Decision optimization

A good forecast tells you what is likely to happen. Deciding what to do about it, such as how many people to put on a shift, which truck takes which order or how much stock to hold, is a separate and usually harder problem.

We build systems that take forecasts, costs and rules and propose decisions that work within them. Planners see why a plan was suggested, can change it and stay responsible for the final call.

Get in touch

What decision optimization does

Most planning problems have too many possible answers to check by hand. A delivery plan with a few dozen stops already has more possible orderings than anyone could compare. Optimization methods search through them systematically and find plans that meet the rules at the lowest cost or highest value.

Machine learning estimates the parts that are uncertain: how long a job will take in practice, how demand will react to a price, which orders are likely to be canceled. Used together, the two produce plans that respect the rules and rest on realistic estimates.

What we build

Scheduling and staff planning
Shift plans, production schedules and maintenance windows that balance workload, skills, working-time rules and deadlines.
Routing and dispatch
Vehicle routes and job assignments that account for time windows, capacity and traffic, and can be re-planned when something changes during the day.
Pricing and promotion planning
Price and discount suggestions based on how demand responds, within the margins, rules and brand limits you set.
Inventory and replenishment
Order quantities and stock levels per item and location that weigh the cost of running out against the cost of holding too much.
Resource allocation
Decisions on where budget, equipment, capacity or people go when there is not enough to meet every request.
Scenario planning
How a plan changes if demand rises, a supplier is late or a rule is relaxed. Options can be set side by side before anyone decides.

When it's a good fit

Optimization pays off on decisions that come up often and involve many options and rules, where small gains add up over time. You may not need it yet when:

there are only a handful of options and an experienced planner already picks well
the rules are not yet agreed, so any plan would be argued over anyway
the decision is made rarely and a spreadsheet comparison does the job

How a project runs

Write down the decision and its rules

We sit with the planners to capture what is decided and how often, and every rule a plan must respect, including the ones that exist only in people's heads.
Replan past periods

The model replans past periods using only the data available then. We compare its plans with the planners' actual plans on cost, service level and rule violations. The planners then go through the differences and point out what the model missed.
Plan alongside the team

The system starts by making suggestions that planners can accept, edit or reject. Their changes feed back into the rules and the model.

Making plans people trust

Planners follow a plan when they can see that it respects their rules and understand why it looks the way it does.

Every rule is checked
Working hours, vehicle capacities, minimum stock levels, contract terms and safety limits go into the model as rules. Each plan is checked against them before anyone sees it, and if no plan can meet them all, the system says which rule is in the way.
A reason for every plan
Planners can see why a plan looks the way it does and what changing it would cost, which makes it easier to accept or challenge.
Tested against disruption
We test plans against delays, cancellations and forecast errors. A late truck or a canceled order should not mean rebuilding the whole plan.

What you receive

A planning model that captures your rules, costs and goals
An interface or connection that puts suggestions where planners already work
The rules written down, in a form your team can update
Code and documentation, handed over to your team

Frequently asked questions

What data does decision optimization need?

Records of past decisions and their outcomes, the rules planners follow, and some time from the people who make these decisions today. If some of the data is missing, we start with what exists.

How soon can we see whether optimization helps?

The first results come once we replan a few past periods and compare the system's plans with the ones your team used. How soon depends mostly on how clear the rules are and how easy the data is to reach.

Can a decision optimization system run on our own servers?

Yes. The planning system can run on your own servers or private cloud. Cost, contract and staff data then never leave your environment.

What happens when our planning rules change?

Your team can update the rules, or we can keep supporting the updates. We also track whether plans are followed and deliver the expected results, and adjust the model when they do not.