Resource leveling vs resource smoothing is usually taught as a pair of definitions to memorise for an exam: leveling may push the finish date out, smoothing may not. Both true, both useless the moment you are standing in front of a schedule that needs fixing, because neither tells you which one your plan can actually give you.
They are not two techniques. They are one trade-off, read from opposite ends. I solved the same 21-task renovation both ways and the two answers landed on exactly the same curve.
The definitions, briefly, and then the useful part
Resource leveling treats your headcount as a hard limit. Nobody works a day that needs more people than you have, and if that means the project finishes later, it finishes later. The date is the output.
Resource smoothing treats the date as the hard limit. Tasks get shuffled within their float to flatten the daily peaks, and the finish is not allowed to move. The crew profile is the output — and if the peaks will not come down far enough, smoothing has simply failed at the thing you asked it for.
That last clause is where the exam definition stops and the real problem starts. Smoothing is described as if it always works, as a gentler alternative you reach for first. On most tight plans it does not work, and the standard advice gives you no way to find out except to try it and squint at the result.
The plan
Twenty-one tasks, a house renovation. Five resource pools: three general crew, two electricians, two plumbers, two painters, one excavator. The critical path arithmetic says 25 November 2026, and that answer assumes anything the dependencies permit to run in parallel does.
Count the bodies that schedule implies and the busiest day wants five general crew. There are three. Four of the five pools are never over capacity at any point — the electricians, plumbers, painters and the excavator all fit inside what exists. The entire problem is one pool, and it is over on fourteen separate days.
Leveling: fix the crew, read off the date
Hold general crew at three and solve for the earliest finish that never breaks it. The answer is 2 December — five working days later than the network promised, proved optimal, meaning no arrangement of those 21 tasks inside three crew finishes sooner.
Eleven of the twenty-one tasks moved to get there, and the daily profile goes from a peak of five to a flat ceiling of three with no day over. The work did not shrink. There are 155 crew-days of general labour in this job before leveling and 155 after — I went through what that redistribution actually looks like in the post on resource loading charts.
Smoothing: fix the date, read off the crew
Now the other direction. Hold 25 November and ask for the smallest daily crew that still hits it.
| Date you hold | Smallest crew that makes it |
|---|---|
| 25–27 November | 5 |
| 30 November – 1 December | 4 |
| 2 December or later | 3 |
Hold the CPM date and you need five. Not four with clever sequencing — five. On this plan, at that date, there is no smoothing available at all: the peak cannot be brought below the five the unlevelled schedule already demanded, because the float in the network is not deep enough to move any of the overlapping crew work anywhere else.
Give it three days of slip and the peak drops to four. Give it five and it drops to three, which is leveling arriving at the same place from the opposite direction. The two techniques produce identical numbers because they are the same question with different variables held constant.
Why they get taught as opposites
Because in most scheduling software they genuinely behave like two different buttons, and the buttons have different failure modes. Tammo Wilkens put the choice plainly in a PMI congress paper on the fundamentals of scheduling and resource leveling: you are choosing whether to “limit the resources” or “constrain the project finish”, and the tool reschedules activities by priority — usually total float — to serve whichever you picked. Two settings, one procedure underneath.
What the two-button framing costs you is the middle. Both buttons return a single point, and the decision almost always lives between them. Four crew is not a compromise nobody thought of; it is a real option that gets you to 30 November, and neither the leveling button nor the smoothing button will ever mention it, because you did not ask about four.
Look at the shape of that curve again. The fourth person buys two days. The fifth buys three more. The sixth buys nothing whatsoever — the plan hits a floor at the critical path and no amount of hiring goes below it. That ceiling is not guessable from either endpoint, and it is the single most useful thing on the chart, because it is the number that tells you when to stop negotiating for people and start changing how the work gets done instead.
The vocabulary, since the exam cares
- Resource leveling — capacity is hard, the finish date may move. Uses float first, then delays whatever it must.
- Resource smoothing — the finish date is hard, only float may be used. Cannot delay the project, and therefore cannot always deliver.
- Resource loading — not a technique at all, just the measurement: how many people each day the current schedule implies. It is what both techniques are looking at.
- Resource allocation — deciding who does what in the first place. Comes before either.
Worth knowing that “resource levelling” with two l’s is the same word — British spelling, same procedure, and searching either will find you the same arguments.
Getting both answers out of your own plan
The free scheduler has both as modes on the same file, which is the only reason this comparison was easy to run. Upload a task list with durations, predecessors and a crew per day, plus a sheet of what you actually have.
- “Work within my crew — when does it finish?” is leveling. Your headcount is the hard limit and the date is the answer.
- “Keep my date — how many people would it take?” is smoothing, extended honestly: if the peaks will not come down to what you have, it tells you the number it does need rather than quietly returning a schedule that breaks your capacity.
Run it both ways and you get the two ends. The strip underneath fills in the middle — one extra person at a time, and what each one buys — which is the part I would look at first and the part no textbook definition contains.
One caveat on all of it. Every number here is proved optimal, which is a stronger claim than most scheduling tools make and a much stronger one than most of them can support. When somebody benchmarked the leveling in seven commercial packages, the median schedule came back 3.6% above optimal and got worse as crews got tighter. If your tool tells you smoothing needs five people, it is worth knowing whether that five is a fact or a guess.
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.