12 minute read
- The problem-led principles apply one evidence-focused decision loop to both everyday work and organisational change and funding decisions. see why — jump to the section that argues this
- Work is worth doing only to the extent that it reduces a named problem for someone the organisation exists to serve. see why — jump to the section that argues this
- The problem-led principles are designed to protect five specific values, which are treated as held commitments rather than hypotheses to test. see why — jump to the section that argues this
- Start from a named problem, tied to what the organisation is for, before choosing any work, change, or investment.
- Count work as done only when it is in use by the people it was for and producing evidence that the problem is shrinking.
- Limit what runs at once: start nothing without capacity set aside, and when you start one bet, name which bet stops or waits.
Why these principles exist
Work exists to solve problems for the people an organisation serves, and the people doing that work can see those problems better than anyone. These principles start from that belief.
In my experience, many organisations ask their teams to work from evidence: try something small, check whether it helped, adjust. Then the same organisations decide what to fund and how to change themselves on plans, opinion and money already spent. The cost is real. Money keeps flowing to work nobody uses, changes run for a year before anyone can say whether they helped, and decisions are defended by how much has been spent rather than by what has been learned.
The problem-led principles apply one discipline to both kinds of decision. It runs as one loop:
a named problem, a hypothesis, something in use, evidence, a decision.
Name the problem you are solving. State what you expect your change or investment to achieve, as a hypothesis you can check. Get something into use. Read the evidence. Then decide, openly, whether to stop, adjust or continue. The loop runs wherever work is chosen or the organisation is changed, from a single team to the whole organisation.
Values
The principles exist to protect five values. These are commitments: they are held, not tested.
- Respect for people. Results come from the system people work in, not from blaming the people in it.
- Outcomes over methods. Methods, frameworks and tools are means. The problem solved is the point.
- Honesty about evidence. Decide on evidence, not on plans, opinion or money already spent, and say plainly what is not yet known.
- Transparency in both directions. Work, limits, decisions and risks are visible to everyone they affect.
- Improve while delivering. The organisation changes in small steps while its work carries on, never by halting that work for a big change.
First principles
A first principle is a starting point for reasoning: it is not derived from anything else on this page. Calling something a first principle does not make it true, so each one says what it rests on, where it applies and what follows from it. Those that rest on observation also say what evidence would show them wrong. The works each one draws on are listed under “Sources”, where they are still being checked.
1. Work exists to solve problems for those the organisation serves
Work is worth doing to the extent that it reduces a problem for someone the organisation exists to serve.
- What it rests on. A commitment. It is the stance these principles take, and no evidence is offered for it.
- Where it applies. Every piece of work, change and investment an organisation chooses to make. Legal, safety and contractual obligations are not optional, so they come first and limit what else can be chosen.
- What follows. Work that cannot be traced to a named problem has no claim on time or money. Problems, not solutions or projects, are what get pursued, so choosing which problem to work on also decides which ones wait.
2. Complex work can be tested but not reliably predicted
Where cause and effect only become clear afterwards, the effect of a change or an investment cannot be reliably predicted. It can be found out by trying it and watching what happens.
- What it rests on. Observation of complex work.
- Where it applies. Changing an organisation, new products and services, and markets. It does not apply to clear, well-understood work, where a known fix can simply be applied.
- What follows. Plans for complex work are guesses to be tested. Commit to the problem, not to the first solution.
- What would show it wrong. Long-range plans for this kind of work that reliably deliver what they predicted.
3. Nothing produces value until it is used
Work produces value only when someone uses it and is better off for it.
- What it rests on. A definition of value. Its worth is in what follows from it.
- Where it applies. All output: products, features, changes to how work is done, and decisions.
- What follows. Built, shipped or announced is not done. Something has to be in use before anyone can tell whether it helped.
- What would make it worth dropping. Evidence that counting unused work as done leads to better decisions about what to pursue.
4. The people closest to the work see what others cannot
The people who do the work, and the people it serves, see problems and effects that those further away do not.
- What it rests on. Observation.
- Where it applies. Finding problems, and judging whether a change is working.
- What follows. Go and see the work where it happens. Ask the people affected. When they decline a change, that tells you something about the change.
- What would show it wrong. Decisions made at a distance that consistently match or beat close observation at finding problems and reading effects.
5. The system largely determines the results
Most of the variation in an organisation’s recurring results comes from its system: its structure, funding, decision rights, measures and the assumptions it runs on. Individual effort accounts for far less.
- What it rests on. Observation, grounded in the theory of variation. Variation that shows up period after period belongs to the system. Variation from a particular event has a particular cause.
- Where it applies. Recurring results, read over a run of periods rather than one. It does not apply to one-off events.
- What follows. When results stay poor despite effort, change the system, not the people. A measure turned into a target distorts the system it was meant to measure.
- What would show it wrong. A run of results that changes mainly through individual effort while the system stays the same.
6. Finite capacity means queues delay use and evidence
An organisation can do and absorb only so much at once. The more it starts, the longer each piece waits before it is in use and its evidence can be read.
- What it rests on. Observation of how work flows through queues.
- Where it applies. All work in progress, at every scale.
- What follows. Every new start delays something already waiting. Keep the number of things running small, and protect some capacity rather than planning every hour.
- What would show it wrong. Organisations that add work in progress, at the same capacity, without lengthening the time it takes for work to be used.
Derived principles
These follow from the first principles, and they are what the problem-led guides put into practice. Each notes the first principles it comes from.
In short, for whoever decides: what to work on (1, 15), what counts as done (3, 4), how much runs at once (5, 11, 16), when to stop or fund more (6, 12, 17), how far a decision can be undone (18), and who decides what (8, 13, 14).
- Start from a named problem, tied to what the organisation is for. (From first principle 1.)
- Every change or investment in complex work is written as a hypothesis that evidence can refute, with a continuous monitor and a timeframe for its expected outcome. It is tested continuously from the start, not only at a review. (From 2.)
- Done means in use. Count nothing until it is used and producing evidence. The organisation sets a minimum Definition of Done that meets this; each team adds to it for what it delivers, and never removes from it. (From 3.)
- Two kinds of evidence: whether it is used, and whether the problem is shrinking. The first is shown by meeting the Definition of Done; the second by the goal’s key results, the evidence measures agreed before the work starts. Read each on its own timescale, and over a run of periods rather than one. (From 2, 3 and 5.)
- Limit what runs at once. Start nothing without capacity set aside for it. (From 6.)
- Stop, adjust or continue, decided explicitly, at the end of the hypothesis’s timeframe or earlier when the evidence is clear. (From 1 and 2.)
- Record every decision: who made it and on what evidence, visible to everyone affected. (From 2 and 4, and the value of transparency.)
- Invite people, and treat their consent as evidence. People are asked whether they agree to a change in their own work. Their answer is evidence the decider weighs, and any disagreement is recorded with the decision. People are given problems, not solutions: a team may refuse a problem it shows to be invalid, and the refusal is recorded with its evidence. (From 4.)
- Measures are evidence, never quotas. (From 5, and Goodhart’s law: a measure that becomes a target stops measuring well.)
- Reach the system, and test the assumptions it runs on. (From 5.)
- Keep each change or investment small and bounded, so that being wrong costs a known, limited amount. (From 2 and 6.)
- A refuted hypothesis or a stop is a result, not a failure. (From 2.)
- Evidence is honest only where it is safe to report, and those with the money, and anyone with authority over how people are treated, are responsible for that safety. (People closest to the work report what they see only when doing so is safe, from 4; the system decides whether it is, from 5.)
- Accountability carries authority, placed by function, not rank. Those with the money decide which opportunities are invested in; those with the skills decide how the work is done; those who know the outcome best judge whether the evidence supports the hypothesis. Each decision is made openly by one accountable person or group holding that function, and the record names who holds each. Those with the money stay accountable for whether those they fund are ready to decide, and the authority to act is checked before starting. A change shown to work may spread to other teams by invitation: a team may take it up without anyone deciding to widen it, and records that it has; a change imposed on others, or one that affects them, stays with the decider. (Those closest see most, from 4; decision rights are part of the system, from 5.)
- Rank problems by evidence: how big each one is, who it affects, and what waiting costs. (From 1 and 6.)
- Starting one change or investment means naming which one stops or waits. Fixed obligations take their share of capacity first. (From 6.)
- Fund in stages. Continuing is itself a funding decision. A change or investment earns a larger stake as evidence arrives, and funding more is an explicit option. (From 2 and 6.)
- Prefer decisions that can be undone. Before deciding, say how the decision would be undone and what that would cost. A decision that cannot be undone needs stronger evidence, is made by someone with authority over its consequences, and is recorded as irreversible. (From 2 and 6.)
Core guides and practice guides
The principles are applied in two kinds of guide.
Core guides apply them to a kind of work:
- The Problem-Led Change Guide, for changing the system that delivers value.
- The Problem-Led Value Guide, for pursuing value people can use.
Practice guides apply them to a practice used within the core guides:
- The Evidence-Led OKR Guide, a problem-led guide to OKRs.
- The AI Adoption Guide, a problem-led guide to AI adoption.
The rule between them: a core guide may add to or narrow a principle, but may not contradict one. A practice guide may not contradict the principles or a core guide it is used within. Bringing every guide fully into line with this page is work in progress.
The engineering and leadership thinking the principles draw on is set out in my book, Engineering as a Leadership System.
Other practices fit within the principles without a guide of their own. None is required, and each must respect the principles. Scrum gives a delivery cadence inside a change or an investment, and an increment counts as done only when it is in use (derived principle 3). Kanban manages flow and the limit on work in progress (derived principle 5). The four Key Value Areas of Evidence-Based Management suggest what to measure (derived principles 4 and 9).
How to use these principles
To test a decision, rule or practice, name the principle it serves, then check it does not break another. A rule that serves no principle does not belong. A decision that breaks one needs a recorded reason or a different decision.
Then open the guide for your situation. If you are changing how your organisation works, start with The Problem-Led Change Guide. If you are deciding what to invest in, start with The Problem-Led Value Guide.
Terms used on this page
- Investment: work pursued to create value for the people an organisation serves. Its stake is a share of a team’s capacity, not a sum of money.
- In use: being used by the people it was for, so that its effect can be observed.
- Definition of Done: the standard work must meet before it counts. The organisation sets a minimum that requires work to be in use and producing evidence; each team adds to it.
- Reading evidence: looking at what the evidence shows, at the time set for it.
- A run: results across several periods, rather than a single one.
- The system: the structure, funding, decision rights, measures and assumptions that shape how work is done.
- Queue: work waiting to be started or finished.
- The limit: the maximum amount of work allowed to run at once.
Sources
These principles are by Martin Hinshelwood, NKD Agility. The values are drawn from the NKD Agility ethos.
© 2026 NKD Agility. This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License. You are free to share this material for any purpose, even commercially, provided that attribution is given and it is shared unchanged. Adapted versions may not be shared without permission.
They draw on the work of others. Each source below is credited as an influence, and is still being checked against the claim it supports.
- Complexity thinking, including David Snowden and Mary Boone’s “A Leader’s Framework for Decision Making” (Harvard Business Review, 2007), for the limits of prediction in complex work.
- Jeffrey K. Liker, The Toyota Way (2004), for going to see the work where it happens.
- W. Edwards Deming and Walter Shewhart, for the theory of variation and the system’s part in results.
- John D. C. Little’s work on queues (1961), for how work in progress affects waiting.
- Charles Goodhart’s observation on measures and targets.
- L. David Marquet, Turn the Ship Around! (2012), for giving control to the people doing the work, rather than taking it.
- Donald Reinertsen, on the cost of delay, for ranking work.
- The Kendall Framework, a problem-first methodology for enterprise AI, which informs The AI Adoption Guide.
- The Evidence-Based Management Guide (2020), for the four Key Value Areas.
- Martin Hinshelwood, Engineering as a Leadership System, for decision testability, including the reversibility of decisions.
Was this useful?
Enjoyed this? One click, no account.
Questions this guide answers
What are the problem‑led principles and how should I use them in my organisation?
The problem-led principles are a set of values, first principles, and derived principles for running work as a loop of: named problem, bet, something in use, evidence, and decision. They guide leaders and teams to start from problems, treat responses as testable bets, count work as done only when it is in use, limit work in progress, and make transparent, evidence-based decisions using core and practice guides such as the Problem-Led Change Guide and Problem-Led Value Guide. You use them by checking each decision or rule against the principles and then applying the relevant guide (change or investment) to structure your bets, evidence, and funding in stages.