Think Build Implement Repeat
London, UK +44 7367 067226
WhatsApp FOLLOW f in X
  1. Home
  2. Blog
  3. Choosing a Forecast Horizon You Can Believe
AI & Machine Learning

Choosing a Forecast Horizon You Can Believe

Accuracy decays with distance, and past a point a forecast tells you nothing you could act on. How to find that point for your own business.

Updated 2 min readBy SpiderHunts Technologies

Free estimateNo obligation

Get a free estimate

Tell us what you need. A senior engineer reads every enquiry.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →

Quick answer — TL;DR

The useful horizon is set by your lead time, not by what the software can produce. Measure how accuracy decays with distance on your own data, compare that against the point at which decisions become irreversible, and forecast to that horizon rather than further.

The horizon nobody chose

Most forecasts run to a horizon inherited from whatever the previous system produced, or from the financial calendar. Neither has much to do with whether the number is still meaningful at that distance.

The consequence is quiet waste: a twelve-month forecast where only the first six weeks change a decision, reviewed monthly by people who know the back half is fiction.

Two questions decide it

The useful horizon sits at the intersection of two things: how long before a decision becomes irreversible, and how far out your forecast still beats a naive baseline.

  1. Work out your true lead time - supplier lead time plus internal processing plus any review cycle. That is the minimum horizon worth forecasting.
  2. Measure accuracy decay: backtest at one week, four weeks, twelve weeks and so on, and see where the model stops beating a seasonal naive baseline.
  3. If the decay point is shorter than the lead time, the honest conclusion is that you cannot forecast far enough ahead to drive that decision, and the answer is buffer stock or a shorter lead time rather than a better model.

What to do beyond the useful horizon

Business planning needs numbers further out than the forecast can support. That is legitimate - it just needs a different label.

Beyond the point where the model adds nothing, use a planning assumption: last year plus agreed growth, or a capacity envelope. Calling it an assumption rather than a forecast changes how people treat it, and stops anyone building an order on it.

DistanceWhat it isWhat it should drive
Within lead timeForecast, monitored for accuracyOrders, rosters, allocations
Just beyond lead timeForecast, wider rangeProvisional plans, supplier warnings
Beyond decay pointPlanning assumptionBudgets, capacity, headcount

Reviewing more often beats forecasting further

Where a long horizon is genuinely needed, the better investment is usually a shorter review cycle rather than a more distant forecast. A twelve-week forecast refreshed weekly is far more useful than one refreshed quarterly, even if the single-shot accuracy is identical.

This is often an organisational change rather than a technical one, and it tends to be the cheaper half of the project.

If the forecast cannot see as far as your lead time, the problem is the lead time.

FAQ

Frequently asked questions

The questions readers ask us after this guide.

Still have a question?

Ask us directly — a senior engineer will get back to you.

Ask about your project

How do I find my accuracy decay curve?

Backtest the same model at several horizons and plot the error against distance. The point where it converges with a naive baseline is where the model stops adding value.

Should I forecast further out for the budget?

Produce a planning assumption instead and label it clearly. It avoids anyone treating a long-range number as an operational forecast.

Does more data extend the useful horizon?

Sometimes, but the limit is usually inherent unpredictability rather than data volume. External drivers can help where they are themselves knowable in advance.

Is a rolling forecast worth the effort?

Usually yes where decisions are made continuously. Refreshing more often typically beats extending the horizon.

Keep reading

More on AI & Machine Learning

Start here

Want machine learning project details from us?

Tell us what you are trying to predict and roughly what data you hold. We will come back with an honest view on whether machine learning is the right tool, what the work would involve and a realistic cost range. If a spreadsheet would do the job, we will say so.

  1. You tell us what you needTwo minutes on the form, or a message on WhatsApp.
  2. A senior engineer reviews itAnd comes back with questions, a realistic range and an honest view on fit.
  3. Free 30-minute scoping callWe talk through scope, options and a realistic estimate — with no obligation.
Free estimateNo obligation

Talk to someone who builds this

Send a short brief and we will come back with an honest view and a realistic range.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →