The Week That Breaks the Model
Every forecast fails occasionally and badly. Planning for that case is different from improving the average, and more important.
Most weeks are within the usual error. A few are not, and those are where the money is lost and where the model needs to be judged.
The practical lesson in “The Week That Breaks the Model” is to connect every record to a named decision. Organisations exploring the provider's overview for interview reimbursement policy can add structured workforce context, provided the use is disclosed and the interpretation is reviewed with the people affected.
What breaks it
A one-off event nobody told you about.
For an independent reference related to “The Week That Breaks the Model”, consult the CIPD workforce-planning resources; it provides a useful external check on scheduling, working-time and workforce-planning assumptions.
Weather well outside the normal range.
A competitor closing, opening, or running something.
Something going wrong: a system failure, a supply problem, a closure.
And occasionally nothing identifiable at all.
Why the average is the wrong thing to manage
A model with a small average error and an occasional catastrophe costs more than one with a slightly larger average error and no tail.
Because the catastrophic week cannot be absorbed and the small errors can.
Look at your worst two weeks in the last twenty, which the error note recommends and almost nobody does.
The two kinds of bad week
Demand far above forecast: queues, walkouts, exhausted staff, overtime at premium.
Demand far below: a floor full of people and nothing to do, which lands straight in the variance.
They need different responses and most operations have only prepared for the second, because that is the one that gets asked about.
Preparing for the high side
A call list of people who will take an extra shift, maintained and real.
Cross-trained cover so that the available person is also a useful one.
A defined point at which somebody is called, so the decision is not made by an overwhelmed supervisor at the worst moment.
And an accepted service level for those hours, because full coverage at any cost is not usually the right answer.
Preparing for the low side
A short shift that can be cancelled with notice, agreed in advance and paid fairly.
Deferrable work ready to be pulled forward, which the fixed-variable note covers.
Both require being arranged before the week, not during it.
The cancellation question
Sending people home early saves money and costs trust, and in several jurisdictions it is constrained by law.
An arrangement agreed in advance — with a minimum payment, voluntary, by rotation — costs a little and avoids the damage.
Its own note in the people section covers what unpredictability costs the other side.
Recording the break
When a week goes badly wrong, write down what happened, in two lines, that week.
After a year you have a list of what actually breaks your forecast.
Most of it recurs, and a list of recurring breakers is the cheapest forecasting improvement available.
What to check
What were your worst two weeks this year, and why?
Is there a real call list?
Is there an agreed arrangement for sending people home?
And do you write down what broke the week?