The window is a number of units, not a number of days
A pack arriving with 45 days on it and a three-day pull rule gives a 42-day selling window. A line selling two a week is selling 0.286 a day, so across those 42 days the column can move twelve units. That is the depth the rate reaches. Anything loaded below unit twelve is waiting past the pull date before its turn comes, and the default column is filled to fifteen — three units too deep. The page settles on 0.9 units pulled every fortnight, 23.5 a year, $16.89 at cost on one machine and $202.73 across twelve.
Read that as a share and it is worse than it sounds: 18.4 percent of everything loaded into that column never sells. Take the fill down to twelve and the write-off goes to zero while the sales do not change at all, because the column was never going to get past unit twelve anyway. Raise the rate to six a week instead and the window reaches 36 units, so fifteen deep is comfortable and the loss disappears from the other direction. Same product, same date, same rule — the sell rate is doing all of it.
Why the two failure modes do not average out
Filling shallow costs margin on sales that never happen. Filling deep costs the cost of goods on units that never sell. They are not the same money and they do not offset. Missed demand costs the gap between the vend price and the delivered cost, so on the defaults that is $1.03 a unit; a written-off unit costs the 72 cents you already paid for it. Whichever is larger for your own product decides which side of the crossing you should err on, and for most vended items the missed sale costs more than the write-off — which is the honest argument for erring deep, right up until the rate falls far enough that the tail stops turning at all. On the defaults the table is flat from four deep to twelve, all of them at $107.41 a year, and the page names twelve rather than four because within a tie the deeper column is the one with a buffer.
Loading in front of the old stock
Switch the loading order on the form and the same column behaves differently. Putting the fresh stock in front means the units at the back stop moving entirely, and once they have stopped they age until the pull date takes them, every cycle, forever. Pulling the column forward first costs a few seconds and removes that permanently. It is the cheapest thing on this whole page and it is the one most often skipped when a driver is behind.
Fill depth and service interval are the same decision seen from two sides. This page holds the interval fixed and moves the depth; the par level and restock interval calculator holds the columns fixed and moves the interval, and names the one selection that is setting it. The 72 cents this page uses as the cost of a unit is itself worked out in the delivered cost per vend calculator.
What the number is not
It is a floor. The model sells at a flat rate and a slow column does not sell at a flat rate — it sells nothing for eleven days and four units on a Thursday. Under lumpy demand more of the tail reaches the date than an even rate says, so measure your own over a quarter, put it next to this figure, and treat the difference as the price of your own variability. That is a more useful number than either one on its own.
And the dates themselves are not this page. It does not know what any date on any pack means, it does not say how long anything keeps, and it does not judge whether something is fit to sell. It reads two numbers you supply and multiplies. Anything beyond that belongs to the manufacturer who printed the date and to whoever regulates the site the machine is standing in.
Questions people ask
How deep can I fill a slow vending column?
As deep as your sell rate reaches inside your own selling window, and no deeper without accepting a write-off. Multiply the daily rate by the days between arrival and your pull date. Two a week over a 42-day window reaches twelve units; six a week reaches 36. Beyond that depth every extra unit is a unit that will be pulled rather than sold, so the question stops being about the column and starts being about whether the selection belongs in the machine.
Is a stockout worse than a write-off?
Usually, in money, because a missed sale costs the whole margin and a written-off unit costs only what you paid for it. On the defaults that is $1.03 against 72 cents. But the arithmetic flips as the cost of goods rises toward the price, and it flips again if a customer who finds the column empty stops trying the machine, which is a real cost this page cannot see at all.
Does it matter whether new stock goes in front of or behind the old?
On a fast column, barely. On a slow one it is the difference between a tail that eventually sells and a tail that never moves again. Loading in front leaves the oldest unit permanently at the back, where it ages until the pull rule takes it, on every single cycle. The page will run both orders so you can see the gap in units and in cost for your own line.
Where do the shelf life and the pull rule come from?
From you. The first is the number of days you count between putting the pack in the machine and the date printed on that pack, which changes with every delivery and is worth checking rather than assuming. The second is your own operating rule for how early stock comes out. This page states no shelf life, describes no requirement, and makes no judgement about whether anything is fit to sell.