Skip to content

Multi-objective resource planning

The cheapest plan isn’t always the best one. We show you what faster delivery costs in staff hours, and you pick the plan.

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

In short: A multi-objective optimisation plans with several goals at once, such as cost and delivery time. It shows every plan where you can only improve one goal by giving up some of another. These plans form the Pareto frontier, and you pick the one that suits you.

The problem

The quarterly plan reaches your desk with one number circled: total cost. It’s the cheapest plan the team could find. Inside it, the second shift at the East warehouse is gone. Customers in that region will now wait two days for their goods instead of one. Nobody decided that. The calculation decided it, because cost was the only number it was asked to bring down.

Why the cheapest plan hides the decision that matters

Almost every plan wants two things at once: low cost and short delivery times. Delivery time is the time from the order until the goods reach the customer. These two goals work against each other: when one gets better, the other gets worse. More warehouse staff and more drivers shorten delivery times, but they cost wages. A truck that only leaves when it’s full saves cost. Your customers wait longer for it.

You know this from booking a flight. Flight 1 to New York costs 600 francs and takes 14 hours with a change. Flight 2 is direct, costs 900 francs and takes 9 hours. Flight 3 costs 950 francs and takes 16 hours. You’d never book flight 3, because flight 1 is cheaper and faster. Whether flight 1 or flight 2 is better depends on what five hours are worth to you.

Your planning works the same way, only with thousands of possible plans instead of three flights. If you only cut cost, you always book flight 1, and nobody was asked. The calculation quietly decides that cost beats everything else. With several goals in view, you see the choice and make it yourself.

Why a single weighted objective misleads

The common fix is one weighted objective, a score that rates possible plans by your criteria. You give each goal a weight, say 70% on cost and 30% on delivery time. The model multiplies each goal by its weight and adds them up. Then it looks for the plan with the best score. The weights are often picked by gut feel. Yet they decide the answer before the model runs.

There’s a technical problem too: cost is counted in francs, delivery time in hours. Take two plans for one quarter. Plan A costs 1 million francs, and goods reach the customer within 48 hours. Plan C costs 1.4 million and delivers within 24 hours. The 24 hours saved count 0.3 × 24 = 7.2 points. The 400,000 francs of extra cost count 0.7 × 400,000 = 280,000 points. Delivery time barely matters, even though you gave it 30%.

The other shortcut only cuts cost and adds delivery time as a fixed rule: “Every order reaches the customer within 36 hours.” Why 36 and not 40? Maybe those four hours cost 200,000 francs a quarter. You never see that, because the calculation hands you just one plan. The decision about what an hour is worth is still there. It’s hidden in the number 36.

What changes

Today, the team hands you one plan, the cheapest. What it gives up in delivery time stays out of sight. With our model, you see all the best plans side by side, each with its price. One costs 250,000 francs more, keeps the second shift at the East warehouse and delivers there the next day. Another is cheaper and shows openly in which region customers wait longer. You choose, and you know exactly what your choice costs.

Multi-objective optimisation: what you get

Multi-objective optimisation counts each goal separately in one model: cost in francs and delivery time in hours. Other goals can join when you need them, such as CO2 or how much money is tied up in stock. Then it calculates the best plans. With each of them, you can only deliver faster if you pay more. You get an interactive view with one slider per goal. Move delivery time from 40 to 32 hours, and you see the extra cost at once. You also see which warehouse needs more staff to make it happen.

The future is uncertain. So the model runs each plan through many possible futures, such as a flu wave in the team or a supplier who delivers two weeks late. For each plan, you see how often delivery then takes too long and how far costs could rise. We test the model on your data, in your hardest quarters. Only then do you commit.

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

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.

What the Pareto frontier is

Plot every workable plan on a chart, with cost on one axis and delivery time on the other. Plan X costs 1.2 million francs and delivers within 48 hours. Plan A costs 1 million and also delivers within 48 hours. So plan A is cheaper and just as fast. In optimisation, we say A dominates X. Plan X is like flight 3: you’d never pick it.

The plans that remain form a curve, the Pareto frontier. It’s named after the economist Vilfredo Pareto. On that curve, you only deliver faster by paying more, and you only save by delivering slower. Every plan off the curve wastes money or time. No plan beats the ones on it. Pareto frontier optimisation calculates every plan on this curve, so you choose from all the good plans. With three or more goals, the curve becomes a surface, and the logic stays the same.

Weighted-sum and epsilon-constraint methods

The plans on the Pareto frontier aren’t simply there. The model has to calculate them, one after another. Two techniques do this in practice. Weighted-sum optimisation runs the model many times with different weights. Each time, it keeps the plans no other plan dominates. It’s fast and fine for a first overview. The epsilon-constraint method turns every goal but one into an upper limit. In optimisation, that limit is called epsilon.

Here’s an example. Cost may be at most 1.3 million francs. Within that limit, the model finds the plan with the shortest delivery time. Then we lower the cost limit step by step, say to 1.25 and then 1.2 million. Each step adds one more plan on the curve. When booking flights, it’s a price filter you lower in steps, noting the fastest flight each time.

This way, the method calculates plans at even intervals along the whole curve. It also finds plans that the weighted sum never shows. Together, the two methods give you all the best plans, the ones called Pareto-optimal. You get the complete menu instead of a sample.

Why some good plans never show up in a weighted sum

Das and Dennis showed in 1997 that weighted sums have two blind spots. First, evenly spaced weights don’t give evenly spaced plans. You often get many near-identical plans at one end of the curve and a big gap in the middle. Second, there are plans that no weight finds at all. Take three plans. Plan A costs 1 million and delivers in 48 hours, plan B costs 1.25 million and delivers in 40, and plan C costs 1.4 million and delivers in 24.

A weight is really a price: it says what one hour of faster delivery is worth to you. Plan B costs 250,000 francs more than A and is 8 hours faster, about 31,000 francs per hour. Plan C costs 150,000 francs more than B and is 16 hours faster, about 9,400 francs per hour. The first 8 hours are the expensive ones, so no weight stops at B. Compare A with C directly: C costs 400,000 francs more and is 24 hours faster, about 17,000 francs per hour. If an hour is worth less than that to you, A wins. If it’s worth more, C wins.

So plan B wins under no weighting at all. Yet B is often the right plan, for example when your budget stops at 1.3 million. A weighted sum never shows it to you. An epsilon constraint finds it at once: cost at most 1.3 million, delivery time as short as possible. In optimisation, we say the curve bends inward at that point.

Supply chain and manufacturing trade-offs

In the supply chain, more safety stock shortens delivery times, because the goods are already in the warehouse. Safety stock is extra stock you keep for surprises, and it ties up money. A cheaper transport lane, such as rail instead of road, saves cost and CO2. The goods may then take a day longer. Pareto optimisation in the supply chain shows what each of these steps means for cost and time.

In manufacturing, longer runs mean lower changeover costs. A changeover is switching a production line from one part to the next, for example changing tools and resetting the machine. Make a part for a week straight instead of one hour every day, and you switch less often. In return, a customer waits longer until their part comes round again. The Pareto frontier shows which run length fits your costs and delivery times.

Stochastic optimisation plans with many possible futures and takes into account how likely each one is. For each plan on the curve, it shows how often the plan runs into trouble, such as several people off sick at once. You pick a plan that still works in a bad quarter.

Common questions.

Something else? Ask us directly.

Look for where the Pareto frontier bends. Say that at first, 50,000 francs more cuts delivery time by eight hours. From a certain plan on, the same 50,000 francs saves only one hour. That bend is a good place to start the discussion. From there, you move along the curve until cost and delivery time match what you’re willing to pay.

The maths handles many. People are the limit. Two goals give a curve you can read at a glance, and three give a surface you can still explore. With more, we show the goals in pairs, or fix some as limits and let you trade off the rest. Most decisions come down to two or three goals that really work against each other: when one gets better, another gets worse.

Yes, as long as each decision in the model has an emission figure attached, such as per kilometre driven, per tonne shipped or per unit produced. Then you see the emissions of every plan next to its cost and its delivery time. You see what each tonne of CO2 saved costs you in money or delivery time. You can also set an upper limit, say 500 tonnes of CO2 a year, and watch how the plan adapts.

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