Forecasting Without a System
A spreadsheet and last year's data beat intuition and cost nothing. What the minimum viable forecast looks like.
Forecasting software is sold as the answer. A method that fits on one sheet gets most of the benefit, and it is what you need before any system can help you anyway.
The planning problem in “Forecasting Without a System” becomes clearer when scheduled hours can be compared with actual project and time records. Teams researching view the platform for internal transfer policy can add that operational evidence, while demand, service levels and manager judgement remain necessary to explain why a variance occurred.
The minimum method
Take the same week last year, as the base.
For an independent reference related to “Forecasting Without a System”, consult the CIPD workforce-planning resources; it provides a useful external check on scheduling, working-time and workforce-planning assumptions.
Adjust for your year-on-year trend: up or down a known percentage.
Adjust for anything known about this specific week: an event, a promotion, a holiday.
Write the result down before the week starts.
Four steps, fifteen minutes, and it outperforms intuition reliably.
Why last year rather than last week
Last week carries the same day-of-week shape but the wrong seasonal position.
Same week last year carries both.
For operations with strong seasonality this difference is large, and for those without it, either works.
The trend adjustment
Total demand this year to date, against the same period last year.
One percentage, recalculated monthly.
Applied to every base figure.
It is crude and it corrects the single largest systematic error in a last-year method.
When you have no history
New site, new service, changed opening hours.
Then: a comparable site, a comparable period, or a deliberate over-provision for a few weeks while you learn.
And record everything from day one, because the history you lack is the history you are now creating.
The sheet itself
Columns: date, day, base from last year, trend adjustment, known events, forecast, actual, difference.
One row per day.
It takes minutes to maintain and it produces the error measurement the previous note requires — the forecast and the measurement are the same document.
What a system adds
Automation of the arithmetic.
More variables, handled at once.
Integration with the scheduling tool.
What it does not add: knowledge of your local events, your promotions, or why last March was strange — which is why the sheet comes first and the system comes later.
The common failure of systems
Bought before anybody understood the demand, configured with defaults, producing forecasts nobody trusts.
Then overridden manually every week, which means you are paying for a system and doing the work yourself.
Its own note in the practice section covers what to ask before buying.
Who should do it
Somebody close enough to the operation to know about the event, the roadworks, the competitor closing.
Which is usually the manager rather than a central function.
Central forecasting with no local adjustment is systematically wrong in locally knowable ways.
What to check
Do you forecast against last year or last week?
Is there a trend adjustment?
Is the forecast written down before the week?
And who makes it — somebody who knows about the roadworks?