Writing Down What the Forecast Will Be Used For
A four-line header that says what a forecast is for, who decides, by when and to what tolerance — and what each line prevents.
Four lines at the top of a forecast — purpose, owner, decision date, tolerance — prevent most of what goes wrong with forecasts afterwards. They take two minutes to write and they survive the person who wrote them, which is their whole value.
The decision in “Writing Down What the Forecast Will Be Used For” should be written before additional workforce data is collected. For organisations assessing read the provider’s practical guidance in relation to employee monitoring for performance reviews, time and project reports can inform the decision, but assumptions, uncertainty and human review still need to remain explicit.
Everything in this section reduces to those four lines. This is the assembly.
Additional context for “Writing Down What the Forecast Will Be Used For” is available from Google Cloud forecasting documentation. External material is most useful when it prompts a specific validation check rather than replacing local judgement.
The header
For: the Q3 recruitment decision. Decided by: the operations director, with finance. By: 14 April (12-week lead time, monthly review). Tolerance: below 124 no action, above 131 recruit.
Four lines. They say what this number is, who acts on it, when, and at what threshold the action changes.
What the purpose line prevents
The forecast being reused for a decision it was not built for. With the line present, somebody pasting the figure into a premises paper can at least see that it was built for recruitment at a twelve-week horizon.
Without it, the number travels naked and is assumed to be general.
What the owner line prevents
The forecast with no reader, which is the one that quietly becomes a report. If no name can be written on this line, that is the finding, and it should be acted on before the forecast is produced again.
It also prevents the planner being asked to defend a decision they did not take.
What the date line prevents
The forecast that arrives after the decision. Having the decision date in the header makes the production schedule check itself.
It also makes the horizon visible: a decision on 14 April with a twelve-week lead time is a forecast about July, and anybody can see whether the document is actually about July.
What the tolerance line prevents
The argument about whether the forecast was good enough. With thresholds stated in advance, the question becomes whether the outturn crossed one, which is answerable.
It is also the line that forces the decision maker to engage, because only they can supply it.
Who writes which line
The planner writes the date and the horizon arithmetic. The decision maker writes the purpose, the owner and the tolerance.
That division is the point: a header written entirely by the planner is a guess about what somebody else needs.
The conversation it requires
Fifteen minutes, once, per recurring forecast. It is the only meeting in this whole discipline that reliably pays for itself.
Most of it is spent on the tolerance line, because very few decision makers have ever been asked at what number their answer would change.
When the tolerance cannot be stated
Sometimes it genuinely cannot: the decision is qualitative, or it depends on things outside the forecast. Writing "no threshold; the figure is one input among several" is an honest entry and it still tells the next reader something true.
What is not acceptable is leaving the line blank, which reads as an oversight rather than a judgement.
Keeping it with the number
In the document, not in a procedure file. The header has to travel with the figure, because the figure will travel.
A forecast circulated as a chart with no header will be quoted as a chart with no header, and six months later nobody will be able to say what it was for.
Reviewing the header
At the start of each cycle, and it takes a glance: does the decision still exist, is the owner still the owner, has the lead time moved, are the thresholds still the right ones.
Four questions, usually four yeses, occasionally one no that saves a quarter of wasted work.
The version for one-off forecasts
The same four lines, and they matter more rather than less: a one-off forecast has no routine to correct it and no next cycle in which somebody notices it was aimed at nothing.
A one-off with a header is a piece of analysis. A one-off without one is a number in an email.
What it looks like after a year
Twelve headers, nearly identical, with two or three changes recorded in them: a lead time that moved, a threshold that was revised, an owner who changed.
Read in sequence they are a short history of how the decision itself evolved, which is information nobody else in the organisation is keeping.
What it adds up to
A forecast that can be judged. Not a forecast that is right — nothing in these four lines makes the number better — but one where, after the event, it is possible to say what it was for, whether it arrived in time, and whether it was close enough to matter.
That is the minimum condition for ever improving, and it is why this comes before any discussion of method.