From counting to visibility: the shift that changes what inventory management is

01 / FIELD NOTE

Keep the decision tied to the operating context.

Inventory management is usually described in terms of counting — how often, how accurately, how quickly. That framing treats the record as something reconstructed periodically from observation, and it accepts that the figure is between counts a description of the past. Visibility is a different proposition: the record is maintained by the movement itself, and the question it answers is what is there now.

The difference is in what can be decided from it. A count supports a reconciliation: it tells you whether the record was right and by how much. A continuously maintained record supports a decision: whether to accept an order, where to send a picker, what to replenish, whether a promised delivery is still achievable. Those are different questions, and only the second is asked during the working day.

Accuracy improves as a consequence rather than as an objective. A record that is written at the moment of each movement has fewer opportunities to diverge from the physical world than one reconstructed at intervals from a memory of what happened. The improvement comes from the frequency of correction rather than from the precision of any single observation, which is why a method that makes counting cheap produces a better record than one that makes counting perfect.

The cost of the old model is mostly invisible, which is why it persists. Safety stock held to cover uncertainty appears as a purchasing decision rather than as a consequence of the record. Labour spent searching, double-checking and reconciling appears as ordinary work. Sales lost because a figure said an item was present appear as a demand problem. None of those is attributed to the record, so the record is not improved.

Frequency is the variable that matters most, and it is the one that the economics of counting determine. A count that takes days happens rarely; one that takes minutes can happen weekly, in small pieces, without displacing the work. The benefit is not that the count is quicker but that the whole rhythm changes, and everything good about the resulting record follows from the rhythm.

Item-level identity is what raises the ceiling. A quantity per product code can say how many of a thing are held; it cannot say which ones, where each is, or what happened to a particular one. The questions a business actually asks during the day — is this specific item available, where is it, what is its condition — are all about specific things, and answering them approximately is not the same as answering them.

The physical limits do not disappear with a change of ambition. Radio behaviour varies with material, density and geometry, a tag that is shielded will not be read however often the count is run, and a deployment that assumes otherwise will discover the assumption at the worst moment. Measuring the intended read points against the actual goods is what turns the ambition into a plan.

The integration is what turns capture into a record. Readers produce observations and the business acts on events, so something has to decide which observations constitute a receipt, a move, a pick or a dispatch. A deployment that leaves that decision unmade produces data that is voluminous, ambiguous and attributed to the wrong process steps.

What the shift changes about people’s work is worth naming, because it is the part that determines whether the change holds. Counting as a separate activity largely disappears, replaced by capture happening as a by-product of the work. What remains is the exception: the tag that did not read, the movement that did not match, the item in the wrong place. That is a different job from counting, and it needs a defined queue and an owner.

The transition should be staged, because the questions only reveal themselves in use. One area, one process, a defined period, measured against the questions the business actually asks rather than against a target accuracy figure. A result in one place produces the evidence for the next, and it produces it before the whole operation has been committed to a method.

The measures worth keeping are the ones that describe the failure rather than the average. How often an order cannot be fulfilled from stock the record said was there, how much stock is present but unavailable, how long between a movement and its posting, how many exceptions are open and how old they are. Those move before a buyer notices, and unlike an accuracy percentage they point at what to change.

The claim worth making is modest and durable. Visibility does not stop goods from being lost and does not make a forecast correct. What it does is keep the record close enough to the physical world that the ordinary decisions taken during a working day are made against something true — which is the difference between managing inventory and periodically finding out how wrong the record was.