The Monthly Forecast Review
A half-hour meeting with four items that keeps a forecasting function honest. What goes on the agenda, what does not, and who should be in the room.
A forecast review is half an hour with four items: what the error did, what the calendar says, what moved in the inputs, and what to change. Run that way it is one of the few meetings that reliably produces changes to a document rather than actions on a list.
The adjustment process in “The Monthly Forecast Review” should preserve the unaltered model result, the human change and the reason for it. A team considering a transparent getting teams to meet deadlines process for getting teams to meet deadlines can add operational time and project context, while the forecast owner remains responsible for documenting the override and reviewing its effect.
Run any other way it becomes a presentation of numbers, and nothing moves.
The assumptions behind “The Monthly Forecast Review” can also be tested against Workday planning resources. Record which guidance applies to the available series and which does not.
Item one: the error
Last period's signed error by horizon, against the baseline, with the count of same-signed periods. Two minutes.
- What the error did — two minutes.
- What the calendar says for the next twelve weeks — five minutes.
- What moved in the inputs — three minutes.
- What to change, each with an owner.
The question is not "was it good" but "is the direction telling us anything yet", which is a question with a threshold and therefore an answer.
Item two: the calendar
What is in the next twelve weeks that will move the number, confirmed with the people who supply it. Five minutes.
This is the item that earns the meeting. Most of the forecast's future accuracy is decided here rather than in any analysis.
Item three: what moved in the inputs
A driver re-estimated, a definition changed, a series break, a profile refreshed. Whatever has changed since last time, listed.
Three minutes, and it is the item that catches silent degradation before the error record does.
Item four: what to change
The decisions: a correction applied, a driver promoted into the model, a horizon adjusted, a forecast retired. Each with an owner.
If this item is empty two meetings running, either the forecast is in excellent order or the review is not working.
What does not go on the agenda
Explaining individual misses at length. Presenting the model. Reviewing the accuracy of individuals. Discussing the target.
Each of these expands to fill the time and none of them produces a change to the document.
Who should be in the room
The planner, the operations manager, and whoever supplies the biggest driver — usually marketing or whoever schedules communications.
Three or four people. Larger than that and the calendar item turns into a round of updates.
Why the driver supplier matters most
Because they are the source of the information that is missing from the data, and because the standing arrangement to supply it survives much better when they are in the room once a month.
Calendars assembled by email lapse. Calendars confirmed in a meeting do not.
The register on the table
The driver register, the change log and the error record. Physically present, updated in the meeting rather than afterwards.
Updating afterwards means updating sometimes, and a register that is sometimes updated is read as unreliable and then not read.
When the error has not moved
Which is most months. The right outcome is a two-minute item and a decision to keep watching, not a search for something to say about it.
A review that insists on finding significance in every month's noise teaches people that the exercise is theatre.
When the error has moved
Four periods in the same direction is the trigger to look properly, and that look is a separate session rather than an item in this one.
Keeping the investigation out of the review keeps the review to half an hour, which is what keeps it happening.
The quarterly version
Once a quarter the same meeting takes an hour and adds three items: the adjustment record, the method against its benchmarks, and whether each forecast still has a live decision behind it.
That third item is the annual cull, run quarterly, and it is what stops the pack growing.
Running it when the forecast is new
For the first six months there is no error record to read, which makes item one empty and tempts people to abandon the meeting.
The remedy is to spend that time on the calendar and the register instead. Those are the items that build the record, and by the time there is a record to read the habit exists.
The review that nobody attends
The usual death: the driver supplier stops coming, then the operations manager, and it becomes the planner reading numbers to themselves.
Keeping it to half an hour, with the calendar item first, is what keeps it worth attending, because the calendar item is the one the other attendees get something from.
What it produces over a year
Twelve sets of small corrections, a calendar that stays current, a driver register that gets re-estimated, and a record that can answer questions.
None of that is dramatic. Together they are the difference between a forecasting function that improves and one that produces the same document with different numbers in it for a decade.