assumes finance owns the reporting. The variance gets resolved by whichever path is least disruptive that month, not by a documented standard.
What it costs
The recurring variance itself is rarely the largest cost. The larger cost is the decisions made on top of numbers that look defensible and are not: pricing that compresses margin, purchasing that ties up working capital in the wrong inventory, capital commitments approved on a gross margin trend the data does not support.
AI makes this worse, not better, when it runs on top of structurally wrong data. It does not surface the error. It summarizes it, in a format the board takes seriously. The finance leaders getting real value from AI fixed the data foundation before the model saw the data.
The fix
The fix is not more reconciliation. A defensible close has 3 properties:
- Cost logic is documented and matches how operations runs the business today, not how the system was set up years ago.
- Definitions are aligned across every system that touches inventory, and that alignment is tested, not assumed.
- An owner is named for every figure where two systems can disagree, and the reconciliation logic is the same every month.
The result
A medical consumables company had booked unexplained inventory adjustments every month for two years. The books balanced and margin looked stable. Purchase orders, supplier invoices and warehouse receipts turned out to be dated on 3 different logics, so cost was landing in the wrong period. Once the timing was aligned across purchasing, receiving and accounting, about $20,000 of adjustments nobody could explain was gone.
"We had reconciled those numbers every month for two years. Nobody had ever mapped the timing across all three."
To see the same pattern on a worked example, open the month-end dashboard. To check your own figures, the free adjustment test says whether a month-end adjustment is noise or a pattern.
None of this is glamorous. It is also the highest-value work a finance leader at this size can do before a lender review, a sale, or a serious AI project: it is what lets the CFO defend every number in the room.
Is this your number?
If your close balances and the number still moves, this is the work: a month-end figure that matches what actually happened.