Delegating the Cafe Inventory Count to Staff

The count is the first chore worth handing off

It is 7:10 on a Wednesday. The cafe opens at 8. You should be dialing in the grinder and pulling the first shots, but instead you are on your knees in the reach-in counting oat milk, because it is a count day and the count is yours.

It should not be. The count is repetitive, it runs on a schedule, and it needs presence more than judgment. About 60 items, three times a week, Monday, Wednesday, and Friday, and none of it requires the owner's specific read of the business. It is the first chore you should hand off.

Owners hold onto it anyway, for a reason rarely said out loud. You keep the count because you do not trust the number your manager will write down. A wrong count means a wrong order: an emergency run at corner-store prices, or three extra cases of oat milk on a shelf. So you keep counting, and you lose the first hour of your Wednesday.

Delegation is a verification problem, not a trust problem

Search for how to hand off inventory and you will get restaurant coaching about letting go: build trust, train thoroughly, learn to hand over the reins. It is the wrong frame for a cafe. Trust is a feeling, and you cannot order supplies on a feeling.

Here is the frame that works. You do not need to trust the count if you can check it. The question stops being "do I believe my manager counted carefully" and becomes "can I see who counted, and does the number hold up against what we sold."

Two mechanics make a delegated count safe. The first is attribution: every count has a name on it. The second is reconciliation: every count checks itself against what your Square sales say should have left the shelves. Get those two, and the person doing the count stops being a matter of faith and becomes a matter of record.

Attribution: every count carries a name

When your manager counts oat milk and writes down six, that entry is stamped with who wrote it and when. If someone corrects it later, from six to nine, the change is logged with the old value, the new value, and the person who made it. Nothing gets quietly overwritten. The record of what was counted, and what got changed, survives.

It is the same reason a good count records what you see, not what you expect: a number is only useful if you know where it came from. When every entry carries a name, a strange count is a question you can ask a specific person on a specific morning, not a mystery you inherit at the end of the month.

At my own shop, as of this July, the manager has run the last 26 of 26 counts. I have not knelt in the reach-in in weeks. What I have instead is a list of counts, each with his name on it, each one I can open and read. The handoff was real, and I did not have to give up knowing what happened.

Reconciliation: the count checks itself against what Square sold

Attribution tells you who counted. Reconciliation tells you whether the count is right, and it does it the same morning, before the order goes out.

The logic is simple to state. Your last count, plus whatever got delivered, minus what your sales should have used, gives you an expected number for each item. Your Square sales are the input to that "should have used" figure, mapped through each drink's recipe and every modifier on the ticket. An iced matcha latte with oat leaves a scoop of matcha and a pour of oat milk, and only a system reading the modifiers knows that. When the counted number and the expected number disagree, something is worth a look.

That gap catches two very different problems, which is exactly why it is useful. Sometimes the count is off: a case in the back never got tallied, or a digit got fat-fingered. Sometimes the count is right and the problem is real, and the reconciliation is the only reason you find out. In my own shop, computed usage for whole milk once read 0.70 bottles a day when the real burn was 9.716, a fourteen-times miss. The count was fine. A set of Square modifiers had been deleted, so the math behind "should have used" had gone blind to every drink that took whole milk. We found the broken modifier mapping and fixed it. Without a number that checks itself, that drift hides until the fridge is empty on a Saturday.

For a delegated count, this is the safety rail. A manager can run the whole thing, and if a bad entry slips in or a real problem surfaces, it shows up the same day next to the item, not four weeks later when you reconcile the spreadsheet by hand.

Keep the money behind your login

Handing off the count does not mean handing off the books. This is where the permission boundary earns its keep.

Your manager sees what they need to do the job: the item list, the counts, the tasks, the supplier order that comes out the other end. They do not see what each item costs, what the margins are, what anyone is paid, or the billing. Those stay behind the owner login. A manager can count 60 items and send an order grouped to each supplier's cutoff without ever seeing that the whole milk costs you a dollar-something a bottle or what the closing barista makes an hour.

That boundary is also where honesty about labor lives. Parly shows you labor numbers, computed from your Square timecards joined to your sales, and it shows them to you, the owner. It does not build the schedule, run payroll, or manage the person doing the count. It will tell you what a shift cost against what it sold, the same way timecards reveal the gap between the schedule and reality. What you do with a person's hours is yours. The tool reports; it does not manage.

What the handoff looks like on a Wednesday morning

Same Wednesday, one change: it is your manager in front of the reach-in, not you.

He opens his phone and taps through the list in walk order, beverages, then paper goods, then chemicals, then pastries. A recent full count at my shop was 58 items in about 8.5 minutes, from a phone. He submits. Reconciliation runs against the week's Square sales and flags one item where the counted number and the expected number are far enough apart to check. He recounts it, finds a case he missed in the back, and fixes it. That correction is logged with his name.

The confirmed count drafts a supplier order, grouped by supplier and timed to each cutoff, and it lands on your phone. You are at the bar dialing in the grinder, which is where you should have been at 7:10 the whole time. You read the order, adjust one quantity, and send it before the Odeko cutoff. The count got done, it got checked, and the money stayed yours to see.

If you want to make this move, do it this week. Pick the next count day and hand the phone to your manager. Do not take it back to "just double-check." Watch the reconciliation flag instead of watching the person, because the flag is what tells you whether the number is right. Once a count is attributed and reconciled against your own Square sales, you have a number you can trust without writing it down yourself. Build the rest of the week on it, and when you sit down to write next week's schedule, you are working from real demand instead of a guess.