A project that crosses three departments rarely fails inside any of them. It stalls in the gaps, where work sits waiting for a team that has not agreed it is urgent.
Each function optimises its own queue
Departments are measured on their own throughput. A design team is judged on design output, not on how quickly it unblocks an engineering team downstream.
So incoming requests from other functions join a queue ordered by that function's priorities. A request that is critical to the sender may sit mid-list for the receiver.
Nobody is behaving badly. Each manager is doing precisely what their own measurement rewards, and the delay appears only when the whole path is viewed end to end.
Waiting time is invisible in every individual report
Status reports record activity, and waiting is not activity. A project spending three weeks in a queue shows up as no progress rather than as a specific, locatable cost.
Because the waiting occurs between reporting boundaries, it also belongs to no one's dashboard. Each department can honestly report that its portion moved quickly.
Aggregate the honest reports and the elapsed calendar time does not reconcile. The missing weeks were spent in transit.
Escalation is the only priority mechanism available
Without shared prioritisation, the sender's remaining option is to escalate to a manager senior to both functions. That works, which is exactly the problem.
Once escalation works, it becomes the standard route, and queue position starts to reflect the seniority of whoever complained rather than the value of the work.
Teams then learn to escalate pre-emptively, and executives spend their time arbitrating sequencing decisions that could have been made two levels down.
Assigned coordinators inherit responsibility without authority
The usual fix is a project manager who tracks the work across boundaries. They can see the delay clearly, which is genuine progress, but they cannot reorder another department's queue.
What they can do is make waiting visible and attach a date to it, which converts a vague delay into a specific request someone must decline in writing.
That is often enough. Much queue delay survives because no one has been asked to explicitly refuse it.
Shared measurement changes behaviour faster than structure
Reorganising functions into a single reporting line solves handoffs by eliminating them, but creates new boundaries elsewhere and is expensive to reverse.
Measuring the end-to-end duration and reporting it to every contributing manager is cheaper. It makes the gap a shared number rather than an unowned one.
The behaviour that follows is simple: managers begin negotiating queue position directly, because the delay now appears on a report with their name attached.