Project crashing is what you do when the date is wrong and the work is not going to shrink on its own: you pick tasks and do them a different, faster, more expensive way. Rent the spray rig. Order the wall panels prefabricated instead of stick-framing them. Hand the drywall to a subcontractor. The version in the textbooks is a tidy table of crash cost per day, and in twenty years of doing this for a living I have never once had a job hand me that table.
So I ran it properly instead of describing it. Same 21-task house renovation I have been using all series, three general crew, six alternatives a real builder could actually buy, and a solver that has to choose which ones are worth it. Here is what crashing bought, change by change, with nothing rounded in my favour.
What crashing is, and what it is not
Crashing means changing how a task gets done so that it takes fewer days. The task stays where it is in the network, the dependencies stay exactly as they were, and the duration drops because you paid for something: more bodies, better equipment, or somebody else’s crew entirely.
Two things it is not. It is not fast tracking, which leaves durations alone and overlaps tasks that used to run one after the other — a different lever with a very different result, which I measured separately and will come back to. And it is not just adding people. On most jobs you cannot conjure a fourth carpenter on Tuesday. What you can do is take a whole task off your crew and give it to someone who arrives with their own.
The mathematics is old. James Kelley set out the mathematical basis of critical-path planning in Operations Research in 1961, and the duration-versus-cost trade-off is in there as a linear programme: assume each task’s cost rises in a straight line as you squeeze its duration, and you can compute the cheapest way to hit any date. It is elegant, and it rests on an assumption that quietly fails on site — that time is available in continuous increments, at a constant price, in any quantity you like. Real alternatives are not a slider. They are a short list of discrete, lumpy options, and most tasks have none at all.
The plan before anybody touches it
Twenty-one tasks, five trades, three general crew, two electricians, two plumbers, two painters and one excavator. Run the critical path arithmetic on it and the answer is 25 November 2026. Schedule the same work so no day ever asks for more than three general crew and it lands on 2 December — five working days later, none of which appear anywhere in the CPM output, because the network was never told how many people exist.
Hold on to those two dates. 25 November is what the plan says. 2 December is what the job does. Everything below is measured against 2 December, because that is the one you would actually miss.
Six other ways of doing something
I gave the plan six alternatives. Each one is a different duration and a different daily crew for a task that already exists — not a new task, not a shortcut, just a second way of doing the same job.
| Task | Normally | The other way | Crew/day |
|---|---|---|---|
| Framing | 10 days, 3 crew | 6 days — prefabricated wall panels | 3 |
| Roofing | 5 days, 3 crew | 4 days — roofing subcontractor | 0 |
| Drywall | 7 days, 3 crew | 5 days — drywall subcontractor | 0 |
| Exterior painting | 6 days, 2 painters | 3 days — spray rig instead of rollers | 2 |
| Driveway | 5 days, 2 crew | 3 days — concrete subcontractor | 0 |
| Landscaping | 6 days, 1 crew | 4 days — landscaping contractor | 0 |
What each one buys on its own
Before letting the solver combine them, I ran each alternative as the only one available, six separate solves. This is the number a foreman actually wants: if I make exactly one phone call, which call is it?
| The one change | Finish | Working days saved |
|---|---|---|
| Drywall subcontracted | 25 Nov | 5 |
| Prefabricated framing | 26 Nov | 4 |
| Roofing subcontracted | 27 Nov | 3 |
| Driveway subcontracted | 30 Nov | 2 |
| Landscaping contracted out | 30 Nov | 2 |
| Exterior painting on a spray rig | 2 Dec | 0 |
The spray rig is the row worth stopping on. It is the biggest proportional saving on the whole list — six days down to three, a task cut in half — and it moves the finish date by nothing at all. Exterior painting is not on the chain that determines the end of this job, so making it faster changes when the painters go home and nothing else. Any tool that ranks your options by how much time each one saves on its own task will put that at the top of the list. It belongs at the bottom.
And notice which change wins. Subcontracting drywall takes two days off a seven-day task, which is less than the framing alternative takes off framing. It still buys more, because it takes three crew off the drywall days entirely, and this job’s problem is not the length of its tasks — it is that too many of them want the same three people at once. That is exactly the gap between the two dates: 25 November is the plan with the crew shortage removed, and one subcontract removes it. The single best crash on this job cancels the resource problem precisely.
Three changes, eleven days — not twelve
Let the solver have all six and it takes three: prefabricated framing, the roofing subcontractor and the drywall subcontractor. Finish 17 November. That is eleven working days off 2 December, and it leaves the driveway, the landscaping and the paint alone.
Those three, individually, were worth 4, 3 and 5 days. Add them up and you get twelve. Together they deliver eleven. Nothing went wrong — the missing day is what it looks like when savings overlap, because two of those changes were relieving the same crowded stretch of the calendar and the second one arrives to find the first has already fixed part of it. Days saved are not additive, which is the whole reason to solve this rather than rank it. A spreadsheet that sums a column of savings will always promise you a date you do not get.
The staircase: how much does each date cost you?
Better question than “how fast can we go” is “what does this date cost me”. So I gave the solver a deadline and asked it for the fewest changes that hold it — the fewest arguments with subcontractors, since that is the currency here. Walking the deadline back a day at a time produces a staircase with four steps and a wall.
| Date you promise | Changes needed | Which ones |
|---|---|---|
| 2 December | 0 | — |
| 25 November | 1 | drywall |
| 19 November | 2 | drywall, framing |
| 17 November | 3 | drywall, framing, roofing |
| 16 November or sooner | Impossible with these six options | |
Read it as a price list. Getting back to the date your CPM chart promised costs one phone call. The next six days cost one more. The two days after that cost a third. And below 17 November there is nothing to buy at any price — you would need a different alternative on the list, not a bigger budget. Being told a date is impossible, and told it in half a second, is worth as much as being told it is achievable.
Why there is no money in this model
You will have noticed the tables have no dollar column, and the classic formulation is all dollars — crash cost per day, cost slope, cheapest path to the target date. I left money out deliberately, and I want to be straight about the trade.
What I have never been able to get from a site team is a credible cost slope. What they can tell you instantly is how many people a job takes and who is available. So the scarce thing in this model is people, the crew caps do the pricing, and the objective is to make the fewest changes that hold your date. A change is a change: renegotiating a scope, calling a sub, committing to a delivery. Ranking by count rather than by an invented cost curve is a smaller lie.
The cost of each change is the one thing you genuinely know and I do not. The tool narrows six options to three and tells you which; whether the drywall sub’s quote is worth five days is your call, and it should be.
The other lever, measured
Crashing is one of two standard ways to compress a schedule. The other is fast tracking — overlapping tasks that were planned end to end. It is usually presented as the cheaper option, since it costs nothing but risk. I ran that on the same plan with the same solver, and the comparison did not go the way the textbook order implies: fast tracking pulled eight working days off the plan and one off the job.
Running it on your own plan
The free scheduler takes the same workbook as the rest of this series — a task list with durations, predecessors and a crew per day, and a resource sheet with what you actually have. Two more things and it will crash it:
- Fill in the alternatives sheet. One row per option: task ID, the shorter duration, and the crew per day it needs. Leave the resource column blank to use the same trade as the task row. Put 0 in crew per day for a subcontractor who brings their own people — that zero is what makes subcontracting different from hurrying.
- Set Mode on the parameters sheet to
crash. Leave the deadline blank and it goes as early as the options allow. Give it a date and it finds the fewest changes that hold it, or tells you the date is impossible.
Most plans need two or three rows on that sheet, not twenty. You already know which tasks have a faster expensive version, because they are the ones you have argued about before. If you have not built the underlying plan yet, the construction schedule template with crew sizes is the same workbook with the trades already filled in, and the critical path template covers the float side of it.
One last thing worth saying about all of this. Every number above came back proved optimal in about a tenth of a second, which means the solver is not offering an opinion about whether three changes is enough — it has established that two cannot reach 17 November and three can. That is a different kind of answer from the one a priority rule gives you, and on a job where each change is a real conversation with a real subcontractor, knowing you are not making a fourth one unnecessarily is most of the value.
Things that I use, like, and am affiliated with:
Mint Mobile offers great cell phone service for $15 flat, get $15 off using the link. Get discounted phones with service activation and no contract.
I never spend money before I check Mr Rebates or Rakuten to get cashbacks, rebates, discounts, coupons or cheaper gift cards.