Custom algorithm development
Packaged planning tools are built for the average problem, so their plans don’t work on yours. We build an algorithm around your problem instead.
- Free 30-minute call
- No commitment
- You own the code
In short: A custom algorithm is optimisation software written for one specific planning problem, as an alternative to a packaged tool. The algorithm encodes the company’s own rules, so its plans work in production without repair by hand.
The problem
At 6 am, the overnight planning run has timed out again. On a good day it finishes, but the plan violates a rule only you know. So you patch it by hand before the first shift. The tool runs, and you still do the planning yourself.
Some problems are too large, too specific or too time-critical for off-the-shelf planning software. Its plans can be valid on paper and still not work on the shop floor. Most advice on buying custom software covers two things: references and who owns the code. Both matter. Few people ask whether the algorithm will still fit your operation in year three.
When custom makes sense
- Your problem is too large to solve exactly in time. Exact methods prove that a plan is the best possible. On your problem, they stop unfinished or return nothing before you need the plan.
- A generic solver’s plans fall clearly short of what your operation needs. The solver runs, but every plan costs you more than it should.
- Your business rules are too specific for a standard tool. You have house rules, logic that grew over the years and exceptions nobody wrote down. The tool gives you no way to enter them.
- Standard settings can’t make the solver fast enough. You need answers in seconds or minutes, and tuning a generic tool won’t get you there.
What changes
Today, you take the tool’s plan and repair it every morning. With our solution, the plan already keeps your rules, so you check it and send it out. The system runs in production, it’s documented, and your team understands how it computes. They can maintain it themselves.
Will a custom algorithm age badly?
It can. Practitioners on Operations Research Stack Exchange, a question-and-answer forum for optimisation, describe the pattern well. A team tunes a model on typical cases. Then the business changes, and bit by bit people stop using the model. One answer adds a sharper point: a hand-built method beat the commercial solvers when it was new. A few years later the solvers had caught up, and the custom code hadn’t moved.
So we keep the custom part as thin as your problem allows. Whatever fits a standard model goes to a standard solver, which gets faster with each release. The custom search covers only what no solver handles in time. You also get the test cases, the sample problems we checked the algorithm on. If your problems today look very different from them, your team knows the algorithm needs an update.
How it works
We write custom optimisation algorithms in Python, the everyday language of optimisation work. Where a standard solver helps, the algorithm calls OR-Tools, Gurobi or HiGHS. We also use LPL, the modelling language our founder built at the University of Fribourg. The honest boundary: if a well-written model on a standard solver finishes before you need the plan, you don’t need a custom algorithm. We’ll tell you, and build that model instead.
When it isn’t the answer
The honest boundary: if a well-written model on a standard solver finishes before you need the plan, you don’t need a custom algorithm. We’ll tell you, and build that model instead.
Handover and tools
We don’t deliver a black box. We document every model and explain how it works. Then we train your team to run it, change it and maintain it. When the project ends, the know-how stays with you.
We aren’t tied to any solver or any vendor. We pick the solver that fits your problem and your deadline: Gurobi, CPLEX, OR-Tools, HiGHS or LPL. When an exact method can’t finish in time, we use metaheuristics. These are search strategies that find good plans fast, without proving they are the best. You never have to reshape your problem to suit a tool.
Where a custom model changes the outcome.
Multi-resource scheduling
You coordinate crews, machines and materials that all depend on each other. No standard tool keeps them in sync. We build a scheduling algorithm that respects every dependency, shift rule and setup time at once. Your plan then works on the shop floor.
Budget and resource allocation
You have a fixed budget or limited resources, and a long list of priorities. We model the allocation as one optimisation problem. You see the best possible split of money or capacity. You also see what each priority gains or loses when you move money between them.
Supply chain network design
You’re choosing where to put facilities, depots or distribution points. Every choice locks in cost for years. We build a network design model that weighs sites, goods flows and delivery speed together. You then choose the set of sites that the model calculates as best for cost and delivery speed.
Embedded optimisation for SaaS
You want optimisation as a feature inside your own product. We build the algorithm as a service your product calls. It answers as fast as your product needs, at your volumes. It’s documented, so your engineers own it.
Common questions.
Something else? Ask us directly.
You do. You get the documented source code, the test cases and the data formats it expects. You pay us no licence fee to keep it running. If you later hire your own optimisation engineer, they start from readable code instead of a black box.
Sometimes there is, and we try it first. Commercial solvers such as Gurobi handle many problems. They struggle when your rules can’t be written in their format or the answer is due in seconds. Then we add a small piece of custom code on top of a standard solver. That is usually the cheapest fix.
We replay your past planning weeks through it. Every plan gets checked against every rule, automatically, and compared with what you actually did. Where we can, we also compute a bound, a number no plan can beat, to show how much room is left. Only then does it run beside your current process.
Sometimes they can, and then we’ll say so. Standard tools are built for an average problem, though. With Custom algorithm development, 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.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
- Heuristic and metaheuristic
- Gurobi
- CPLEXIBM CPLEX
- HiGHSHiGHS
- OR-ToolsGoogle OR-Tools
- LPLLPL modelling language
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