Product 41 (an oak sideboard, Atelier Mira’s bestseller) has started selling at nearly twice its usual pace. Good news, with a question inside it: when does stock become the problem? Ask most software and you get a number. Day 22. Ask SENA and you get something that looks less impressive and is worth more: a range, with its width shown honestly.
The seduction of the single number
A point forecast feels like knowledge. It fits in a cell, survives being pasted into a deck, and gives everyone downstream something firm to plan against. The trouble is that the firmness is an editorial choice, not a property of the future. Somewhere inside every “day 22” is a distribution, a range of outcomes the model actually considers plausible, and publishing only the middle of it deletes the one thing an operator most needs to price: how wrong could this be, and in which direction?
Deleting the width doesn’t make the forecast more accurate. It makes the reader more confident than the forecast ever was. That gap, between the model’s honesty and the cell’s confidence, is where reorders land late, where cash buffers get sized wrong, where “the forecast said we were fine” gets said in meetings.
Eight weeks of history, then the honest part: a forecast that admits how wide it is.
Reading a band like an operator
Here’s what the band actually says. The middle of it crosses Mira’s reorder point around day 22. But the honest window runs from day 16 to somewhere past day 35: if the hot streak holds, the reorder point arrives nearly a week earlier than the “official” number; if it fades, much later. Now add the one fact the chart doesn’t show: the hardwood supplier’s lead time is about ten days, and, as this library’s other pieces cover, that supplier has been slipping.
Run the decision both ways. On the point forecast, day 22 minus ten days of lead time says order early next week; no rush. On the band, the early edge says the risky-but-plausible world hits the reorder point in sixteen days, which makes this week’s order the cheap insurance. Same model. Same data. The width changed the decision. That’s the entire argument for ranges, in one move: a point forecast can only tell you what’s likely; the band tells you what it would cost to be wrong.
A single number answers “what will happen?”, which is not a question anyone can answer. The band answers “what should I do?”, which is the only question that was ever yours.
The band isn’t decoration on top of a point estimate; it’s computed from named sources of doubt: demand variance in the recent history, how thin the data is (one strong week widens it; eight would narrow it), and supplier lead-time variance on the replenishment side. When the band is wide, you can ask why, and get an answer.
Uncertainty is not indecision
The reflex worth resisting is that showing a range is somehow weak: hedging dressed up as method. It’s the opposite. It takes more model, not less, to hold a distribution honestly, and it takes more respect for the reader to show it. What SENA declines to do is convert its own uncertainty into your false confidence, because that conversion always settles in the same currency: a decision you’d have made differently if you’d seen the width. When the numbers firm up (Friday, when one strong week has either become two or hasn’t), the band narrows on its own, and the watchlist entry that’s been tracking Product 41 says so in the morning brief.