Your register counts drinks, not bottles

Someone orders a rum and coke. The bartender pours it, the register rings "1 rum and coke," a card gets charged, and as far as the point of sale is concerned that transaction is finished. It did its job. Somewhere behind the bar, the bottle that pour came out of still has about twenty-four ounces in it, and may sit there part full for another two or three weeks, depending on how often someone orders from it. The register has no way to know that, and it was never built to.

That's the fact everything else here comes back to. A POS decrements units. A bottle isn't a unit, it's a reservoir: one object that gets a little emptier with every pour, sits open on a shelf between pours, and doesn't finish being sold until the last ounce is gone, sometimes weeks after the first one left. Ring up a sale and the system can tell you a drink went out the door. It can't tell you what's actually left in the bottle that supplied it, because nothing about a sale event carries that information forward.

A Negroni pulls partials from three bottles. One drink, three spirits, each pulled at a different rate. The register logs "Negroni, 1." It has no idea it just touched three separate reservoirs, so nothing about that sale tells you the gin moved faster than the vermouth this week, or that the Campari you're running low on came almost entirely from cocktails, not from anyone ordering it on its own.

A single is usually 1.5 ounces. Not always. The POS records the sale at whatever size is on the menu. The bottle empties at whatever size actually left the spout, and those two numbers drift apart a little more with every shift.

"Make it a double" changes two things, and the register only sees one. The price goes up, that part's easy. The depletion should double too, and on most systems it doesn't, because the item on the register is still just "Negroni" with a modifier, not two units of gin coming out of the well.

Spills, comps, staff drinks, and tastings are real consumption with no sale attached. A round on the house, the pour someone tastes to check a keg, the glass that gets knocked over: all of it is liquid actually leaving a bottle. None of it rings anything up, so none of it shows up anywhere the register looks.

None of this is dramatic on any single night. A quarter ounce here, a comp there, a double that only counted once. The cost is that you can't see it. To know what a bottle should have left, you need every drink that draws on it, and the register only tells you drinks.

You can work it out. Take the gin. Find every item on the register that pours from it, and what each takes:

That is one bottle, for one week. Now do the vodka, the rum and everything else on the back bar. So most weeks you take the count as the answer, maybe with a hunch that the gin is moving faster than it should, and nothing to hold that hunch against.

Your pour cost, your margin on a category, the variance report that says you're three bottles short: all of them start from a picture of the shelf that the register was never able to see. You aren't being lied to. You are being told, accurately, about something else.

So we stop treating a sale as one event and start treating it as what it actually was. The register says "Negroni, 1." A Negroni is equal parts, an ounce each of gin, Campari and sweet vermouth, so that sale was three withdrawals from three different reservoirs, not one unit of anything.

We hold that mapping, so a sale coming off the register turns into amounts leaving specific bottles. A single of the same gin takes an ounce and a half. A double takes three. The Negroni takes one. All of it comes off the same gin, because it is the same gin.

The register carries on doing the job it is good at, which is telling us what sold. It was never built to know what each sale used up from the bottles.

What those withdrawals add up to is a question about the bottle rather than the register, which is why counting containers is the wrong unit. This gap is also part of why a pour cost worked out in January is already suspect by March, even when nobody has done anything wrong. See the arithmetic, and why nothing tells you to redo it.