Skip to content

Project and task scheduling

Your Gantt chart looks fine until two projects need the same crew. We schedule every project against the people and equipment that are free on the day.

  • Free 30-minute call
  • No commitment
  • You own the code
Typical industries
Construction and project-based, Public sector
Problem type
Resource planning
You receive
A documented model, run on your data

In short: A resource-constrained project schedule sets a date for every task in your projects. Each task starts only when the tasks it depends on are finished. And no person or machine is planned for two tasks at the same time. Every deadline then accounts for which people and machines are free at the time.

The problem

On Monday, you meet to go through all running projects. The plan says the foundations are poured on Wednesday. The crane is booked on another site until Friday. The welding crew is still finishing a job that ran late last week. On the Gantt chart nothing has moved. In reality, three projects are now waiting for the same crane and crew.

A Gantt chart shows each task as a bar on a timeline. It quietly assumes every task gets its engineer, crane and budget the moment it needs them. Projects share people and equipment, though. The moment two projects need the same person or machine, the dates on the chart are wrong.

Why Gantt plus spreadsheet fails

The usual approach plans the chart and the resources separately. A project tool puts the tasks in order. It looks at which task has to wait for another one and how long each task takes. Then you check in a spreadsheet or by phone who’s free. Each project is planned alone, as if all the shared people and equipment belonged to it. When two projects need the same scarce resource at the same time, the plan stops working, and you start re-planning by phone.

What changes

Today, you plan each project on its own and phone around to find out who’s free. Our model plans all projects at once, with the people and equipment they share. Each task starts when its people and equipment are actually free. You see a realistic finish date, and which resource keeps you from finishing earlier, say the only crane.

What you get

Project scheduling optimisation gives you a plan your shared people and equipment can deliver. The work is spread evenly across all projects running at the same time. We build the resource constrained project planning model on your data: your tasks, which task waits for which, and the people and equipment you have. The model takes three things as rules. A task waits for the tasks before it. No person or machine does two jobs at once. And several projects share the same people and equipment. Its goal is to meet the deadline. We test it on your running projects before you give a customer a new date.

  • A model written around your own project and task scheduling rules.
  • Results measured in the numbers you already track.
  • It slots into how you already work, and you can check every answer it gives.

What the levelling button actually does

Most project tools have a levelling button, and most guides tell you to press it. An over-allocation means a person or machine is booked for more work than fits in the time. The button clears it by delaying or splitting tasks. You choose the order it works in: by task number in the list, by priority or by its standard rule. You also decide whether the project may finish later. If not, the tool moves a task only as far as it can without making another task wait. That room to move is called slack.

So the button follows a simple rule: it clears one over-allocation at a time, each on its own. It never checks whether a different order across all your projects would be better. That goes wrong when it moves a task with no slack. Those tasks form the critical path: if one of them finishes later, the whole project finishes later.

Take two projects that need the same crane for two days. Task A is on project one’s critical path, and task B has five days of slack in project two. Level by task number with B listed first, and A waits. Project one now finishes two days late for no gain.

Our model for the resource-constrained project scheduling problem (RCPSP) looks at both projects at once. It lets B wait instead, because B has slack. With two tasks you’d catch it yourself. With three hundred tasks across six projects you won’t, and neither will the button.

How it works

We write your resources, the work they have to cover and your rules into one model. You get a plan that says which resource does what and when. It’s the plan that fits your goals best, such as the lowest cost or the shortest waiting time. Every rule is kept at the same time, and that’s the part a spreadsheet can’t do.

Resource-constrained project scheduling (RCPSP)

In planning, this is called the resource-constrained project scheduling problem, or RCPSP. In plain words: plan tasks when people and equipment are limited and shared. The model uses the same dependencies between tasks as a Gantt chart. It also records how many people and machines there are. Those limits decide when a task can happen, and the model plans around them.

Precedence: which task waits for which

The first layer is precedence, the rule that some tasks can’t start until others finish. A foundation comes before the frame, a design before the build and a test before the release. If a task waits for several others, it can start only when all of them are finished. A Gantt chart shows these dependencies as arrows between the bars. What it doesn’t draw is a task that’s ready but blocked, because its resource is busy elsewhere.

Resources: why RCPSP is hard

Resources are the second layer, and the reason for “resource-constrained” in the name. Each task needs specific resources for its duration: a welder, a crew of a certain size, a machine or a slice of budget. The model also decides which suitable person or machine does each task. So a project can finish late even though no task takes longer than planned. If two tasks need the same welder at the same time, one of them has to wait.

Multi-project resource levelling

The third layer appears as soon as you run more than one project. On a single project, a resource conflict affects only that project. With several projects running at once, it affects all of them. They all draw on the same engineers, equipment and budget.

Resource levelling smooths the peaks: weeks with more work booked than the people and machines can handle. It shifts tasks within their slack. No resource is asked to be in two places at once. Push a task back by hand in project one, and it may land in the week project two needs the same crew. Multi-project resource levelling in one model uses up a little slack. In return, you get a schedule you can actually run.

Makespan and deadlines

The model also needs an objective, the goal it pursues. The usual one is makespan, the time from the first task’s start to the last task’s finish. Minimising makespan means finishing as early as people and equipment allow. Deadline-constrained scheduling treats a fixed date as a hard limit. It asks which assignment of people and equipment gets you there. Both use the same model: one asks for the earliest finish, the other for a plan that meets the fixed date.

One model for the whole programme

Precedence, resource capacity and competition between projects become constraints, rules the plan must keep. Mixed-integer linear programming (MILP) is maths for yes-or-no choices, such as whether a task starts on a given day. You get the earliest feasible finish, the earliest date that keeps every rule.

For large programmes we add a second method. It’s called constraint programming. The model first rules out every start date and assignment that would violate a dependency or a capacity limit. Only then does it search the rest for the earliest finish. That’s what our RCPSP model does. You measure the result in makespan, in how busy each person and machine is and in the number of late tasks. That lets you set the new plan beside the old one line by line.

Where it matters most

Construction is the classic case. Trades such as bricklayers, electricians and painters work in a fixed order: the painter comes only when the electrician is done. Across several sites they compete for the same cranes, crews and site access. Engineering and R&D have the same problem, with designs and prototypes queueing for a few specialists and test rigs. In event planning, stages and suppliers must be ready on fixed dates, often with one crew. In the public sector, one overloaded group of project managers runs programme after programme.

Common questions.

Something else? Ask us directly.

Because levelling solves clashes by moving tasks to a later date. A clash means that two tasks need the same person at the same time. Some tasks have no slack: if one of them starts later, the whole project finishes later. Those tasks form the critical path. Most project tools decide by a fixed order which task to move, such as task number or priority. A different order can sometimes clear the same clash with no delay at all.

Usually not. A typical project tool solves one clash after another by a fixed rule. That’s fast and good enough for many single projects. But it doesn’t compare the possible task orders to find the earliest finish across all your projects. For a programme of projects that need the same crews, an optimisation model can find orders the rule never tries.

No. Your project managers keep the tool they know for tasks, dates and reporting. The model reads from the tool which tasks there are, which task has to wait for which, and which people and equipment each task needs. It plans the people and equipment that your projects share, across all projects. Then it writes the start dates back. Your project plan stays in the tool where everyone looks for it. Only now its start dates take into account which people and equipment are free.

That depends on what the contract pays for. If the contract sets a date after which you pay a penalty, that date becomes a fixed deadline. The model then finds the cheapest way to meet it. If finishing early frees crews for the next job, the earliest finish fits better. The model then makes the makespan as short as it can, the time until the last task is done. You can calculate both and compare. It’s the same model with two different questions.

Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.

You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.

You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.

Tell us what you’re planning.

In a free 30-minute call with a specialist, you find out whether optimisation can help.

  • Free, no commitment
  • You talk to a specialist directly
  • A clear go or no-go