The inventory is human attention.

"WFM is often viewed as the scheduling department. It needs to be viewed as Strategic Inventory Management. The inventory is human attention. The job is to ensure enough inventory to meet demand without bankrupting the company, or burning out the inventory."

(excerpts from my working manuscripts, here and below)

That is a reframe, and reframes are cheap. This one earns its keep because it survives being taken literally. Push on it term by term and it does not turn into metaphor. It turns into the actual mechanics of the function.

The forecast is the inventory plan. The schedule is the allocation. Shrinkage is the carrying cost. Occupancy is the spoilage limit. Variance is the demand signal the operation has not accounted for yet.

Take those one at a time, because each carries a consequence.

The forecast as inventory plan means it is a commitment about supply, not a guess submitted for approval. A plan that changes because the number is uncomfortable was never a plan. What a plan can do (and a guess cannot) is carry its assumptions on the outside, so that anyone who wants a different output has to name which input they want moved and accept the trade that comes with moving it.

The schedule as allocation means every hour in it was assigned to something at the expense of something else. There is no neutral hour. This is why coaching that is not in the schedule is not real: the hour was allocated to handling contacts, and the operation decided that whether or not it wrote the decision down.

Shrinkage as carrying cost is the term that most changes the conversation. Carrying cost is not waste. It is what you pay to hold inventory that will still be usable when you need it. Training, coaching, breaks, leave and development time are the cost of keeping human attention in a condition to be spent. An operation that treats shrinkage purely as leakage is not economizing. It is refusing to pay the carrying cost and then acting surprised when the inventory spoils.

Occupancy as spoilage limit is the same idea with a ceiling attached. There is a level of sustained occupancy above which attention degrades: slower recovery between contacts, thinner notes, shorter patience, worse judgment on exactly the contacts that need judgment. Running above it does not create capacity. It borrows capacity from next quarter at a rate nobody prices.

Variance as an unaccounted demand signal is the most useful of the five, because it stops variance from being a blame event. A miss is information about demand the model did not know it had.

There is one place where the inventory analogy breaks, and the break is the important part: human attention cannot be warehoused. An interval that is overstaffed does not bank the surplus, and an interval that is understaffed does not simply record a stockout and move on. Queued work crosses the interval boundary and lands as a live input to the next interval, while most planning treats each interval as though it starts clean. That is where the plan and the floor begin to disagree.

Even operations that have accepted the reframe miss this:

"A forecast can be mathematically clean and operationally false because the interval is wrong. When average handle time is large relative to the interval, the work crosses the interval wall: the model reports coverage while the floor carries hidden backlog… That is not a failure of effort. It is a failure of geometry: shorter intervals give you more detail, not automatically more truth."

An inventory plan denominated in the wrong unit of time will report a healthy position while the shelves are empty. Nothing in the standard practice looks careless: forecast the day’s volume, spread it across the intervals along an arrival curve, staff each interval to the result. When it misses anyway, two reflexes follow: shorten the interval to fifteen minutes, then to five, or fit a better curve to another year of history. Both buy accuracy on arrivals, and an arrival curve only says when work begins; the floor runs on when work occupies. The two are nearly the same curve while handle time stays short against the interval. When it runs long they separate, and the load lands downstream of where it arrived. A requirement that means to describe the floor has to be solved on the workload each interval actually holds, duration and carryover included, not on the volume that arrived in it. The two curves keep pulling apart, because the contacts reaching the human layer are the longer, more complex ones.

When the plan and the floor disagree, there is a default worth adopting:

"WFM sees the data. Operations sees the floor. When those two views disagree, the floor is almost always right and the data is catching up."

That is an unusual thing for a workforce management function to concede, and it is the correct default. The floor is where the unaccounted demand signal appears first. The data gets there second, once whatever happened has been categorized, entered, and rolled up. A WFM function that treats the disagreement as a challenge to its numbers loses the earliest warning it will ever get.

What keeps the plan improving is what happens after the miss, and it needs to be narrower than a discussion:

"…what was expected, what happened, what the delta was, and what drove the delta. Not a story. Not a posture. An attribution. The miss was volume, or AHT, or shrinkage, or staffing, or adherence, or an outage, or an upstream event, or interval geometry… Attribution that protects a function does not prevent the next miss. It teaches the model the wrong lesson."

Attribution is what turns a variance into inventory intelligence rather than a meeting. It is also the discipline that makes the earlier reframe safe: the moment shrinkage is a carrying cost and occupancy is a spoilage limit, someone will be tempted to attribute every miss to conditions nobody controls. A fixed category list makes that harder to do quietly.

None of this makes workforce management more important than the floor it serves. It makes it legible. When the function is the scheduling department, its requests arrive as departmental preference and get traded away first. When it is inventory management for the only asset the operation actually sells, coaching hours, recovery time and exception capacity can be argued in the same sentence as service level, which is the sentence where they have always needed to be argued and rarely have been.

The arithmetic behind the requirement, live: service level against agents on one interval.