Skip to content
Find your tool
Guides · 6 min read

Why a percentage on top is not a risk budget

Ten percent on top sounds prudent. On two fictional infrastructure projects it holds in only about a third of outcomes, and each needs a different amount. A percentage cannot tell you that; a simulation can.

By Mike Bouman, Quantil · November 4, 2026

"Add 10% for contingency." It's the most common sentence in project estimating, and it sounds prudent. The trouble is what it leaves out: how likely is it that the budget will hold?

A percentage answers the wrong question

A project board doesn't really ask "how much should we add?" It asks "how sure can we be that this budget holds?" A flat 10% or 20% can't answer that, because the same percentage gives very different confidence on different projects. On a repeat project with a well-defined scope, 10% may be generous. On a first-of-a-kind tunnel, it may be far too little.

The percentage doesn't look at the project at all. It doesn't know which items are uncertain, which risks could hit, or how far costs can run. So when an auditor or a board member asks what the reserve is based on, the only honest answer is "convention".

What 10% buys on two fictional projects

The way to find out is to simulate the project. Each cost item and each risk gets a range instead of a single number, and the project is run thousands of times. The result is a distribution of possible outcomes, so you can read off what share of those futures stays within a given budget.

Fictional caseBase estimateBase + 10%Chance the budget holds
Riverside High-Voltage Grid Construction (the simulator on our homepage)$120.0M$132.0M33%
Cascade Highway Tunnel (QRCIM reference case)$416.0M$457.6Mabout 29%

On both projects the rule falls short. On the grid project, 10% on top holds in only about one in three futures; reaching 80% confidence takes $20.5M more, 27% on top instead of 10%. On the tunnel, 10% on top leaves a chance of overrun of more than 70%, before the project even goes to market. Even the tunnel's full deterministic budget, with class-based allowances per item, reaches only about 40% confidence. And the right amount differs per project: the grid project needs 27% on top for P80, the tunnel 22%. No single percentage fits both.

You can check the first row yourself: run the simulator on the homepage with its default values, then change a probability and watch the confidence move.

Why the base estimate sits low

Most people expect the base estimate to sit somewhere in the middle of the possible outcomes. It usually does not. Cost risk is skewed: an item can turn out much more expensive than planned, but rarely much cheaper. If you add up the most likely value of every item, you get a total that real projects exceed more often than they undershoot it.

In the tunnel case, the simulation reaches the base estimate of $416M or less in only about 3% of the runs. The mean outcome is $478.2M, more than $60M higher, without any change in scope. A percentage on top tries to cover that gap blindly. A simulation shows how big the gap actually is.

A practical tip. The 10% is rarely anyone’s estimate. It’s a habit, sometimes copied from the previous project. Run a first simulation on the register and compare: the gap between the 10% and the P80 is seldom small, and it differs a lot from one project to the next. That’s the whole point: one percentage can’t fit projects with very different risks.

What to use instead

Replace the percentage with a confidence level. Instead of "base plus 10%", the budget becomes "the P80": the amount the project stays under in 80% of the simulated futures. That changes the conversation in three ways.

A common misreading. P85 doesn't mean "base plus 85%" or any fixed percentage. In the tunnel case, P85 happens to be 23.9% above the base. On a better-defined project, the same P85 could be 8% above it. The distance follows from the shape of each project's distribution.

Is a percentage ever fine?

Yes, in two places. Very early, at screening stage, there may not be enough information for meaningful ranges, and a class-based allowance is an honest placeholder. And within a probabilistic budget, a fixed allowance per item still has a role for known unknowns in the estimate itself. What a percentage shouldn't be is the whole answer to "how much reserve do we need?" once the project has a real risk register.

Getting started

You don't need specialist software to make this change. If your risks already live in a register, the Cost & Schedule Risk Analyzer simulates it and returns the reserve in money and in weeks at the confidence level your board uses. If you want the whole budget build-up, with cost items, risk events and correlation, QRCIM takes the estimate from base to a structured, auditable budget. The next guide walks through that process: how to set a P80 contingency. To take the result to your board, read why boards need ranges.