Interdependence & Coordination

Coordination Fit

Many programmes are run as though work passes cleanly from one specialist to the next. That works when the handoff is stable. It fails when a change in one part of the product changes what another part can build, fund, sell, secure, or operate.

The practical question is whether new information adds to an existing commitment or requires that commitment to be revised. The answer determines where coordination belongs and how quickly it must happen.

The structure of the work comes first

James D. Thompson distinguished three forms of task interdependence. The categories describe relations between activities, not the character of people or teams. Each calls for a different coordination mechanism.

Pooled

Independent contributions combine

Teams contribute to a common result without changing one another's work. Standards, rules, and a final merge are usually enough.

Coordination fit: shared rules and aggregation.

Sequential

Known output moves downstream

One unit's output becomes another's input. The handoff and its order matter, but a valid earlier decision remains usable.

Coordination fit: interface, schedule, and causal handoff.

Reciprocal · additive

Mutual learning accumulates

Teams inform each other repeatedly, while earlier conclusions remain valid. The loop helps performance without necessarily invalidating work.

Coordination fit: fast exchange, shared visibility, and proportionate integration.

Reciprocal · retractive

Mutual adjustment changes the answer

A new constraint, trade-off, or discovery makes existing commitments invalid. The work must be revised across the boundary before more dependent work accumulates.

Coordination fit: local decision rights, high-bandwidth context, and a short adjustment loop.

Feedback is the boundary

Reciprocal work is not always retractive. A research team may give a design team more evidence without invalidating its current choice; that is additive feedback. A hardware tolerance, security finding, capacity limit, market commitment, or architecture decision may instead rule out work that was previously acceptable. That is retractive feedback.

Ju formalises one part of this distinction through monotonicity. Under the paper's task representation, pooled work and non-retractive sequential work can remain correct as history grows. Retractive feedback cannot: a previously valid conclusion may have to be withdrawn, so coordination is required for correctness. The result does not justify coordination everywhere; it identifies the cases in which omitting it risks an invalid outcome.

Governance has to carry the adjustment

Galbraith treats uncertainty as a demand for information processing. Rules and plans help when exceptions can be anticipated. When they cannot, the organisation needs a way to process new information where the relevant context is held: direct contact, an integrating role, a cross-functional team, or a different boundary.

A monthly programme review can be suitable for a stable handoff. It is a poor fit when a contradiction can leave several weeks of dependent work built on an outdated commitment. The resulting delivery slip is often rework rather than a simple planning error.

Sources and limits

This module is a screening aid and an explanatory model. The cost section uses a symmetric, two-activity rework model with a Design Structure Matrix dependency structure. It is not a formal task verification or a budget estimate. Results depend on how the work and its boundary are defined.