The cost hiding in the handoffs

01The owned and the unowned

Ask any organisation where its process is weak and the answers will name steps: the machine is slow, the review takes too long, the form is confusing. Steps attract attention because steps have owners, and an owner is a person who can be asked, measured and helped. Over the years this attention works. The steps of a mature process are usually in reasonable shape, each one polished by the person who lives inside it.

The spaces between steps receive none of this care. No one is measured on what happens after their step ends and before the next begins. No one's name is on the interval. So while the steps improve, the intervals quietly fill with waiting, ambiguity and error, and the process as a whole gets no faster. The cost of a process rarely sits where the work is done. It accumulates where the work changes hands.

02What a handoff really transfers

A handoff looks like the movement of a thing: a pallet to the next station, a file to the next desk, a payment instruction to the next system. But every handoff transfers four things, and only the first is visible. It transfers the material or the data. It transfers responsibility, the question of who now owns the outcome. It transfers context, everything the sender knew about this particular unit that made it different from the standard case. And it transfers assumptions, everything the sender believed the receiver would check, and everything the receiver believed the sender had already checked.

The material almost always arrives. Responsibility usually arrives, eventually. Context and assumptions are dropped constantly, and they are dropped silently, because their absence produces no immediate signal. The error they cause surfaces later, somewhere else, attached to someone else's step, which is precisely why the handoff that caused it is never blamed.

03The tax nobody itemises

Every process pays a rework tax, and almost no organisation itemises it. Rework hides inside ordinary activity: the clarification call, the correction, the re-approval, the second passage of a unit through steps it has already visited. None of these appears as a line in any account, and all of them consume the same hours, machines and attention as first-time work.

The rule that governs the tax is distance. An error caught at its own station costs the minute it takes to fix. The same error discovered three steps downstream costs a multiple: the work built on top of it, the tracing back, the units that followed it through before anyone noticed. The multiple grows with every step of distance, and it grows fastest across handoffs, because a handoff is where the knowledge needed to catch the error is separated from the unit that carries it.

04Queues are frozen cash

Between two steps there is almost always a queue, and a queue is not neutral. Everything waiting in it is cash in a frozen state: materials paid for and not yet sold, invoices earned and not yet issued, settlements initiated and not yet final. The queue also freezes information, because a defect sitting in a queue is a defect not yet discovered, ageing towards the expensive end of the rework rule. Organisations tolerate queues because queues look like buffers, and buffers feel like safety. Priced honestly, days of waiting wrapped around minutes of work, they are usually the largest single cost in the process, and the least defended, because no one owns them.

05Fixing a place nobody owns

Fixing a step is an engineering problem. Fixing a handoff is an organisational one, and that is why handoffs stay broken. A handoff belongs to two owners, which in practice means it belongs to no one: the sender considers the work done, the receiver considers it not yet started, and the interval between those two positions has no budget, no metric and no advocate. Improving it requires two departments to change at once, and each can plausibly wait for the other.

What it takes is modest and specific. Make the transfer explicit: written, not assumed, stating what is handed over and in what condition. Date it, so the interval becomes measurable and the queue becomes visible. And give the defect a single owner, one name accountable for errors found at the boundary, wherever they were made. None of this is sophisticated. It is governance applied to a place governance does not normally reach, and it is usually worth more than any improvement inside the steps.