Demand You Created Yourself
A large share of workforce demand is generated by the organisation's own actions, which makes it forecastable from internal plans rather than from the demand series.
A substantial share of the demand a workforce handles was created by the organisation itself: by a letter it sent, a system it changed, a process that failed, a bill it issued. That demand is forecastable from internal plans rather than from the demand history, and it is usually forecast from the history anyway.
The method in “Demand You Created Yourself” is only as useful as the definitions and source records behind it. A team exploring open this published workflow for remote workforce management software can add timely work and project context, with transparent settings, limited access and a documented correction route around every consequential use.
Separating it out is the difference between predicting the weather and reading a diary.
The assumptions behind “Demand You Created Yourself” can also be tested against PwC workforce resources. Record which guidance applies to the available series and which does not.
The four sources
Communications sent out. Changes made to systems or processes. Failures — defects, outages, errors needing correction. And the backlog the operation itself created by not completing work earlier.
- Communications sent out — bills, statements, notices, campaigns.
- Changes made to systems or processes.
- Failures: defects, outages, errors needing correction.
- Backlog the operation created by not completing work earlier.
Each is knowable in advance from somewhere inside the building.
Communications
Bills, statements, renewal notices, service messages, marketing. Each has a dispatch date and a response rate that is estimable from the last several.
An operation that knows its dispatch calendar and its response rates can forecast a large part of next month from first principles.
Changes
A release that alters a screen produces questions for a fortnight. A process change produces questions for longer. Both are in the change calendar and neither is usually joined to the demand forecast.
The effect size is estimable from the last few comparable releases if anybody recorded them.
Failures
The least comfortable category and often the largest. A defect that affects two per cent of customers produces contacts from a share of them, and the shape of that arrival is well understood once measured.
Incident records exist. Joining them to demand is rarely done because the two live in different teams.
Self-generated backlog
Work not completed today arrives again tomorrow, often in a more expensive form — a chase, a complaint, an escalation. This is demand created by the staffing decision itself.
It makes the forecast and the staffing plan circular, which is real and worth modelling rather than ignoring.
Why the circularity matters
Under-staff a week and the following week's demand rises. The forecast for the following week, built from history, does not know this.
Operations that have measured it typically find a meaningful multiplier on deferred work, which means a short week costs more than the hours it saved.
Forecasting self-generated demand
Not from the series. From the plan: dispatch calendar times response rate, release calendar times question rate, incident volume times contact rate, backlog times chase rate.
Four small calculations, each needing one rate estimated from history and one schedule obtained from a colleague.
What it leaves behind
Genuine external demand, which is usually more stable and more forecastable than the total. Separating the self-generated part often reveals that the underlying series is far better behaved than it looked.
That is a common and welcome surprise.
The lever this creates
Self-generated demand is controllable. A dispatch moved by a week, a release held back from a peak, a defect fixed — each reduces demand rather than requiring it to be staffed.
A forecast that identifies the self-generated share puts that option on the table, and it is usually cheaper than cover.
Who needs to see it
Whoever schedules the communications and the releases. Shown that a dispatch date creates four days of pressure, they will frequently move it, because they had no idea.
This is the most direct value a planning function can deliver and it requires no modelling at all.
Measuring the rates
One quarter of joining calendars to demand. The response rate to a dispatch, the question rate after a release, the contact rate after an incident. Three numbers.
They drift, so they are re-estimated annually, and they are far more stable than the demand series they explain.
The dispatch nobody counted as demand
Worth stating plainly: a letter is not a cost of printing. It is a printing cost plus a share of recipients contacting you, and the second is usually the larger figure.
Pricing dispatches that way changes how many get sent, which is a result no forecasting improvement can match.
Measuring the deferral multiplier
The hardest of the four rates and the most valuable. Compare weeks where the backlog grew against the following weeks' volume, controlling for the usual.
It needs a year of data and it answers a question operations managers argue about constantly: what short-staffing actually costs later.
What this does to the forecast's character
It stops being a prediction about customers and becomes, in large part, an arithmetic statement about the organisation's own schedule.
That part of the forecast is not uncertain in the same way. It has its own band, which is much narrower, and it should be shown separately because it is the part that can be acted on directly.