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 Retractive feedback cannot: a previously valid conclusion may have to be withdrawn, so coordination is required for correctness.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
- Thompson, J. D. (1967). Organizations in Action: Social Science Bases of Administrative Theory.
- Galbraith, J. R. (1973). Designing Complex Organizations.
- Ju, H. (2026). “When Coordination Is Avoidable: A Monotonicity Analysis of Organizational Tasks.”
- Smith, R. P. & Eppinger, S. D. (1997). A Predictive Model of Sequential Iteration in Engineering Design.
- Ford, D. N. & Sterman, J. D. (1998). Dynamic Modeling of Product Development Processes.
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.