# Adaptive Planning

> Hold the goal steady, treat the plan as a hypothesis, and revise the route as work produces evidence.

- HTML version: https://robbiepalmer.me/ideas/adaptive-planning
- Source: https://www.gov.uk/government/publications/uk-defence-doctrine-jdp-0-01

Adaptive planning holds a goal steady while repeatedly changing the route used
to reach it. The plan is the best current hypothesis about what to do. It will
be wrong because it rests on incomplete knowledge and because doing the work
changes what is known.

This makes a plan valuable without making it sacred. Its job is to coordinate
the next action, expose dependencies, and create evidence. When the evidence
contradicts the plan, revise the plan. Defending it would confuse consistency
with progress.

Changing the goal is a different act. Within one planning loop, the goal defines
success and remains fixed. If the goal changes, begin a new loop and judge the
existing plan against the new goal.

## Fixed ends, revisable means

Military doctrine makes this distinction explicit. The UK's
[Defence Doctrine](https://www.gov.uk/government/publications/uk-defence-doctrine-jdp-0-01)
describes strategy in terms of ends, ways, and means. It says that execution will
unavoidably require adaptation: the ends should generally remain constant while
the ways and means are adjusted.

The same distinction appears in project research. Turner and Cochrane's
[goals-and-methods matrix](https://doi.org/10.1016/0263-7863%2893%2990017-H)
separates how well the goal is understood from how well the method is understood.
A project can have a clear goal while the method for achieving it remains
uncertain. In that case, discovering the method is part of the work.

Military mission command deals with this condition by giving people the purpose
and desired end state, then allowing initiative when events depart from the
order. The US Army's
[description of commander's intent](https://www.army.mil/article/215297/mission_command_requires_sharp_commanders_intent)
emphasises the end state and gives subordinates room to choose the ways and
means. A detailed order cannot anticipate every event, but a shared intent can
still coordinate action.

## A plan is a set of assumptions

The fallibilist strand of [epistemology](/ideas/epistemology) explains why action
does not require certainty. A belief may be justified while remaining open to
correction. Planning applies the same stance to action: proceed on the strongest
account available and state what would cause it to be revised.

RAND's
[assumption-based planning](https://www.rand.org/pubs/monograph_reports/MR114.html)
makes the logic operational. It identifies critical assumptions whose failure
would force a major change, signposts that warn an assumption is failing,
and actions that shape or hedge against the consequences. McGrath and MacMillan's
[discovery-driven planning](https://hbr.org/1995/07/discovery-driven-planning)
similarly treats milestones as tests of assumptions. Under high uncertainty,
the point of early work is to buy useful knowledge before committing more
resources.

This changes what belongs in a plan. Record the goal, constraints, dependencies,
current assumptions, and tests that could disprove them. Describe distant work
only as far as the evidence warrants. PMI's account of
[rolling-wave planning](https://www.pmi.org/learning/library/manage-innovation-programs-rolling-wave-3515)
uses detailed plans for the current horizon and fills in later work as uncertainty
is removed. Premature detail gives weak assumptions the appearance of fact and
creates a plan that costs money to maintain.

## The graph is provisional

A [directed acyclic graph](/ideas/directed-acyclic-graph) can represent the
current route to the goal. Its dependency edges expose the current critical path
and show which work can proceed in parallel. They do not turn the route into a
prediction.

Downstream nodes inherit unresolved questions from their ancestors. They should
therefore become less detailed and easier to replace as those questions
accumulate. Graph depth is only a warning. A routine task may sit deep in the
graph, while an immediate experiment may have a highly uncertain result. The
actual sources of uncertainty are the assumptions attached to the nodes and
edges.

The critical path is provisional too. Herroelen and Leus found that most project
scheduling research assumed complete information and a static environment,
although real activity durations are uncertain and become known during
execution. Their
[survey of scheduling under uncertainty](https://doi.org/10.1016/j.ejor.2004.04.002)
covers reactive, stochastic, and robust approaches. An adaptive planner should
recompute the critical path when evidence changes durations, dependencies, or
the work itself.

## Keep the queue fluid and work in flight stable

Unstarted work is an option. Replenish and reprioritise the queue whenever new
evidence changes the best route. When capacity appears, pull the
highest-priority eligible item. Commitment begins when work starts.

This follows the flow logic in the
[Kanban Guide](https://kanbanguides.org/the-kanban-guide/). Controlling work in
progress creates a pull system in which new work starts when capacity is
available. Kanban measures cycle time from an item's start to its finish and
bases its service-level expectation on historical cycle time. That is the useful
forecasting boundary: estimate the delivery of work in flight rather than assign
dates to an uncertain chain of future tasks.

Work in flight deserves stability because interruption discards context and may
leave partial work behind. Stability still has a limit. Stop or redirect active
work when evidence shows that completing it will no longer advance the goal, or
when the cost of continuing exceeds the cost of interruption.

Herroelen and Leus call the disruption caused by frequent rescheduling
**nervousness**. Adaptive planning therefore has two different change costs:
reordering an unstarted queue should be cheap; changing active work should clear
a much higher threshold. This asymmetry permits continual adaptation without
continual thrashing.

## Test the plan before committing to it

Wargaming creates evidence about a plan without paying the full cost of executing
it. The UK Ministry of Defence
[Wargaming Handbook](https://www.gov.uk/government/publications/defence-wargaming-handbook)
places wargaming in planning and executive decision-making. Its guidance on
[wargaming and red teaming](https://moderncivilservice.blog.gov.uk/2023/11/06/skills-wargaming-and-red-teaming-how-the-mod-is-challenging-defence-thinking/)
uses structured simulation to expose flawed logic, hidden assumptions, and
consequences.

Experiments, prototypes, rehearsals, and simulations perform the same planning
function. Each should target a decision. If no plausible result would change
the plan, the exercise produces information without adaptation.

Adaptive planning can therefore be stated as a short loop:

1. Fix the goal and the constraints that define acceptable action.
2. Build the smallest plan that exposes ready work, dependencies, and the
   current critical path.
3. Attach critical assumptions to the parts of the plan that depend on them.
4. Pull eligible work when capacity becomes available and protect it from weak
   priority changes.
5. Treat results as evidence. Update assumptions, replace affected branches,
   recompute the critical path, and reprioritise the queue.
6. Repeat until the goal is reached or explicitly replaced.

The discipline lies in doing both halves. Act despite uncertainty. Abandon a
route when the evidence turns against it.

## Sources

* Ministry of Defence. [*Joint Doctrine Publication 0-01, UK Defence
  Doctrine*](https://www.gov.uk/government/publications/uk-defence-doctrine-jdp-0-01).
  Sixth edition, 2022.
* Perna, Gustave.
  ["Mission command requires sharp commander's
  intent"](https://www.army.mil/article/215297/mission_command_requires_sharp_commanders_intent).
  US Army, 2019.
* Turner, J. R., and R. A. Cochrane.
  ["Goals-and-methods matrix: coping with projects with ill defined goals and/or
  methods of achieving them"](https://doi.org/10.1016/0263-7863%2893%2990017-H).
  *International Journal of Project Management* 11, no. 2, 1993.
* Dewar, James A., Carl H. Builder, William M. Hix, and Morlie Hammer Levin.
  [*A Planning Tool for Very Uncertain
  Times*](https://www.rand.org/pubs/monograph_reports/MR114.html). RAND, 1993.
* McGrath, Rita Gunther, and Ian C. MacMillan.
  ["Discovery-Driven Planning"](https://hbr.org/1995/07/discovery-driven-planning).
  *Harvard Business Review*, July-August 1995.
* Githens, Gregory D.
  ["Manage Innovation Programs With a Rolling
  Wave"](https://www.pmi.org/learning/library/manage-innovation-programs-rolling-wave-3515).
  *PM Network* 15, no. 5, 2001.
* Herroelen, Willy, and Roel Leus.
  ["Project scheduling under uncertainty: Survey and research
  potentials"](https://doi.org/10.1016/j.ejor.2004.04.002).
  *European Journal of Operational Research* 165, no. 2, 2005.
* Coleman, John, and Daniel Vacanti.
  [*The Kanban Guide*](https://kanbanguides.org/the-kanban-guide/). May 2025.
* Ministry of Defence.
  [*Wargaming Handbook*](https://www.gov.uk/government/publications/defence-wargaming-handbook).
  2017\.
* Savage, David, and Rupert Mott.
  ["Wargaming and Red Teaming: How the MoD is challenging defence
  thinking"](https://moderncivilservice.blog.gov.uk/2023/11/06/skills-wargaming-and-red-teaming-how-the-mod-is-challenging-defence-thinking/).
  Ministry of Defence, 2023.

## Related ideas

- [Directed Acyclic Graph (DAG)](https://robbiepalmer.me/ideas/directed-acyclic-graph.md): A directed graph with no path that loops back to its starting node, useful for representing dependencies and executable plans.
- [Epistemology](https://robbiepalmer.me/ideas/epistemology.md): How we know what we claim to know, what justifies a belief, and when new evidence should change it.
- [Goodhart's Law](https://robbiepalmer.me/ideas/goodharts-law.md): When a measure becomes a target, pressure to improve the number can break its connection to the outcome it once represented.

## Where it appears

- Project: [Work Graph](https://robbiepalmer.me/projects/work-graph.md)

---

Markdown index of this site: https://robbiepalmer.me/llms.txt