Skip to content

Mathematical modelling and OR consulting

Many optimisation projects fail at the formulation, how the problem is written as maths. We get that right before anyone chooses an algorithm.

  • Free 30-minute call
  • No commitment
  • You own the code
Tools
Chosen per problem
Used in
50 documented use cases
Delivery
Documented source code, yours to keep

In short: A mathematical model in operations research (OR), the maths of planning, writes a planning problem as equations. It sets out the decisions, the goal each plan is judged by and the rules every plan must keep. Only then does anyone pick an algorithm.

The problem

A project starts with a tool. Six months later the model runs, but its plans don’t match how your business works. The project team wrote the problem down wrong at the start, and every line of code since was built on that. Many operations research (OR) projects fail like this. OR is the maths of planning decisions. The projects fail at the formulation, how the problem is written as maths, and not at the algorithm.

What changes

Today, the method comes first, and the problem gets bent to fit it. With our solution, we write the problem down correctly first, and the method follows from it. You know which approach fits before we write a line of code. Nobody spends months building the wrong thing.

How it works

Our consulting starts with three questions. What does the model decide? What makes a plan good? Which rules must every plan keep? We analyse the problem and identify the decision variables. These are the choices the model makes, such as which truck takes which order. We formulate the objective function, the score every plan is judged by, such as total cost. We model the constraints, the rules every plan must keep. Then we refine the model with your team, round after round.

A well-formulated model guides the choice of algorithm. It also shows whether the problem can be split into smaller parts that solve faster. With more than 20 years of formulation work, we see early which problems suit MILP. That’s mixed-integer linear programming, the standard method for yes-or-no decisions. We also see which need constraint programming, built for many hard rules, and which need something custom.

How do you tell a good formulation from a bad one?

Two models of the same problem can both be correct and still differ hugely in speed. The first test uses a simpler version of the model. In it, the solver may answer a yes-or-no choice with a fraction, such as a half-open depot. We then compare this answer with the best plan without fractions. The closer the two are, the less searching is left. In optimisation, this is called a strong bound.

The second test is the stopwatch, done fairly. We pick the test cases before we know which model they favour. Both models run on the same cases, on the same machine, in several realistic sizes. A thread on Operations Research Stack Exchange has a useful warning about the travelling salesman problem, the shortest round trip through a list of stops. There, the model that wins the first test can run out of memory on large cases. Then the other model is the better choice. You get the results of both tests before you commit to a model. The honest boundary: if your problem is a textbook one, a standard tool with a stock model costs less than consulting. We’ll point you to it.

When it isn’t the answer

The honest boundary: if your problem is a textbook one, a standard tool with a stock model costs less than consulting. We’ll point you to it.

Common questions.

Something else? Ask us directly.

Show them a week they remember. Put the model’s plan next to the plan they actually ran, and point to where the two differ and why. Write the rules in plain words, one line each. A top answer on Operations Research Stack Exchange adds a practical point: tell managers early when a change that looks small will take a long time to build.

Three things help most. One week of your data, even if it’s messy. A list of the rules you follow, including the unwritten ones. And one plan that went wrong, with what made it wrong. From those, we can sketch the decisions, the goal and the rules in the first session.

An exact method eventually proves its plan is the best possible. A heuristic gives no such guarantee, however good its plans look. Between them sit approximation methods. They prove how far their plan can be from the best one at most, as answers on Operations Research Stack Exchange spell out. Ask any vendor which of the three you’re buying.

Sometimes they can, and then we’ll say so. Standard tools are built for an average problem, though. With Mathematical modelling and OR consulting, we start from yours: its size, its rules and the exceptions nobody wrote down. That is why the plan fits your operation better.

You don’t need either to start. We build the model from the data you’ve got, messy or not, and cleaning it up is part of the job. You get a documented system your team can maintain, and we train them until they can run it long after we’ve left.

It starts 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. You’ll see results before you commit to anything bigger.

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