Two arguments are often made about the same new initiative, sometimes by the same person in the same meeting. The first is that it cannot be planned properly because too much is unknown. The second is that it cannot be funded properly because nobody has planned it. Both sound prudent. Together they produce a familiar stalemate in which the most important and uncertain work a business takes on gets the least disciplined thinking.
The stalemate is not really caused by uncertainty. It is caused by an assumption about what a plan is for. If a plan is a forecast, a prediction of what will happen, then genuine uncertainty does make planning pointless, and the first argument wins. If a plan is something else, the stalemate dissolves. Project management writer Stanley Portny, writing for scientists in 2002 about research whose outcome cannot be known, argued that planning should start at the very beginning of such work, because uncertainty is the reason to plan rather than the obstacle.
This article explains a different job for plans under uncertainty: detecting early when beliefs are wrong. It covers the common ways planning goes wrong on uncertain work, how to find the few assumptions that carry most of the value, how to stage commitment so that learning comes before large spending, and how to set clear triggers for stopping, changing course or scaling up.
A plan that detects, not one that predicts
For routine work, a plan is a reasonable prediction. A competent team can describe the result, estimate the cost and be held to it. For uncertain work, such as a new product, a new market, a new technology or a major change in how the business operates, a detailed prediction is mostly guesswork dressed up as commitment.
Under uncertainty, a plan does a different job. A plan is an instrument of detection, not prediction. Its value is not that it tells you what will happen. It is that it writes down what you currently believe, so that when reality turns out differently, the difference is visible early and to everyone, rather than late and only to the people closest to the work.
This changes what a good plan looks like. A predictive plan is judged by how closely results match it. A detecting plan is judged by how quickly it shows that they do not. These are different design goals and produce different documents: shorter, clearer about assumptions and specific about what will be watched.
A plan also helps people work together when much is unclear. The less certain the situation, the more it matters that everyone involved shares the same understanding of where things stand, where they are heading and why. That shared understanding is what allows a team to act quickly on early information.
Three ways planning goes wrong on uncertain work
Treating the plan as a promise. When a plan is seen as a commitment to a specific future, every revision looks like failure. People defend plans long after they stop being useful, because changing them carries a personal cost. The business loses the adaptability the plan was meant to provide.
Confusing detail with rigour. A hundred-line schedule for work nobody understands is not more rigorous than a ten-line one. It is less honest, because it spreads false precision across more lines. Rigour under uncertainty means being clear about what is known, what is assumed and what is being watched.
Leaving the definition of success until later. Businesses often commit money to initiatives whose definition of a good outcome is left to be settled once things become clearer. They rarely do become clearer. Instead, success is defined by whoever is most persuasive at the point the question can no longer be avoided, usually after the money is spent and the options have gone.
Exempt uncertain work from forecasting, not from thinking
Because detailed forecasts do not suit uncertain work, many businesses exempt such work from planning altogether. Research, new ventures and anything labelled innovation are allowed to proceed with light oversight. The intention is reasonable: it seems unfair to hold exploratory work to a forecast it cannot meet.
But the exemption is usually applied to the whole discipline of planning, when it should apply only to forecasting. Removing planning from uncertain work does not free it. It removes the one mechanism that would show, early and cheaply, whether the work is going anywhere.
The front-end questions
Before committing significant money to uncertain work, answer five questions:
- What result would justify this investment, stated clearly enough that you would recognise it if it arrived?
- Whose view of that result decides, and whose only informs it?
- What work does this need, who would do it, and do they have the capability?
- What else must we have or acquire for this to be possible?
- What could reasonably turn out differently, and what would it cost us if it did?
None of these requires knowing the future. All of them require deciding what you currently think, which is far more achievable.
The front end is also where changing direction is cheapest. Early on, a change of direction costs a conversation. Later it costs contracts, equipment, hiring decisions and credibility. Putting off these questions does not keep options open. It uses up the period when options are cheap without getting anything for it.
Find the load-bearing assumptions
Every plan for uncertain work contains a chain of beliefs: customers will respond, the capability can be built, suppliers will perform, people will adopt the change, the costs will hold and competitors will not quickly copy the idea. The plan may look certain even though everything in it depends on these conditions.
Not every assumption deserves equal attention. A load-bearing assumption is one that is both uncertain and capable of destroying the value of the plan if it is wrong. Common examples include how many customers will buy and at what price, unit costs, whether a process can reach the required quality, regulatory approval, supplier capacity and whether a new technology will work with existing systems.
A simple assumption register helps. For each important assumption, record:
| Field | Question |
|---|---|
| Assumption | What are we believing? |
| Importance | If this is wrong, does the plan still work? |
| Evidence | What do we actually know, and how confident are we? |
| Test | How could we find out before committing the full cost? |
| Trigger | What result would change our decision? |
| Owner | Who is watching this? |
Focus testing effort on the assumptions that are both important and uncertain. Teams naturally test what is easy to measure rather than what is decisive. Ask which single assumption, if wrong, would most damage the plan, and test that first.
Commit in steps that follow learning
Large commitments should follow the reduction of important uncertainty where possible. A pilot, prototype, supplier trial, limited launch or staged investment can turn a belief into evidence before the business crosses a threshold that is hard to reverse.
This changes the conversation from “approve or reject?” to “what is the next justified commitment?” Staging can make progress look slower, but it speeds up learning and protects money. The article on risk as the opportunity you miss discusses how to separate the decision to learn from the decision to scale.
Test the plan against more than one future
A plan does not need to succeed in every possible future, but the business should understand how it behaves under plausible changes in demand, cost, timing, regulation, technology or competitor response. Ask what happens to the plan if sales are half the forecast, if a key cost rises by a third or if the launch slips six months. A plan that only works under one forecast is fragile, however good its expected return looks. Robust plans keep reasonable value across several futures, or include affordable ways to adapt.
Decide your triggers in advance
Managing assumptions is incomplete until the business decides what evidence will cause action. A trigger connects an observed result to an agreed response: stop, change course, continue as planned or invest more.
| Trigger type | Example |
|---|---|
| Stop | Fewer than a third of trial customers reorder within a month |
| Change course | Customers reorder but delivery costs are too high |
| Continue | Results within the expected range; next review in a set period |
| Scale | Reorders and margins above agreed thresholds |
Deciding triggers before results arrive matters, because once money has been spent, people find ways to reinterpret bad news. Agreeing in advance means accepting that a favoured plan may need to change.
Separate the case for investing from the estimate of delivery
A business case usually does two jobs at once. It argues why the work deserves money compared with alternatives, and it estimates what the work will take and when. Combining them puts pressure on teams to invent precision in the estimate in order to win approval for the argument.
Separating them helps. Approve uncertain work on the strength of its logic and its plan for learning, while being open that its estimates are provisional and will be refined at agreed points. This produces more honest estimates and earlier warning when they change.
Make revising the plan normal
None of this works if revising a plan is treated as failure. A business that punishes changed plans will get stable plans and unstable results. The capability worth building is not better forecasting. It is faster, less defensive decision-making when evidence changes.
Practical habits include reviewing assumptions at every significant review, recording what has changed and why, and praising people who bring early evidence that a plan is wrong. The bad news early article looks at how a business teaches people whether it is safe to do that.
A readiness test before funding
Before committing significant money to uncertain work, check six things:
| Test | Question | Ready when |
|---|---|---|
| Recognisable result | Would we know the intended result if it arrived? | Two people would independently agree whether it had happened |
| Authority | Who may change what this initiative is for? | The answer is written down |
| Beliefs | What must be true for this to work? | Assumptions are written down and separated from facts |
| Detection | How would we learn we are wrong? | A way of detecting it exists and someone owns it |
| Reversibility | What does this commitment rule out? | The options it removes are listed |
| Exit | What would make us stop? | Stop conditions are defined before funding |
Work that fails the recognisable-result or exit test should not receive full funding. It may deserve a small, bounded amount to resolve the uncertainty, but that is a different decision and should be described as one.
A worked example
This is an illustration. A commercial bakery supplying bread to local cafés wants to launch a range of frozen, part-baked pastries that cafés can finish in their own ovens. The owners are enthusiastic but uncertain. The full launch would need a blast freezer costing about $85,000, new packaging and a refrigerated delivery arrangement.
Instead of a detailed five-year forecast, the owners write a short plan built around their assumptions. They identify four load-bearing ones:
- Demand: cafés will reorder regularly at a price that gives an acceptable margin.
- Freight: frozen delivery can be done for no more than 12% of revenue.
- Quality: the products will still be good after three months in a café freezer.
- Capacity: production can fit around existing bread production without overtime.
They design an eight-week trial with 15 existing café customers, using a hired freezer and a refrigerated courier, and set triggers in advance:
- Scale: at least 9 of the 15 cafés reorder within four weeks, and freight is no more than 12% of revenue. Then buy the blast freezer.
- Change course: reorders are strong but freight or quality falls short. Then redesign the weak part and retest.
- Stop: fewer than 5 cafés reorder. Then end the trial.
A shelf-life test runs alongside, and the production manager records the hours used.
After eight weeks, 11 of the 15 cafés have reordered, and trial revenue is $18,000. Quality holds, and production fits within existing hours. But freight has cost $2,880, which is 16% of revenue, above the 12% threshold. Under the agreed triggers, this is a change-of-course result, not a scale result.
The owners redesign delivery: frozen orders are consolidated into one weekly run instead of deliveries on demand. A further four weeks shows freight would have been about $1,980 on the same revenue, or 11%. With all four assumptions now supported by evidence, the owners buy the blast freezer. Their original enthusiasm would probably have led to the same decision, but with a delivery model that would have quietly eroded the margin.
How this applies to a small Australian business
Small businesses often take on uncertain work with little formal planning, or with a forecast nobody believes. Practical steps:
- Write a short plan for uncertain work that states what you believe, not what you predict.
- Answer the five front-end questions before committing significant money.
- List the load-bearing assumptions and test the most important one first.
- Stage commitments so learning comes before large spending.
- Check the plan against several futures, not just the expected one.
- Set stop, change and scale triggers before results arrive.
- Separate the case for investing from the estimate of delivery.
- Treat revised plans as good management, not failure.
The when the project succeeds and the strategy fails article looks at how to keep testing whether a project still deserves its place after it has started.
Signals worth watching
- Plans for uncertain work that never change, which suggests nobody is using them.
- Bad news arriving fully formed and late.
- Assumptions made at approval that have changed without anyone revisiting the decision.
- The purpose of an initiative drifting without a decision.
- Estimates for very different pieces of work showing suspiciously similar confidence.
- Negative evidence producing discussion but no action.
- Commitments locking in faster than uncertainty is falling.
Common mistakes
- Skipping planning because the future is uncertain.
- Writing detailed forecasts for work nobody understands.
- Leaving success undefined.
- Testing the easy assumptions instead of the decisive ones.
- Making large commitments before learning.
- Setting triggers after results arrive.
- Treating revised plans as failure.
Frequently asked questions
How long should a plan for uncertain work be? Often one or two pages: the intended result, the load-bearing assumptions, the tests, the triggers and the next review date. Length is not a measure of quality.
What if we cannot test an assumption before committing? Look for a partial test, such as a smaller version, a proxy measure or a conversation with people who have tried something similar. If no test is possible, reduce the size of the commitment or make it easier to reverse.
Who should set the triggers? The person who will make the decision, ideally with input from someone who is not the initiative’s main champion. Record them before results arrive.
What if the result falls between triggers? Treat it as a signal to look closer, not as permission to continue indefinitely. Agree a short further test and a date for the decision.
Does this apply to small decisions? Scale the effort to the consequence. A quick list of assumptions and one stop condition is enough for small initiatives. Large or irreversible ones deserve the full approach.
Questions to ask
- For our largest uncertain initiative, what observation in the next 90 days would make us change course, and who is watching for it?
- Which assumption carries the greatest share of the expected value?
- Where have we exempted work from forecasting, and did we accidentally exempt it from thinking?
- What would make us stop, and is it written down while stopping is still cheap?
- Which commitments have closed off options we did not consciously decide to give up?
- When someone revises a plan here, is that treated as good management or as a problem?
Bringing it together
Uncertainty is the strongest reason to plan early, as long as the plan is designed to detect rather than to predict. Write down what you believe, find the assumptions that carry the value, test the decisive ones first, stage commitments behind learning, check the plan against more than one future and agree stop, change and scale triggers before results arrive. The test of a plan under uncertainty is not whether events matched it. It is how quickly the business found out they had not.
Source: KEVOS notes, drawing on writing by Stanley Portny on planning uncertain work (2002) and on teaching material on strategic assumptions, staged commitment and decision triggers. Examples and figures in this article are illustrations.
