Skip to content

Resource allocation and assignment

You have more work than people, machines and budget. You get the allocation that does the most with what you have and keeps every rule.

  • Free 30-minute call
  • No commitment
  • You own the code
Typical industries
Manufacturing, Healthcare
Problem type
Resource planning
You receive
A documented model, run on your data

In short: A resource allocation optimisation assigns limited people, machines and budget to tasks so the overall result is as good as possible. The model considers every rule and criterion at the same time. It doesn’t fill one task after another.

The problem

On Monday at eight, the week’s work is already bigger than your team. You have twelve people, four machines and a budget that won’t stretch. Thirty jobs are waiting for someone to do them. By lunch everyone has something to do, and nobody can tell you whether that was the best use of what you’ve got.

That’s the resource allocation problem. You have a fixed number of people, machines and vehicles, and a fixed budget. And you have more work than they can cover. Who does what, where and when? With five people and five jobs you can see it at a glance. With twelve and thirty you can’t, and you know it.

Why one-slice-at-a-time allocation fails

Usually you solve it step by step, in a spreadsheet. First you assign people to jobs by who is available. Then you give the machines to whoever is free. Only at the end do you check whether the budget is enough. Each decision sees part of the picture and limits the next one.

Tighten a skill requirement and fewer people qualify. Move a machine and one shift has to work without it. Add a fairness rule and the cheapest assignment stops working. A spreadsheet can’t weigh all of that at once. So a plan that looks balanced on paper stops working on Wednesday, when a job needs the welder you already sent elsewhere.

What changes

Today, you fill the plan slice by slice: people, then machines, then the budget. Our model weighs every person, machine and budget line at once. You start from a complete allocation and check it. Where you know better, you give a job to someone else. Then you send the plan out. The plan keeps every skill requirement, capacity limit, budget line and fairness rule.

Task assignment optimisation: what you get

Resource allocation optimisation weighs all your rules and limits together. You get the allocation that does the most with what you have. We build the task assignment optimisation model on your own cost, capacity and demand data. Sometimes each person gets exactly one task, sometimes one person takes many. The model handles both, exactly as they happen in your operation.

For every plan you see four numbers. Utilisation is how much of their time people and machines are busy, say 85%. You also see the cost, how many jobs stay uncovered and how evenly the work is shared. Ask what one more order would cost in people and machines, and you get a number. We test the model on your figures, in your hardest weeks, before it goes live. After that, you work with an allocation that works in daily practice.

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

When a spreadsheet stops being enough

Most guides tell you to try Excel Solver first. It’s fair advice, and for small cases it works. Take the Monday example: twelve people and thirty jobs. Every person-job pair is a yes-or-no decision, so that’s 360 decisions. The Solver that ships with Excel is limited to 200 decision variables and 100 constraints. Decision variables are the cells it may change, and constraints are the rules it must keep.

So the textbook spreadsheet model doesn’t fit this week’s plan, and you haven’t added a single machine yet. You can split the problem by team or by day, and many do. Split by team, and team A’s spare welder never shows up in team B’s plan, even on the day team B is one welder short. Split by day, and Monday takes the only free machine that Thursday needed.

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.

The assignment problem

The cleanest form of resource allocation is the assignment problem. You have agents, tasks and a cost for every agent-task pair. An agent is whoever does the work: a person, a machine or a van. The one-to-one version asks for the cheapest pairing where each agent gets exactly one task.

The classic method is the Hungarian algorithm, published by Harold Kuhn in 1955. It finds the cheapest pairing exactly. It also runs in polynomial time, which means the effort grows steadily with size instead of exploding. For a pure one-to-one case, a plain assignment problem solver is all you need. Most operations aren’t one-to-one, though.

Many-to-one and bin packing

A shift has ten people and forty tasks. A depot has three forklifts and a hundred movements. The textbook trick: if a person can take four tasks, you list that person four times. Ten people then become forty entries for forty tasks. Then you run the Hungarian algorithm as usual. That works while every task takes the same time. Once one job takes four hours and another twenty minutes, the trick stops working.

What you have then is the generalized assignment problem. Each task takes a certain number of hours, and each person has a limited number of hours. The problem is NP-hard. In plain words, no known method solves every large case both quickly and exactly. A closely related problem is bin packing. There you fit items of different sizes into containers with limited room. Here the items are tasks, and the containers are people, machines or shifts.

In practice, there are special cases. Work arrives one job after another. Some shifts cost more than others, and some tasks have to stay with the same person. Each special case changes the best answer. A good model takes all of them into account at once, so you don’t have to correct the plan by hand afterwards.

Multi-criteria resource assignment

In most cases, several criteria count at once. You match on skills, cost and time windows. You also match on location, fairness and seniority, meaning how long someone has worked for you. All of it counts at once. Nurses aren’t interchangeable, and neither are cranes. Multi-criteria resource assignment weighs those criteria together and returns the allocation that scores best on the measures you care about.

One model, every constraint

We solve the whole allocation as one optimisation. Each criterion becomes a constraint or an objective. A constraint is a rule the plan must keep. An objective is a goal the model pushes as far as it can, such as the lowest total cost. For simple cases we write our own algorithms and feed them your costs. Examples are the Hungarian algorithm, matching methods and bin packing heuristics. Heuristics are fast rules of thumb that find good answers without checking every option.

For everything else we use a second method. It’s called mixed-integer linear programming (MILP), maths built for yes-or-no decisions such as whether this person takes this task. With MILP, one model covers all three: one-to-one, many-to-one and multi-criteria assignment. You get the best feasible plan, meaning the best one that keeps every rule.

Where the problem shows up most

You’ll find this problem wherever there is more work than people and machines can cover. In manufacturing, you assign machines to jobs and maintenance crews to assets. In healthcare, you assign doctors to wards and operating theatres to surgical teams. There, a poor match means an operation is delayed. In the public sector it’s officers to cases and inspectors to sites. Budgets there often shrink while caseloads stay the same.

Common questions.

Something else? Ask us directly.

Only in a special case. The Hungarian algorithm gives each person exactly one task. Say each person can take three tasks. Then you enter that person three times and run it as usual. Once tasks take different amounts of time and each person has a limit in hours, that no longer works. That’s the generalized assignment problem. It needs an integer model, which works with yes-or-no decisions. Or it needs a heuristic, a fast rule of thumb.

For small cases, yes. Build a grid of yes-or-no cells, one per person and task. Each task is assigned once, and nobody gets more work than they can do. The Solver that ships with Excel is limited to 200 cells it may change and 100 constraints. Twelve people and thirty tasks already make 360 cells, so you’ll reach that limit early.

Then those pairs are simply forbidden. In the Hungarian algorithm you give them a cost so high it never gets picked. In an integer model you leave out the yes-or-no decision for that pair. That also makes the model smaller. Either way, build the list of forbidden pairs from your skills and certificates, not from memory. Writing that list down is worth doing before any model exists.

Add it as a rule or as a goal. A rule caps everyone’s workload, say at 10% above the team average. A goal asks the model to shrink the gap between the busiest and the quietest person. The rule is easier to explain to the team. The goal leaves the model room to trade a little evenness for a cheaper plan.

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