Skip to content

Facility location optimisation

Where you put your sites decides what you pay for the next ten years. Our model picks them on cost and delivery time together.

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

In short: A facility location optimisation chooses where to open warehouses, plants or service sites, and which customers each one serves. It weighs building and running costs, transport, distance to customers and risk across all candidate sites at once.

The problem

The lease for the new warehouse is on the table. The shortlist came from a map and a spreadsheet that ranked each site on rent and distance. Sign it, and the cost is locked in for ten years. Every quarter after that you live with the result, whether the volumes and rents you assumed still match or not.

Few planning decisions last this long. A warehouse, depot, clinic or charging hub fixes what your transport costs. It also fixes how far you are from your customers. It also ties up capital: money you can’t use for anything else. Those costs run for as long as the site does.

Why site-by-site ranking fails

The usual approach ranks each candidate on rent and distance, picks the cheapest few and hopes the allocation works out. Allocation means deciding which customers each site serves. The catch is that location and allocation are one decision. Choose a site and you change which customers it can serve cheaply. Change who it serves and you change whether it earns back its fixed cost. Fixed cost is what you pay even on a quiet day, such as rent and staff.

A spreadsheet can’t see when a slightly more expensive site lets you serve your customers much more cheaply. And it doesn’t check whether every customer gets the goods on time or whether a site can handle the volume. So the plan that looks cheap per site leaves a region unserved, overloads a hub or opens a facility that never pays for itself. Each of those mistakes runs for the length of the lease.

What changes

Today, you rank sites one by one and hope the customer allocation works out. Our model chooses the sites and the customers each one serves in one step. You see every possible combination of sites with its total cost, travel times and spare capacity. Then you test it against the changes that worry you, such as a new large customer or 30% more volume. Only then do you sign.

What you get

Facility location optimisation treats a site as what it is, a commitment for years. You get the set of sites that does best on cost and delivery time together. With distribution centre location optimisation, you don’t simply pick the sites that look cheapest one by one. The model decides sites and customers together. We build it on your delivery addresses and volumes, your site and transport costs and your delivery promises. Then we test the chosen sites before you sign a single lease.

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

Why the centre of gravity isn’t the answer

Most warehouse location guides start with the centre of gravity method. You compute the middle point of all your customers on the map. A customer with a large volume pulls the point closer than a small one. It’s quick, and it’s a fair first sketch. But it measures in straight lines, gives you one point and knows nothing about rent, opening cost or capacity.

Straight lines hurt more in Switzerland than on a flat map. A depot and a customer on opposite shores of Lake Zurich are close in a straight line and far apart by road. Carling and colleagues tested this in rural Sweden in 2012. Location plans built on straight-line distance came out 2 to 7 per cent worse than plans built on road distances. Worse means customers had that much farther to go by road to their nearest site. Swiss lakes and passes won’t shrink that gap.

Time windows matter even more. Say your trucks leave at 05:00 at the earliest, and your customers want their goods by 07:30. That leaves two and a half hours. So a site’s service area is whatever a truck can reach in that time. We model it as travel time on the road network, with your departure rules built in.

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.

P-median, p-centre and fixed charge

Three classic models describe the problem, each with a different goal. The p-median facility location model starts from p, the number of sites you plan to open. It places them so the total distance from every customer location to its site is as small as possible. Each distance is weighted. Weighted means a busy customer counts more than a quiet one.

The p-centre model keeps the longest trip as short as possible instead. That is the trip to the customer who is farthest from the nearest site. The fixed-charge model adds the cost of opening a site at all, which you pay whether it runs full or not. Ignore that cost, and the model scatters capacity across too many half-empty sites. Which of the three fits depends on what hurts you most: a long average distance, a single customer far from the nearest site or fixed costs for half-empty sites.

Coverage, capacity and phasing

Covering constraints keep every customer, patient or vehicle within a set travel time of an open site, say 45 minutes. That way the network keeps its service promise. Capacity limits stop the model sending more demand to a site than it can handle. Planning over several years lets you open sites in stages. You open as many as today’s peak needs and keep room for expected growth. You don’t commit capital now for volume that may never come.

One mixed-integer model

We solve the whole location problem as one optimisation. Fixed costs, coverage and capacity become constraints and objectives. A constraint is a rule the plan must keep. An objective is a cost or service goal the model pushes as far as it can. The tool is mixed-integer linear programming (MILP). It handles yes-or-no choices, such as open this site or not, together with how much demand flows where.

You can see why the model chose each network. For every proposed network you see which constraint is binding, meaning the rule that holds the plan back most. You also see what relaxing that rule would cost. We use the same method for supply chain network design. It decides how goods flow between suppliers, sites and customers. The location model picks the sites, and network design plans the flows between them.

Testing sites against the future

Nobody knows the demand of the coming years exactly. So we test the chosen network before it’s built. Scenario planning calculates the network with the developments that matter to you: a jump in volume, demand moving from one region to another, a new large customer. You see which sites still pay off and which would be left with too little work. You end up with sites you’d still choose if a large customer arrives or a region grows slower than planned.

Where facility location matters

Warehouse location optimisation is the obvious case. The number and placement of distribution centres set transport cost and delivery speed for everything that moves. Store and clinic chains have the same question. Where should the next store or clinic open? It should win new customers, not simply take them from your existing sites. Charging stations for electric vehicles add a coverage constraint. Chargers go where many people drive, and no driver is too far from the next one.

Common questions.

Something else? Ask us directly.

As a first sketch, yes. The method calculates the midpoint of all your customers, and large customers count more. But it measures in straight lines and gives you a single point. It knows nothing about rent, opening cost or capacity. Use it to draw the search area. Then compare the candidate sites with road travel times in a proper location model.

That depends on your delivery promise more than anything else. Fix the promise first, for example next-morning delivery to almost every customer. Then let the model find the cheapest network that keeps it. Run it again with a looser and a tighter promise. The comparison shows how many more sites and how much more money the tighter promise costs.

Road distances or, better still, travel times. Carling and colleagues compared both on a location problem in rural Sweden in 2012. Straight lines gave plans 2 to 7 per cent worse, measured by the road distance from customers to their nearest site. The gap grew as more sites were placed. In Switzerland, lakes, valleys and passes make road time the only honest measure.

Demand by location, ideally a year of deliveries with postcode and volume. Possible sites with rent, the cost of fitting them out and running costs. Transport rates, or a cost per kilometre and per hour. And your delivery promise as a number: the longest drive from a site to a customer, in hours. You probably have most of it already, spread across your ERP and your freight invoices.

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