Optimisation & the solver mindset

Free visual lesson · about 5 minutes · short quiz at the end

ccPlanning academy · advanced

Optimisation & the solver mindset

Behind every “auto-schedule” button is a maths problem worth understanding.

The big idea

Scheduling is a constrained optimisation problem.

“Cover the requirement at lowest cost, without breaking any rules” is the textbook shape of an optimisation problem: an objective to minimise, subject to constraints. WFM auto-schedulers are solvers doing exactly this — and understanding the framing makes you their master, not their servant.

The three ingredients

Objective, variables, constraints.

Objective

what to minimise (e.g. cost, under-cover)

Variables

what you can choose (shifts, starts)

Constraints

rules that must hold (contracts, rest)

Every solver, simple or fancy, is juggling these three.

What solvers are brilliant at

Searching a vast space fast.

There are billions of ways to arrange a month of shifts. A solver explores that space far faster and more thoroughly than any human, finding low-cost rosters that satisfy every hard rule. For big, constraint-heavy problems, this is genuinely beyond manual scheduling.

What solvers can’t fix

A wrong objective, optimised perfectly.

A solver answers exactly the question you pose. Tell it to minimise cost and it’ll happily produce a brutal, unliveable roster that’s technically optimal — and drives attrition. Garbage objective, optimal garbage. The hard part isn’t the algorithm; it’s defining what “good” means.

Hard vs soft constraints

Must-haves and would-likes.

Hard constraints can never break (legal rest, max hours). Soft constraints are preferences you’d pay to honour (shift preferences, fairness, weekends off). Good optimisation weights the soft ones thoughtfully — over-tighten and there’s no feasible roster; ignore them and you get an efficient, hated schedule.

The planner’s real job

Frame the problem; sanity-check the answer.

The solver does the search; you set the objective, encode the constraints honestly, and judge whether the output is humane and sensible. An optimiser is a power tool — it amplifies a good problem definition and a bad one equally. Never ship a solver’s roster without a human read.

Optimal ≠ good

“Minimise cost” — be careful what you ask

Tell the solver one thing — minimise cost — and it obeys perfectly: split shifts every day, lunch at 3pm, weekends for the newest staff, nobody’s preferences honoured. Technically optimal. Also unliveable, and it drives the attrition that blows the plan apart within a quarter.

The algorithm did its job; the objective was wrong. The craft was never the maths — it’s defining what “good” means before you press solve.

The takeaway

The maths is easy; the framing is the craft.

Scheduling is objective + variables + constraints, and solvers search that space superbly. But they optimise exactly what you ask — so the value is in defining “good” well, balancing hard and soft constraints, and sanity-checking the result. Master the framing and the solver works for you.

Now test yourself ↓

1 / 8

Slides done? Here’s the same idea in a bit more depth — the part worth keeping.

In depth: the maths behind the auto-schedule button

“Cover the requirement at lowest cost, without breaking any rules” is the textbook shape of a constrained optimisation problem: an objective to minimise, subject to constraints. WFM auto-schedulers are solvers doing exactly this, and understanding the framing is what makes you their master rather than their servant. Every solver, simple or fancy, is juggling three ingredients — an objective (what to minimise, like cost or under-cover), variables (what you can choose, like shifts and start times), and constraints (rules that must hold, like contracts and rest).

What solvers do brilliantly — and can’t fix

There are billions of ways to arrange a month of shifts, and a solver explores that space far faster and more thoroughly than any human, finding low-cost rosters that satisfy every hard rule — for big, constraint-heavy problems this is genuinely beyond manual scheduling. But a solver answers exactly the question you pose: tell it to minimise cost and it’ll happily produce a brutal, unliveable roster that’s technically optimal and drives attrition. Garbage objective, optimal garbage. The hard part isn’t the algorithm; it’s defining what “good” means.

Hard vs soft constraints, and the planner’s real job

Hard constraints can never break (legal rest, maximum hours); soft constraints are preferences you’d pay to honour (shift preferences, fairness, weekends off). Good optimisation weights the soft ones thoughtfully — over-tighten and there’s no feasible roster, ignore them and you get an efficient, hated schedule. So the planner’s real job isn’t the search, which the solver does: it’s setting the objective, encoding the constraints honestly, and judging whether the output is humane and sensible. An optimiser is a power tool that amplifies a good problem definition and a bad one equally — never ship a solver’s roster without a human read.

The principle to remember: the maths is easy; the framing is the craft. Scheduling is objective + variables + constraints, and solvers search that space superbly — but they optimise exactly what you ask, so the value is in defining “good” well, balancing hard and soft constraints, and sanity-checking the result.

Optimisation & the solver mindset