The Far Horizon, and What It Can Honestly Say
What a forecast a year or more out can and cannot support, why scenarios beat point estimates there, and how to publish a far number that survives contact with a budget process.
A forecast a year or more out can honestly say what level the operation is likely to be running at, within a wide band, if nothing structural changes. It cannot say what any particular week will look like, and it cannot anticipate the changes that will dominate the outcome.
The horizon in “The Far Horizon, and What It Can Honestly Say” should determine which current records are relevant and how much detail the forecast can honestly support. A planning team evaluating explore the published product guidance for workforce analytics software can connect recent time and project context with short-term capacity questions, while keeping longer-range assumptions in a separately governed model.
Everything that goes wrong at the far horizon comes from publishing it as though it could do more than that.
Teams documenting “The Far Horizon, and What It Can Honestly Say” may use Pigment planning resources as an outside reference. Test the method against the real forecast horizon, known events and operational constraints.
What survives at a year
The trend, if there is a real one and the series is long enough to see it. The seasonal shape, which is usually the most stable thing in the data. And announced structural change: a contract starting, a site closing, a service moving.
- Survives: the trend, if real; the seasonal shape; announced structural change.
- Does not: events, campaigns, competitor behaviour, weather, anything to do with price.
That is the entire content. Anything else in a far forecast is assumption wearing arithmetic.
What does not survive
Events, because the calendar does not exist yet. Campaigns, because marketing plans a quarter at a time. Competitor behaviour. Weather. Anything to do with price.
These do not merely weaken the forecast; at a year they are usually larger than the trend.
Why the band is wide and should stay wide
Because it is honest. A ±35% band at a year is not a sign of a poor method; it is a measurement of how much the world moves in a year.
The pressure to narrow it comes from the budget process, which wants a number, and narrowing it does not reduce the uncertainty — it moves it somewhere it cannot be seen.
Scenarios instead of a point
At this distance the useful output is usually three cases with their drivers named: what happens if the contract renews, if it does not, if it renews at reduced volume.
Each case is internally coherent and has consequences anybody can price. A single point between them describes a world nobody expects.
How many scenarios
Three. Two reads as optimistic and pessimistic and gets averaged. Four or more stops being read.
The three should be built around the one or two decisions that actually dominate the outcome, not around arithmetic variations of a central case.
Naming what each scenario rests on
One line each: what has to be true. "Contract renews at current volume; no change to the channel mix." That line is what lets somebody in six months say which scenario the world is tracking.
Without it the scenarios are three numbers and the exercise is decorative.
The far forecast in a budget process
It will become a commitment whatever anybody intends. The band will be dropped, the central case will be adopted, and the figure will be reported against monthly.
Knowing this in advance, the thing to negotiate is not the number but the trigger: at what observed level does the budget get revisited.
Trigger levels rather than revisions
"If weekly volume exceeds 131 for four consecutive weeks, the headcount plan is re-opened." One sentence, agreed in advance, removes most of the argument that otherwise happens in month seven.
This is the far horizon's most useful product and it is almost never asked for.
What a long series buys
Mostly the ability to see whether a trend is real. Two years of data cannot distinguish a trend from a slow cycle; five usually can.
Where the series is short, the honest far forecast is flat with a wide band, and saying so is better than fitting a line to noise.
Reviewing a far forecast
Quarterly, against the scenarios rather than against the point. The question is which case the world is tracking, not whether the number was right.
That review takes ten minutes and it is the one that actually catches structural change early.
The forecast that is really a capacity question
Often what is wanted at a year is not demand at all but whether the estate, the systems or the recruitment pipeline could cope if demand rose. That is a different question and it has a better answer.
Separating it out early saves the forecast from being asked to carry a decision it cannot support.
The far forecast that was right for the wrong reason
Worth watching for in the review. A central case that lands accurately because two large errors cancelled is not evidence that the method works.
Checking the components rather than the total is what distinguishes a sound far forecast from a lucky one, and at this horizon there are too few observations for luck to be averaged out.
What to put on the page
The three cases, what each rests on, the trigger levels, and one line saying what the forecast cannot see at this distance.
That last line is what stops the number being read as a plan, and it is the only defence a far forecast has once it leaves the room.