Dependency is the weakest of the five clarities, and it is where execution gives way first.
Why it matters when the plan changes
Execution breaks horizontally. A plan is a chain of commitments between units, and the chain is only as good as its least reliable link. Sull, Homkes and Sull found that only 9% of managers can rely on colleagues in other functions all the time, against 84% inside their own line. Brooks's law explains why this gap widens as a plan grows: the number of communication channels between people rises faster than the number of people added, so each new interface raises the cost of keeping commitments in step. The plan assumes the horizontal ones hold regardless.
The tension is completeness against use. A map that lists every interface is unreadable and goes stale within a quarter. A map that lists only the interfaces the current plan depends on, with a named owner and a date on each, stays short enough to be argued with in a meeting. The second kind changes decisions; the first kind decorates a wall.
In practice
A launch depends on a legal review, a data feed from finance and a partner contract, in that order. The plan shows all three as parallel workstreams with owning functions rather than owning people. Mapping them reveals the contract cannot start until the data feed is specified, which nobody had sequenced. The launch date was never achievable, and that was knowable in week one.
Evidence
Cross-unit commitments are reported as the least reliable part of execution, well below commitments inside a reporting line.
Donald Sull, Rebecca Homkes and Charles Sull, Why Strategy Execution Unravels and What to Do About It, Harvard Business Review (2015)The number of paths between people grows faster than the number of people, so every added party raises the cost of keeping a plan in step.
Brooks's law, after Frederick P. Brooks, The Mythical Man-Month (1975)
What it cannot tell you
Dependency mapping shows which interfaces a plan relies on, not whether the people named at each one will keep their commitment. It records structure, not behaviour under pressure. A map with the right names and dates can still fail if workloads shift, priorities change elsewhere, or an owner is replaced without the map being updated.
Questions
Only what this plan depends on. Each entry names the handoff, the person answerable on each side, and the date it must clear. Sull, Homkes and Sull, writing in Harvard Business Review in 2015, found only 9% of managers trust colleagues outside their own function, which is why a function's name is never enough.
A project plan sequences tasks inside one delivery, where everyone reports to the same owner. A dependency map records commitments between units that do not. Frederick P. Brooks's The Mythical Man-Month observed that communication channels multiply faster than the number of people involved, which is why cross-unit maps need named owners and dates to stay honest.
Whenever the plan changes, and whenever an interface changes owner. A new dependency is one of six events that warrant a fresh reading of the organisation. Between those points a map ages quietly, and an out-of-date map is more dangerous than none because it is still trusted.
Because it belongs to nobody in particular. Direction, priority and role all have an obvious owner in the line. An interface sits between two owners, so the work of keeping it clear is nobody's main task, and it is the first thing to lapse when both sides are busy.
It converts an assumption into a commitment with a name and a date, which can then be checked. It also makes sequencing errors visible early, when the plan can still be resequenced, rather than at the milestone where the only remaining options are late delivery or reduced scope.