Skip to main content

6. How Long Should the Forecasting Horizon Be?

Teaching rewrite — the Objective step: which future periods deserve the team’s attention? Original wording; numbers are illustrative.

Introduction​

Planners can’t polish a three-year forecast for every product. Most manufacturers populate 18 months automatically, but management usually tracks accuracy on one or two lags — e.g. “we focus on M+2 because production is planned two months ahead”, or “we track M+3 and M+12”. Those choices are often made by habit, not by analysis of the supply decisions they serve.

This lesson covers:

  • Lags and the risk horizon (lead time + review period)
  • Why the right focus is a range (M1…M4), not a single lag
  • A worked order-sizing example that links forecasts to orders
  • Why freezing forecasts is a bad practice
  • Looking beyond the risk horizon: safety-stock targets and supplier collaboration
  • How lost sales vs backorders change what matters

6.1 Lags​

A lag is how many periods ahead a forecast is. By convention lag 0 is the current period. For a monthly forecast made in January: lag 0 = January, lag 1 = February, lag 2 = March…

The puzzle​

You order monthly, on the first day of each month. Your supplier’s lead time is 3 months: an order placed on day 1 of M1 arrives at the start of M4 (and can’t serve demand from earlier months). Your next order goes out at the start of M2.

Which lag(s) should demand planners focus on to support this ordering decision? M+3? M+4? A range?

Most planners answer M+3 or M+4. The better answer is M1 to M4 (cumulative) — and a little beyond.

6.2 Theory: the risk horizon​

Almost all real inventory policies are periodic: you order on a calendar (daily, weekly, monthly), not at any instant. For a periodic policy the stock must protect you over the risk horizon:

Risk horizon = lead time + review period

Here: 3 months lead time + 1 month review period = 4 months. To size today’s order you need the forecast for every month M1–M4.

flowchart LR
  O[Order placed<br/>start M1] -->|lead time 3 months| R[Receipt<br/>start M4]
  R -->|review period 1 month| N[Next receipt<br/>start M5]
Why practitioners get this wrong

Textbooks spend more time on continuous-review policies (where the risk horizon is just the lead time), and some planning software forgets the review period. In periodic policies, forgetting it under-protects you.

6.3 Reconciling the forecast with the order​

Set-up: start-of-M1 stock 150. Orders already in transit: +40 arriving start of M2, +30 arriving start of M3. Demand forecast: M1 50, M2 75, M3 25, M4 50. Policy: finish each month with 100 units (the safety stock) before the next receipt.

MonthOpening stockReceipt (start of month)ForecastClosing stock
M1150—50100
M2100+40 → 1407565
M365+30 → 952570
M470+X (today’s order)5070 + X − 50

Today’s order can only affect M4. To close M4 at 100: 70 + X − 50 = 100 → X = 80.

Now change any forecast in M1–M4 and X changes one-for-one. If M2 rises from 75 to 100, M3 closes at 45 instead of 70, so X must rise to 105 (+25).

Equivalently: order = target + total forecast over the risk horizon − stock − pipeline = 100 + (50 + 75 + 25 + 50) − 150 − (40 + 30) = 100 + 200 − 220 = 80.

So every month in the risk horizon matters equally.

Don’t freeze the forecast

Many companies freeze the near-term forecast (e.g. M1) because the plan for M1 can no longer change. But the plan (what you will do) and the forecast (what you expect customers to want) are different things. Updating M1 still changes the right order quantity for M4. Freeze plans if you must — keep the forecast current.

Checkpoint: lags and risk horizon​

Quiz

Checkpoint — sections 6.1 to 6.3

1 / 4

Weekly orders, supplier lead time 5 weeks. Risk horizon?

Pick an answer to see if you are right.

6.4 Looking further ahead​

Safety stock depends on what comes next​

How much stock is sensible at the end of M4 depends on the outlook for M5:

ScenarioM5 expected demandSensible end-of-M4 stock
10 (product ending)Little or none — leftovers risk becoming obsolete
21,000A larger buffer is cheap insurance

Many companies set safety stock as weeks of cover (“3 weeks for our top items”), which makes the end-of-M4 target a function of the M5 forecast. So the useful horizon is risk horizon + a few periods.

Supplier collaboration​

A longer view helps suppliers plan capacity, cut lead times and costs, and raise their service level. Share expected supply requirements (net of your stock and policies) — not raw demand forecasts.

6.5 Lost sales vs backorders​

SituationWhat happens to unmet demandWhat matters in the forecast
Lost sales (common in B2C, FMCG)Disappears — customer buys elsewhereEach period individually, because shortages in a given month lose demand
Backorders (customers wait)Carried forward until stock arrivesThe cumulative total over the risk horizon; carry unfulfilled forecast into later periods
Hybrid (typical B2B)Some wait, some leaveBoth: each period and the total

Worked example: a demand surprise in M1​

Same set-up, but M1 demand jumps from 50 to 200.

Lost sales:

MonthOpeningReceiptDemandClosingLost
M1150—200050
M20+4075035
M30+30255—
M45+X50100—

X = 100 + 50 − 5 = 145 — only +65 versus the original 80, because 85 units of demand were lost rather than carried forward.

Backorders: all 150 extra units are still owed, so the order rises by the full +150 to 230 (100 + 350 total demand − 150 stock − 70 pipeline).

Pro tip

Even during a shortage, keep forecasting demand, not the sales you expect to manage. Track backlogs explicitly so carried-over orders aren’t mistaken for new demand.

Checkpoint: beyond the horizon​

Quiz

Checkpoint — sections 6.4 and 6.5

1 / 4

A product is discontinued after M4. What should happen to the end-of-M4 stock target?

Pick an answer to see if you are right.

Rank: horizon practice​

Quiz

Choose and use the horizon

1 / 2

Rank the steps to set a team’s focus horizon.

OrderDrag the handle to rearrange — top should be first.

  1. 1Store every forecast version so multi-lag accuracy can be measured.
  2. 2Set the focus horizon to the risk horizon plus a few periods.
  3. 3Track accuracy over that whole range, not a single lag.
  4. 4List the supply decisions the forecast supports (orders, production, capacity).
  5. 5For each, find the lead time and review period.

Drag items into the correct order (or use ↑↓), then check.

Check: forecasting horizon​

Quiz

How long should the forecasting horizon be?

1 / 5

In a monthly forecast made in March, lag 2 is…

Pick an answer to see if you are right.

Takeaways​

Remember
  • Risk horizon = lead time + review period; size orders from the forecast over the whole range
  • Every period in the risk horizon changes the order one-for-one — don’t focus on a single lag
  • Don’t freeze forecasts; freeze plans if needed
  • Look a few periods beyond the risk horizon for safety-stock targets and supplier collaboration
  • Lost sales → each period matters; backorders → the cumulative total matters; hybrid → both
  • Multi-lag accuracy tracking needs stored forecast versions — use proper tools, not spreadsheets

Further Reading​