Forecasts Nobody Acts On
How a forecast becomes a monthly ritual that changes nothing, the four reasons it happens, and how to end it without ending the forecasting.
A forecast nobody acts on is a report, and it costs the same as a forecast. Most planning functions produce at least one, and it usually survives for years because stopping it would require somebody to say out loud that it never mattered.
The decision in “Forecasts Nobody Acts On” should be written before additional workforce data is collected. For organisations assessing see the official forecasting workflow in relation to task switching cost, time and project reports can inform the decision, but assumptions, uncertainty and human review still need to remain explicit.
The useful question is not whether people like the forecast. It is whether any decision in the last twelve months went differently because of it.
As a separate benchmark for “Forecasts Nobody Acts On”, see Forecasting Principles resources. Keep source data, model, override and resulting decision visible as distinct steps.
The four reasons it happens
The decision it was built for disappeared. It arrives after the decision has to be taken. It arrives in a form the decision maker cannot use. Or it was never aimed at a decision at all and was built because the previous planner built one.
- The decision it was built for has gone.
- It arrives after the decision has to be taken.
- It arrives in a form the decision maker cannot use.
- It was never aimed at a decision at all.
Each has a different remedy, and they are not interchangeable.
The forecast that outlived its decision
The commonest. A recruitment programme ended, a service was closed, a contract moved elsewhere — and the monthly pack kept the page.
Nobody removed it because removing a page requires a decision and leaving it requires none.
The forecast that arrives too late
Produced on the tenth for a decision taken on the fifth. The planner knows the date; the decision maker assumes the planner has a reason.
This one is pure scheduling and it is fixed in an afternoon, which is why it is worth finding first.
The forecast in the wrong form
A monthly total supplied to somebody who has to staff Tuesday afternoons. A half-hourly profile supplied to somebody deciding a headcount budget.
The number may be excellent. In that form it answers a question nobody is asking, and the recipient quietly builds their own.
The shadow forecast
Which is the real symptom. Somewhere in the organisation there is a spreadsheet that the decision maker actually uses, built by somebody who needed a number in a usable form and could not get one.
Finding it is the fastest diagnosis available. It shows exactly what the official forecast should have been.
Asking without putting anybody on the spot
"What did you do with last month's figure?" is an interrogation. "If this had come a fortnight earlier, would it have changed anything?" is a design question, and people answer it honestly.
A handful of those conversations locates every dead forecast in the pack.
Stopping one
Announce it, do not simply drop it. "The regional volume page is being withdrawn from next month; it fed the expansion decision, which closed in March. If anybody is using it, tell me and it stays."
Somebody usually is, for a reason nobody knew, and that reason is worth more than the page.
What replaces it
Often nothing, and that is a legitimate outcome: an hour a month back, and a pack short enough that the remaining pages get read.
Sometimes it is replaced by something much smaller — a single figure with a threshold — which is what the decision needed all along.
The forecast that is really an accountability record
Worth naming, because it is a real use and it is usually disguised. The monthly number exists so that somebody can later show they flagged the rise.
That is a legitimate purpose but it is a different document with different requirements, and calling it a forecast means it gets judged against a standard it was never trying to meet.
The forecast that feeds somebody else's decision
Occasionally the answer to "who acts on this" is outside the department: a supplier plans against it, a partner staffs to it, a shared service sizes itself from it.
That is a real use and it is usually undocumented, which means the forecast can be changed or withdrawn without anybody realising who it reaches. Writing the external reader into the header costs a line.
Keeping one that produces nothing
There is one defensible case: a forecast kept running deliberately to build the error record for a decision that does not exist yet but will. A new service, a planned expansion.
That is an investment and it should be labelled as one, with the date at which it is expected to acquire a reader.
Starting again afterwards
Where a forecast is withdrawn and the decision later returns, the series and the error record should not be thrown away with the document.
Restarting with three years of history and a record of past accuracy is a different proposition from starting from nothing, and the cost of keeping them is a folder.
The annual cull
Once a year, for each recurring forecast: what decision, whose, by when, and when did it last change an answer. Four columns.
The first time it is run it typically removes a third of what the function produces, and the remaining two-thirds get noticeably better because there is time to do them properly.