Running a Cafe by the Numbers: The Operator's Guide

By 7 AM you have already made the three decisions that decide the week: what to order, who opens Saturday, and whether the new drink stays on the menu. Most owners make all three from memory.

Search for advice on running a cafe and you will find guides with thirteen sections, four of them about hiring, one paragraph about inventory, and zero mentions of par levels. That ratio is not an accident. Those guides are written by companies that sell scheduling software, so the guide bends toward the schedule. I run a cafe, and my week does not look like their table of contents. Staffing matters, and it gets its own honest section below. But the work that decides whether the month ends in the black is quieter: counting what you have, knowing what each drink costs to make, and ordering the right amounts before each supplier's deadline.

This guide is the spine of that work, with the deeper method for each piece linked along the way.

The three numbers that run the place

Your register already tells you what sold. That number is loud, it feels like the score, and it is the least useful of the three.

The numbers that run the place are quieter:

1. What you used. Sales minus shelf. Twelve gallons of whole milk left Monday; the register only explains nine of them. The gap between what you counted and what you sold is where waste, spills, and unrecorded drinks live.
2. What each drink costs to make. Not the category average. The actual cost of the 16oz oat latte, syrup and lid included. Recipe costing turns "we are busy" into "we are busy selling the drink that earns the least."
3. What to order. Usage over time divided into each supplier's calendar. Get this one right and the other two start paying rent; get it wrong and you meet it again as an emergency run at corner store prices.

Everything else in this guide is a method for producing one of those three numbers on schedule.

Count first: everything starts at the shelf

Every number above starts with a count. Skip the count and the rest of the system runs on fiction.

A count does not need to eat your morning. A recent full count at my shop was 58 items in under 10 minutes, done on a phone, walking the shelves in the order the stockroom is laid out. The method is not clever: count in shelf order, not list order, keep the item list tight, and use a count sheet built for a cafe instead of a blank spreadsheet.

Cadence matters more than heroics. Monday, Wednesday, Friday covers a cafe well: Monday reads the weekend damage, Wednesday catches the drift, Friday sets up the next weekend. And the count does not have to be yours. My manager has run every scheduled count for months, 26 out of 26, without me standing next to him. Handing off the count is the first delegation that actually returns time, because the team sees counts and tasks while the money stays behind your login.

Know what every drink costs you

If someone asks what your best seller earns you and your answer starts with "probably," you have a guess wearing a number's clothes.

Costing a drink is arithmetic, not accounting: every ingredient at its per-unit cost, summed per recipe. A recipe costing template gets you through the menu in an afternoon. What the afternoon buys you:

• Your real food cost percentage, computed against a denominator you trust, instead of an industry number you inherited.
• The difference between your best-selling drink and your best-earning drink. They are rarely the same drink.
• Prices set from cost, so a supplier raising a case price by two dollars becomes a pricing decision you make once, on purpose.
• An early answer when food cost creeps up, because you can see which recipe moved instead of interrogating the whole menu.

The costing only stays true if it reflects how drinks actually leave the bar. At my shop every espresso drink is a double shot; if I costed singles, every latte on the menu would look sixty cents better than it is. Cost the drink you actually make, modifiers included.

Order from the count, not the gut

Ordering is where the count and the costs turn into money saved or money parked.

The mechanics: know your usage per day for each item, know each supplier's order deadline and delivery days, and set par levels so the reorder point is a written number instead of a feeling. Then the order writes itself from the count: usage since last delivery, minus what is on the shelf, padded by the par.

Both failure modes cost real money. Order light and you are improvising before the morning rush, paying markup for whatever the closest store carries. Order heavy and the surplus is cash sitting on a shelf until some of it expires. Long-lead items punish guessing hardest; matcha ordered on feel either runs out mid-week or ages in the cabinet.

Seasons move the targets. Iced season roughly doubles cold cup usage and shifts milk volume, and the time to reset pars for the season is three weeks before it arrives, not the first hot Saturday.

Read the week like a ledger

A cafe produces a week of data whether you read it or not.

The reading habit is short. Five reports, weekly: sales, usage, cost, labor, waste. A daily profit number small enough to check at close. A Monday briefing that takes minutes because the numbers were kept all week instead of reconstructed.

Two patterns pay for the habit by themselves. First, day-of-week shape: your Tuesday is not your Saturday, so averaging them produces an order that is wrong twice. Second, trend versus blip: thirty days of data is enough to tell a rush from a fluke, and a month of history makes next month legible instead of a hope. When you want the fine grain, the margin on individual transactions shows which orders, hours, and modifiers actually carry the day.

Staffing, honestly

Here is the honest version of the section the other guides lead with.

Labor is usually a cafe's largest cost, and it deserves numbers too: a schedule built from your sales curve instead of last month's copy, timecards read against the plan so drift shows up in dollars, and revenue per labor hour as the one benchmark that survives contact with a real shift. Your team cares about their own numbers more than yours; showing tips and hours plainly buys more goodwill than a pizza party.

What I will not tell you is that software runs your staffing. Mine reads the labor numbers; it does not write the schedule or tell anyone to go home. The staffing decisions stay human. The numbers just mean you make them with your eyes open, and a shift with a plan beats a shift with a vibe every single week.

Waste is a number too

Waste hides because nothing rings when it happens. Milk goes down the drain a splash at a time; pastries age out one case at a time; nothing appears on any receipt.

The count catches it anyway. Waste is the residue left when you subtract sales from usage, and tracked over a few weeks it stops being a vague guilt and becomes a list: which items, which days, how many dollars. A cafe that has never measured usually finds its waste concentrated in two or three items, which means the fix is small and specific instead of a lecture to the whole team.

What 7 AM looks like when the numbers run it

The point of all of this is a shorter morning, not a longer one.

Tuesday, 6:55. The count from last night is already in, because the closer ran it in shelf order on a phone. The order draft is waiting: usage since Friday, minus the shelf, against pars, sorted by which supplier's deadline comes first. You scan it, change two lines because you know a catering order is coming, and send it before the first customer orders a cortado. The register can have the rest of the day.

That is the whole system: count, cost, order, read the week. Start with the count. Everything else is derived.