In the Schedule Assessment Guide, GAO-16-89G, a baseline schedule is supposed to carry a buffer: extra time, which the guide calls schedule contingency, so the plan can absorb risk. Picture the shape they describe. The last real activity ends on a planned date. The finish milestone, the date you are willing to promise, sits later. The gap between those two dates is the contingency. When the last activity slips, the gap shrinks. That shrinkage is the contingency being spent. It is visible. It is not a secret stash of afternoons you feel slightly guilty about.
Calling that gap laziness confuses a reserve with idleness. Idleness is time with no job. Contingency has a job. Its job is to be consumed by the risks you already know are plausible, under a rule, instead of by whatever task feels urgent on a Tuesday.
Contingency is calculated, then guarded
GAO says the size of the gap should come from a schedule risk analysis, not from a mood. You compare the date in the ordinary schedule with the date from a simulation at the confidence level you are willing to promise. The guide's example: an organization that wants about an 80 percent chance of finishing on time sets contingency as the distance between that 80th-percentile date and the ordinary finish date. Some projects treat the 80th percentile as a conservative promise. Others accept 65 or 55 and, the guide notes, take more chance of an overrun. Project type, contract type, and how mature the technology is all change how much risk an organization will swallow. There is no universal "add 20 percent" in that section. If you do not have a simulation, do not invent a false percentile. Say you are using a simpler rule, and say what it is.
A simpler rule that still respects the shape: look at how late the last few similar deliveries were, pick a gap that would have covered most of them, and put that gap on the calendar as its own item named contingency, sitting in front of the promised date. The historical lateness is the poor cousin of a risk analysis. It is still better than a gap you plan to fill with extra scope because empty space makes you nervous.
The guide is strict about custody. Contingency is held by the project manager and released through change control, including when some of it is handed to a contractor or a partner. Using it should be traceable. If anyone can eat the gap without a note, you no longer have a reserve. You have a slush fund, and slush funds disappear into Parkinson's elastic work. The 1955 essay is about staff growth, not about your Friday. The narrower lesson GAO already states without Parkinson: dispersing the reserve in a careless way lets it get consumed early. The guide allows the reserve to sit as one block before the finish, or as several blocks before key milestones, for example before an integration that depends on outside inputs. It prefers that dispersion be deliberate so the time is not spent on the first loud request.
Float is not the same gap
Total float is the amount a predecessor can slip before the program finish moves. Activities on one path share it. GAO says, in so many words, that float should not be treated as schedule contingency. Float is a property of the network. Contingency is a reserve placed against prioritized risks. If you "use the float" to start a side idea, you have not spent contingency. You have tightened every later activity that shared that float, often without telling them.
The PMI lexicon, version 5.0, draws a related line with different nouns. A contingency reserve is time or money inside the schedule or cost baseline for known risks that already have response strategies. A management reserve is time or money set aside in addition to the baseline, released for unforeseen work that is still inside the scope of the project. Both are reserves. Neither is a character flaw, and neither is permission to add out-of-scope wishes. Unforeseen work inside scope is not the same sentence as a new feature. If the work is a new feature, it does not come out of either reserve until someone changes the scope on purpose. That decision lives on the scope page.
Where people hide the gap, and why the date then lies
Three common disguises:
- Padding each task in secret. Every activity grows a private cushion. The cushions are invisible, so they cannot be managed, and they often get used as start-late permission. GAO wants the reserve identifiable so its consumption shows up as a shrinking gap, not as a vague sense that things are fine.
- Padding nothing and calling the ordinary finish the promise. That is a promise at whatever low percentile the ordinary schedule happens to be. The guide's point about the 55th percentile is that it can be a choice, but it should be a choice you name, not an accident you discover during the apology.
- Advertising the gap as free capacity. Once other people can book meetings into it, the reserve is gone and the promise date is fiction. Protect it the way you would protect a milestone. It is a milestone. It is the milestone called "nothing, on purpose."
An illustrative sketch, not a project you can look up: a newsletter due to a printer on the 20th. The last editing activity, based on the last four issues, usually ends between the 14th and the 17th. You set the planned finish at the 17th and the promise at the 20th. The 18th and 19th are a contingency activity named "printer gap," owned by you, released only if a late draft or a bad link appears. They are not a slot for a new column. If the 17th holds, the gap remains and the printer still gets the file on the 20th, or earlier if you choose to spend the reserve on an early delivery. Early delivery is a decision. Filling the gap with a new column is scope, and it deletes the protection you just built.
Laziness is a story about motive. Contingency is a story about risk.
You can be lazy and also have no buffer. You can be diligent and have a buffer because the last four times reality arrived late. Motive is a bad measuring tool. The measuring tool is whether the gap is sized from evidence, named on the schedule, and spent only when a listed risk shows up. If you spend it because you started late, that is not the reserve working. That is the reserve masking a start problem. Record the difference so the next gap does not grow for the wrong reason.
Shared work makes the custody rule matter more. If a partner can slip into your printer gap without asking, your 80th-percentile promise silently became their storage. The handoff page, working with other people's deadlines, is about whose slip moves the finish. This page is about the hours you refuse to pretend are spare. The estimating page, estimating without lying, is how you stop building the ordinary finish out of best-case durations and then acting surprised that the gap vanished.
Limits of a government scheduling guide
GAO-16-89G was written for capital programs and auditors. A two-day personal buffer is the same shape at a much smaller scale, not a claim that you performed a Monte Carlo simulation. Do not cite an 80 percent chance unless you actually ran an analysis that produces percentiles. PMI's two reserves are glossary terms for projects, not tax advice and not a ruling on whether your employer owes you idle time. If a contract defines liquidated damages or a service level, the buffer you keep internally does not rewrite that contract. Ask someone qualified before you promise a date you do not control.
The moralized version of this topic says unused time means you did not care. The schedule version says unused contingency means the risks did not fire. Those are different sentences. Which of your promised dates still has a named gap in front of it, and who is allowed to spend it?
Sources for the gap
- U.S. Government Accountability Office. (2015, December 22). Schedule Assessment Guide: Best Practices for Project Schedules (GAO-16-89G). Schedule contingency and the warning not to treat float as contingency. gao.gov/assets/gao-16-89g.pdf
- Project Management Institute. (2026). PMI Lexicon of Project Management Terms (Version 5.0). Contingency reserve and management reserve. PMI lexicon PDF