An owner asks for an early estimate of a factory fit-out. The person preparing it says “about $380,000”, and if asked would add that it could plausibly land anywhere between $300,000 and $480,000, because the design is not finished and the building has not been surveyed. Nobody asks. The figure goes into a loan application as $380,000, then into the budget, and a year later the conversation is about why the job came in at $450,000.
Nothing improper happened. The estimate was competent and the budget was reasonable. What went missing was the range, and with it the only honest signal about how much anyone actually knew. There was never a meeting at which someone decided to turn a range into a commitment. It happened through formatting: a column was dropped as the number moved from one document to the next.
This article explains why estimates should carry their method and range, how to make the move from estimate to commitment a deliberate decision, and how to set rules for the money held back for surprises, separating contingency for expected problems from a reserve for changes that alter the job. It is general information; your accountant can help with financing and budgeting decisions for significant projects.
Accuracy depends on definition
Estimating guidance commonly describes how accuracy improves as a project becomes better defined. An order-of-magnitude figure at the concept stage might be expected to be within about 30%; a preliminary estimate within about 20%; estimates for well-defined work within 5 to 10%. The exact figures vary by industry and source, but the pattern is consistent: ranges are wide early and narrow as the design, site and scope become known.
Two points follow:
- Accuracy comes from definition, not effort. No amount of estimating skill closes a concept-stage range, because the information does not yet exist.
- A single early number borrows confidence it does not have. The reader supplies the certainty the estimator never claimed.
The perverse result in many businesses is that the most careful estimator, the one who has thought hardest about what is unknown, produces the least confident-looking figure and loses the argument to someone with a rounder number and less analysis.
Every number should carry three things
No estimate should reach a decision without:
- Its method. Was it based on one similar past job, a rate per square metre, supplier quotes, or a firm tender price? These are very different kinds of claim. Two options compared side by side may look equivalent when one is a firm quote and the other is someone’s recollection of a job five years ago.
- Its basis. What scope, assumptions and exclusions does it rest on?
- Its range. What is the plausible low and high, and what would drive each?
Where a range genuinely cannot be given, say so. That is itself useful information.
Estimating and committing are different acts
An estimate is a technical judgement about what something is likely to cost. A commitment is a statement that the business will be held to a figure: in a budget, a loan, a contract price or a promise to a customer. Turning one into the other involves choices:
- where in the range to commit;
- how much contingency to hold, and who controls it;
- what probability of exceeding the figure the business is willing to accept.
Each of those is a judgement. In a well-run business each is made deliberately, by a named person, on a known date. In many businesses they are made by copying a number into a different document.
There are usually four moments when the number can harden: when the estimator hands it over, when it first appears in a business case or loan application, when the budget is approved, and when a firm price is obtained from the market. Many businesses let it harden at the second, in a step nobody regards as a decision, and then simply confirm it at the third. A better habit is to decide which moment is the commitment and to record who made it, where in the range they chose to sit and why.
Contingency and reserve are different funds
Most projects hold some money back for things that go wrong. Larger organisations often distinguish two kinds:
- Contingency sits inside the project budget and covers identified risks of an anticipated type: ground conditions worse than expected, an inspection finding, a delivery delay. The person running the job should be able to use it quickly, within agreed limits, because stopping for approval every time would cost more than the problem.
- Reserve sits outside the project budget and covers events that change the job itself: a new regulatory requirement, a supplier withdrawing a product, a customer changing what they need. Using it is a decision to change the project, and it belongs with whoever approved the project in the first place.
They need opposite properties. Contingency should be fast and delegated. Reserve should be slower and kept with the owner or decision-maker. Put them in one pot and speed wins: the immediate, concrete delivery problem uses the money, and when something genuinely unexpected happens later, there is nothing left.
Rules for each fund
A useful set of rules answers five questions for each fund:
| Question | Contingency | Reserve |
|---|---|---|
| What may it pay for? | Identified risks of an anticipated type | Events that change the job’s scope or premise |
| What may it never pay for? | Scope additions, estimating errors, price rises that should be handled separately | Changes a customer or supplier should pay for |
| Who can release it? | The person running the job, up to a set amount | The owner or whoever approved the project |
| What evidence is needed? | A listed risk that has actually occurred | A written change covering both scope and budget |
| Who tops it up? | The owner, after reassessing the remaining risks | The owner, as a new decision |
The exclusions matter most. A contingency with no exclusions is unallocated budget with a respectable name.
How contingency is often sized, and the trap in it
A common method multiplies each identified risk’s probability by its cost and adds up the results. That gives the expected cost of the risk list: the average across many similar projects. It is a reasonable guide for an organisation running dozens of similar jobs. A single project runs once. It will experience some of its risks and not others, and the total of the averages describes an outcome it will almost certainly not have.
Two other traps are common:
- Double counting. If a risk has been treated, for example by paying for a survey or a second supplier, that treatment has its own cost in the budget. If contingency is still calculated on the untreated risk, the same event is funded twice.
- Quiet drawdown. Contingency used without evidence that a listed risk actually occurred is not risk being realised. It is the range being spent unnoticed. Release should require the same kind of evidence as the original approval: what happened, and what is now known.
The near-certain problems need a budget article covers another gap in contingency: items so likely they are removed from the risk list altogether.
Learn from your own outcomes
Most businesses never compare their early estimates with what jobs actually cost. Doing so, even roughly, across ten or twenty completed jobs is instructive. It shows which estimating methods work in your hands, which kinds of work are consistently underestimated, and how wide your concept-stage ranges really are in your business rather than in a textbook. That replaces an argument about optimism with evidence.
A worked example
This is an illustration. A small manufacturer plans to fit out a leased building for a new production cell. The first estimate is $380,000, based on one similar fit-out the owner did three years earlier. The person who prepared it says the plausible range is $300,000 to $480,000, because the building’s power supply and floor condition are unknown.
The owner records the method and range in the loan application rather than a single figure, and asks the bank to approve finance on the basis of a range with a firm decision point once quotes are in.
Two identified risks drive most of the uncertainty:
- Power upgrade: about a 40% chance it is needed, costing about $50,000.
- Floor levelling: about a 50% chance, costing about $24,000.
The usual method gives a contingency of $32,000: 40% × $50,000 plus 50% × $24,000. But the project will face one of four actual outcomes:
| Outcome | Probability | Extra cost |
|---|---|---|
| Neither needed | 30% | $0 |
| Floor only | 30% | $24,000 |
| Power only | 20% | $50,000 |
| Both | 20% | $74,000 |
A $32,000 contingency covers the outcome only 60% of the time; $50,000 covers it 80% of the time. The owner chooses instead to buy information: an electrician’s assessment and a floor survey cost about $3,000 together and resolve both questions before committing. The survey finds the floor needs levelling; the power supply is adequate. The uncertainty on those items disappears, and the contingency can be sized for what remains.
Quotes then come in. Firm prices cover about 80% of the scope at $412,000, with the remaining items estimated at about $38,000. The owner commits to $450,000, records that decision, and sets two funds:
- Contingency of $25,000, controlled by the project lead, for listed risks such as delivery delays and minor building issues. Each use is logged against the risk it covers.
- A reserve of $30,000 held by the owner, to be used only for changes to the job itself, such as a customer requirement that alters the cell’s layout, and only with a written change covering scope and budget together.
At the end, the owner compares the original $300,000 to $480,000 range with the outcome and records which assumptions proved wrong, for the next estimate.
How this applies to a small Australian business
- Ask for the method and range with every significant estimate.
- Do not let a range become a single figure without a decision.
- Decide when you commit, and record who did, where in the range and why.
- Separate contingency from reserve, even in a simple spreadsheet.
- Write down what each fund can and cannot pay for, and who can release it.
- Check for double counting between risk treatments and contingency.
- Buy information where a small survey or test can resolve a large uncertainty before you commit.
- Compare estimates with outcomes on completed jobs.
Signals worth watching
- Estimates presented as single numbers with no method.
- Firm quotes and rough guesses compared as equals.
- Contingency used without reference to any listed risk.
- One pot of “spare money” with no rules.
- Budgets that only ever grow at the end of a job.
- No comparison of estimates with actual costs.
Common mistakes
- Dropping the range as the number moves between documents.
- Treating an estimate and a commitment as the same thing.
- Sizing contingency as a sum of averages for a one-off job.
- Mixing contingency and reserve in one fund.
- Funding a treated risk twice.
- Never checking estimating accuracy against outcomes.
Frequently asked questions
Won’t a range make a lender or customer nervous? Sometimes. But a single number that later proves wrong is more damaging. Many lenders and customers respond well to a clear range, a decision point and a plan to narrow it.
How large should contingency be? It depends on the risks that remain and how much certainty you want. Consider the actual combinations of outcomes, not just the average.
Is a reserve necessary on small jobs? Not always as a separate fund. But the principle still applies: changes to what the job is should be decided by the owner, not absorbed quietly.
What if we are quoting a fixed price to a customer? The same thinking applies in reverse. Your estimate has a range; your price is a commitment. Decide deliberately where in the range to price and how much risk to carry. The fixed price or not article covers choosing a pricing model.
How do we start comparing estimates with outcomes? Pick your last ten completed jobs, note the first estimate and the final cost, and look for patterns.
Questions to ask
- What method and range lie behind this estimate?
- When did we commit to this figure, who decided, and where in the range did they choose to sit?
- What can our contingency pay for, and what can it not?
- Do we hold any money back for changes to the job itself, and who controls it?
- Have we funded any risk twice?
- How accurate have our estimates been on past jobs?
Bringing it together
Every serious estimate has a range, and the range is the most honest information it contains. Keep the method, basis and range with the number until someone deliberately decides to commit, and record that decision. Hold money back in two forms with different rules: contingency for expected problems, released quickly by the person running the job, and reserve for changes to the job itself, released only by whoever approved it. Size contingency for the outcomes you could actually face, not the average, and buy information where it can narrow the range cheaply. Then compare estimates with outcomes, so each new estimate is better grounded than the last.
Source: KEVOS notes, drawing on teaching material on project cost estimating, budgeting and reserve analysis, on earned value practice for management reserve, and on G. Winch, A. Usmani and A. Edkins (1998) on estimating ranges during design development, Construction Management and Economics. Examples and figures in this article are illustrations. This article is general information, not financial advice.
