Optimisation means finding the best feasible result for a defined objective. A business might minimise delivery costs, maximise profit or choose a combination of products within its available capacity.
The US spelling is optimization. In a mathematical model, feasible means meeting the stated conditions, or constraints. An improvement is meaningful only in relation to the objective being measured.
Objectives, choices and constraints
An objective states what is to be minimised or maximised. Decision variables are the choices that can change, such as production quantities or delivery routes. Constraints limit those choices, for example through available labour, vehicle capacity or deadlines.
The NEOS Guide to Optimization describes these as the main elements of a model. A feasible solution satisfies its constraints. An optimal solution has the best objective value among the feasible solutions being considered.
Maximisation is a form of optimisation. It does not inherently disregard costs. A profit-maximisation model can deduct costs from revenue and include limits on staffing, materials and output. Maximising output is a different objective from maximising profit.
A delivery example
Suppose a fictional business has three possible routes for the same deliveries. Route A costs £80 but misses a required deadline. Route B costs £95 and meets it. Route C costs £110 and also meets it. Assume all other requirements are satisfied.
If the objective is to minimise cost while delivering on time, Route A is excluded. Route B is the best feasible choice among the three. The business is not simply choosing the lowest number.
Google’s OR-Tools introduction uses delivery routing to explain how objectives and constraints work together. This article’s prices are invented to illustrate that distinction.
What an optimised result does and does not show
The result depends on the alternatives, data and assumptions in the model. A solution that improves on nearby choices may be a local optimum without being the best across the whole feasible set. NIST’s definition of a local optimum explains this distinction.
Automation can implement a process without improving its objective. Conversely, a process can improve through fewer errors or lower costs even if its completion time is unchanged. Buying software alone does not demonstrate optimisation.
Where objectives compete, such as cost and service quality, the result depends on the trade-off chosen. There need not be one operating point that simultaneously gives the lowest cost and the highest performance on every measure.
Last checked and updated on November 2, 2026 by the MBN editorial team.