Guides
How Long Should You Forecast? Match the Horizon to Your Lead Time
A 30-day forecast cannot answer a 60-day buying question. The window you project over should be set by how long you will be waiting, not by the calendar.
The horizon is a question, not a setting
Open a forecast and it tends to arrive pre-set to a month — because a month is the unit calendars come in, not because it answers a question anyone asked. The horizon is the window you project demand over, and it should be derived, not inherited from a dropdown.
What you derive it from is the buying question. When you place a purchase order you are committing to survive a stretch of demand on the stock you have plus the stock you just bought. The question, then: how much will sell between now and the moment more units can actually be on the shelf? A window shorter than that stretch is not a conservative forecast. It is a confident answer to a different question — what will sell this month, while you are deciding about the next quarter.
That is not what a reorder point answers. A reorder point asks which signal fires the order; the horizon asks how wide a window that signal reads.
Two clocks set the horizon: the wait and the gap
Two clocks run between the moment you look at a number and the moment new stock is sellable. The first is replenishment lead time — the gap between placing the order and units being on the shelf and pickable: production, transit, customs, receiving, put-away. Note that lead-time spread does different work from lead-time length: variability sizes the buffer you hold, while length, added to your review period, sets how far out you look.
The second clock gets dropped: your review cycle. If you sit down to buy every Monday, anything you decline to order this morning stays uncovered for another seven days — and only then does the lead-time clock start ticking. Review weekly against a 30-day supplier and your true exposure is 37 days, not 30.
Add them and you have the window. Inventory theory calls it the protection interval: in a periodic-review system, the review period plus the lead time, R + L. The practitioner version says the same thing — forecast horizon equals lead time plus planning cycle, and a longer lead time requires a longer horizon.
forecast horizon = replenishment lead time + order cycle (the protection interval, R + L)
Suppose a supplier quotes 45 days door-to-door and you run your buying session once a month. An item you skip today waits up to 30 days for the next session, then 45 more before the replenishment lands — so the stock on your shelf has to carry roughly 75 days of demand, not 30. Forecast 30 and you have sized the order against 40% of the exposure you are actually carrying. (Illustrative numbers.)
The corollary is the useful part: the horizon is not a fact about your business, it is the sum of two things you can change. A faster supplier shortens it; so does buying more often — increasing review frequency reduces the inventory you need to hold and makes you more responsive. If a 75-day horizon feels further out than you can forecast credibly, the fix is a shorter cycle, not a shorter forecast.
Longer horizons cost accuracy, so spend only what you need
There is a real reason people flinch at long horizons: forecast accuracy degrades as the horizon lengthens. Every additional week is more room for a trend to bend or a season to arrive early — forecasts naturally become more accurate closer to the sales period.
That is an argument for not forecasting further than you need. It is not an argument for forecasting less than you need. A planning horizon shorter than the cumulative lead time leaves part of the cycle unplanned: you have committed to demand you never projected. Truncating does not remove the uncertainty; it moves it out of the forecast and into the stockout.
And the two errors are not symmetric. A longer horizon carries a known cost you can buffer against. A short one carries a hidden one: the forecast looks tight and clean, and the demand it could not see arrives later as a shelf gap you blame on something else. It also distorts how you grade yourself — judging a forecast at a 30-day lag when you order 75 days out measures a number nobody bought against. Measure accuracy at the lag where the decision was made, which your lead time sets.
| Horizon shorter than the wait | Horizon = lead time + order cycle | |
|---|---|---|
| Covers the supplier wait | Partly, at best | Yes — by construction |
| Covers the gap to your next order | No — the review cycle is invisible to it | Yes — the review period sits inside the window |
| Demand it cannot see | Everything past the window edge — already committed | Only what falls beyond the next replenishment |
| Accuracy cost | Low — but precise about the wrong span | Higher at the far end; a known cost you buffer against |
| How it fails | Under-orders long-lead items; they go dark between arrivals | Order sized to the exposure you actually carry |
One horizon for the whole catalog is wrong for most of it
If your suppliers have different lead times, no single horizon is right for all of them. A domestic supplier who ships in a week and an overseas one on a ten-week ocean route are two different buying questions wearing the same label. Set the horizon for the long one and you over-order the short one; set it for the short one and the container SKUs go dark between arrivals. So do the bucketing yourself: group the SKUs that share a wait, and run the forecast once per group over that group's window.
Picking the window is the judgement call; what happens after it is mechanical. Assisty forecasts what each product — and each variant — will sell over the horizon you choose, and because buying decisions happen at variant level, so do the forecasts. Out-of-stock days can be excluded, so the projection reflects real demand rather than the days you had nothing to sell. The forecast then marks which items run out before the period ends and which are overstocked, and at-risk items carry straight into the Smart Buying Table — seeing a risk and fixing it are the same motion.
| Step | What Assisty does |
|---|---|
| Pick the window | Forecasts what each product — and each variant — will sell over the horizon you choose (Forecasting) |
| Judge the demand | Out-of-stock days can be excluded, so the forecast reflects real demand (Forecasting) |
| Spot the exposure | Marks which items run out before the period ends and which are overstocked (Forecasting) |
| Hand off | At-risk items carry straight into the Smart Buying Table (Forecasting) |
| Size the order | Every suggestion is computed from your sales velocity, lead times, and incoming stock — then rounded to your case sizes and capped at your maximum stock levels (Smart Buying Table) |
| Test a change | Scenarios apply instantly — the forecast and the reorder needs update in front of you, no spreadsheet copies (Forecasting) |
See how the projection is built in Forecasting, and where the horizon turns into quantities in the Smart Buying Table — whose math panel walks its own example through a 14-day cover period, an illustration of the arithmetic rather than a number to adopt.