Divide each tub's price by the scoops you really get from it, not the scoops the arithmetic promises, then add the cone or cup, napkin and toppings to cost every serve. Price each item so its cost stays under the share of sales your own overheads allow. AI builds the sheet; you supply measured yields and invoice prices.
Price calculators ask you for a cost per scoop, which leaves the step that moves the answer most to you: yield. A 5-litre tub and an 80 ml scoop suggest 62 scoops, and a real tub gives fewer, because of heaped scoops, the layer left on the tub walls, tasting spoons and drops. Skip it and every price downstream rests on a number the shop never achieves.
Measure scoops per tub before you cost anything
Portion scoops are numbered by how many fit into a quart, so a #12 holds about 80 ml (2.7 fluid ounces) and a #16 about 59 ml. Capacity varies a little by maker, so fill yours from a measuring jug of water once and write the true figure on the sheet. Choose the scoop you use for a single serve and use only that scoop for the test.
- Open a fresh tub at normal serving temperature and serve it through ordinary trading, tasting spoons included.
- Keep a tally sheet by the freezer. Every scoop counts, including the ones that were dropped or remade.
- When the tub is scraped out, write the total beside the flavour and the date.
- Do three tubs from each family of flavours, because smooth flavours and ones with inclusions scoop differently.
Here is an illustrative result for a 5-litre tub and an 80 ml scoop. The ceiling is 5,000 ml divided by 80 ml, which is 62.5 scoops.
| Tub | Scoops served | Share of the 62.5 ceiling |
|---|---|---|
| A | 54 | 86% |
| B | 56 | 90% |
| C | 55 | 88% |
| Average | 55 | 88% (12% under the ceiling) |
Use the average for costing. Keep the lowest result as a tripwire: if tubs in service later come in three or more scoops under it, look at portioning before you blame the supplier.
Drift is the quiet leak. Staff who heap scoops to be generous move an 80 ml scoop towards 90 ml without anyone deciding to. On a $32 tub that cuts the yield from 55 to about 49 and lifts the cost per scoop from 58 cents to 65 cents, which is 12.5% more. Across 1,800 scoops a month at the blended cost used further down, that is about $156 leaving the shop with no change to the menu.
Turn invoice lines into unit costs, then check the AI's sums
Costing needs a price for one single thing: one cone, one millilitre of sauce, one gram of topping. This is the dull part, and a chat assistant such as ChatGPT, Claude, Gemini or Copilot does it quickly if you make it show its working. Type in lines from your last supplier invoices (or photograph them and read the transcription back before you trust it), and ask for a multiply-back check.
Here are lines from my ice cream shop's supplier invoices. For each line give: pack size, pack price, unit (each, ml or g) and cost per unit to four decimal places. Then multiply each unit cost back by the pack size and show that it matches the pack price. If a line is ambiguous, ask me instead of guessing.
Vanilla ice cream, 5 L tub, 2 tubs, $64.00
Wafer cones, case of 200, $28.00
Waffle cones, case of 120, $31.20
Chocolate sauce, 2.5 L, $14.00
Rainbow sprinkles, 1 kg, $6.80
Chopped nuts, 1 kg, $14.00
8 oz paper cups, sleeve of 50, $5.90A realistic first answer is right on six lines and wrong on one. This is an illustrative output, with the check applied by hand:
| Line | Unit cost the assistant gave | Correct? |
|---|---|---|
| Vanilla tub | $32.0000 each | Yes (two tubs for $64.00) |
| Wafer cone | $0.1400 each | Yes (200 x $0.14 = $28.00) |
| Waffle cone | $0.2600 each | Yes (120 x $0.26 = $31.20) |
| Chocolate sauce | $0.0560 per ml | No: should be $0.0056 |
| Sprinkles | $0.0068 per g | Yes |
| Chopped nuts | $0.0140 per g | Yes |
| Paper cups | $0.1180 each | Yes |
The assistant read 2.5 L as 250 ml, and its own multiply-back agreed with the slip, so asking for the check was not enough on its own. Two habits catch this. The first is plausibility: at $0.056 a millilitre, a 30 ml drizzle of sauce costs $1.68, and nobody believes that. The second is doing the multiplication yourself with the pack size printed on the invoice: 2,500 x 0.056 is $140, not $14.
Once the unit costs are right, have the assistant lay the sheet out.
Set up a costing sheet in Google Sheets with four tabs: Ingredients (item, pack size, pack price, unit cost), Tiers (tub cost, tested scoops per tub, cost per scoop), Menu (one row per component with quantity and cost, totalled per item) and Summary (cost, price, cost share, floor price at a 30% cost share). Give the formulas for row 2 of each tab and say which cells I type into.Enter one item by hand first, such as the single cone, and confirm the sheet agrees with your own sum before you fill in the rest. The same layout works in Excel; the recipe costing spreadsheet tutorial goes further on ingredient-level recipes if you also run a kitchen.
Cost the flavours in tiers, then blend by what sells
Not every flavour costs the same, and one price per scoop hides that. Group your flavours into three tiers by tub price, run the yield test on each family, and work out the cost per scoop for each. The mix you actually sell, read from a till report, turns three numbers into one blended cost for the standard menu.
| Tier | Tub cost | Scoops per tub (tested) | Cost per scoop | Single cone cost | Share of a $4.00 price | Share of scoops sold |
|---|---|---|---|---|---|---|
| Standard (vanilla, chocolate) | $32.00 | 55 | $0.58 | $0.74 | 18.5% | 60% |
| Mid (cookie dough, salted caramel) | $41.00 | 55 | $0.75 | $0.91 | 22.6% | 25% |
| Premium (pistachio) | $58.00 | 55 | $1.05 | $1.21 | 30.4% | 15% |
The single cone cost includes a $0.14 wafer cone and a $0.02 napkin. Weighting the three scoop costs by the sales mix gives a blended $0.69 per scoop, and that is the figure used for the rest of the menu.
The premium tier is the finding. At a flat $4.00, a pistachio cone earns 47 cents less gross profit than a vanilla one, and its cost share is already at 30.4%. A 75-cent premium-flavour surcharge takes it to $4.75, a 25.6% cost share, and puts the extra cost on the customer who chose it.
Cost every item on the menu, cone to pint
Build one row per thing you sell, with every component on it, including the ones that feel too small to count. A napkin is small. A sundae carries eight small things, and together they cost 94 cents.
| Item | What goes in | Cost |
|---|---|---|
| Single cone | 1 blended scoop $0.69, wafer cone $0.14, napkin $0.02 | $0.85 |
| Double cone | 2 scoops $1.39, wafer cone $0.14, napkin $0.02 | $1.55 |
| Sundae | 3 scoops $2.08, sauce (30 ml) $0.17, whipped cream $0.15, nuts (15 g) $0.21, cherry $0.06, wafer stick $0.08, sundae cup $0.22, spoon $0.03, napkin $0.02 | $3.02 |
| Hand-packed pint, standard flavour | 473 ml at $0.0064 a ml $3.03, tub and lid $0.45, label $0.05 | $3.53 |
Two things show up once everything sits on one sheet. The cone is paid for by the first scoop, so the second scoop costs only the ice cream, $0.69, and a double earns more dollars than a single at a similar cost share. And the pint has no cone, yet it hands over 473 ml in a single sale, so its cost share runs higher than the scoop items.
The premium tier changes the pint more than the cone: on the same arithmetic a pistachio pint costs $5.99 and a mid-tier pint $4.38, so pints need a price per tier, not one price.
The pint line is costed at the tub's full volume, which assumes nothing is lost to the tub walls and the packer does not press in extra. Packing squeezes out air pockets, so a firmly packed pint can hold more ice cream than its 473 ml suggests. Check it by weight: weigh a full tub, subtract the weight of an empty one and divide by 5,000 to get grams per millilitre, which tells you what a pint should weigh. Then weigh five packed pints. If they average 10% heavy, the pint's ice cream costs 10% more, about $3.33 instead of $3.03, and the sheet should say so. Use your container's real volume, too: an imperial pint is 568 ml.
Where the margin target comes from
A target taken from someone else's guide is someone else's shop. Work backwards from yours: take three months of accounts, write each overhead as a share of sales, and see what is left for ingredients and packaging.
| Block | Share of sales (illustrative) |
|---|---|
| Staff wages | 30% |
| Rent and service charge | 11% |
| Utilities, insurance, card fees, software, repairs | 12% |
| Owner pay or profit target | 14% |
| Total before ingredients | 67% |
| Left for ingredients and packaging | 33% |
The 33% is the most ingredients and packaging can take. I would round it down to 30% to leave room for waste, staff scoops and the odd free cone, and treat 30% as the ceiling. For any item, cost divided by 0.30 gives the lowest price that stays under it. If your overheads are heavier than these, the ceiling falls, so prices have to rise or costs fall; an assistant cannot invent your accounts, so the shares must come from them.
Set prices between your floor and your neighbours' ceiling
The floor comes from your sheet; the ceiling comes from the street. An assistant has no view of other shops' boards or of what your customers will pay, and one that quotes local prices is guessing. Walk to three other shops, photograph their price boards, and type the single, double, sundae and pint prices into the sheet.
| Item | Cost | Floor at 30% | Neighbours (illustrative) | Price now (cost share) | New price (cost share) |
|---|---|---|---|---|---|
| Single cone | $0.85 | $2.85 | $3.75 to $4.50 | $4.00 (21.3%) | $4.00 (21.3%) |
| Double cone | $1.55 | $5.16 | $6.00 to $7.00 | $6.50 (23.8%) | $6.50 (23.8%) |
| Sundae | $3.02 | $10.06 | $9.00 to $11.00 | $9.50 (31.8%) | $10.50 (28.8%) |
| Pint, standard flavour | $3.53 | $11.76 | $10.00 to $12.50 | $10.50 (33.6%) | $12.00 (29.4%) |
Two items fail. The sundae sits at 31.8% and the pint at 33.6%, both over the ceiling, and both can move to a price that still lands inside the neighbours' range. The cones pass with room to spare, but that is headroom, not a reason to cut: they already sit inside what the street charges. If a price you want falls outside the neighbours' range, write down why customers would accept it, such as a bigger scoop or a better product, before you print the board.
When the numbers say a price should rise, test the move before committing. The menu price increase tutorial shows how to roll out and measure a change item by item, and the menu engineering tutorial covers which items to promote.
Price the extras so nothing goes out free by accident
Toppings and upgrades are where small leaks hide, because they are rung up inconsistently or not at all. Cost each one from the unit costs you already have, then decide which are free and which are charged.
| Extra | Cost to you | Charge | Cost share |
|---|---|---|---|
| Waffle cone instead of a wafer cone | $0.12 more ($0.26 against $0.14) | $1.00 | 12% |
| Second scoop (single to double) | $0.69 | $2.50 more ($4.00 to $6.50) | 28% |
| Sprinkles, 10 g | $0.07 | $0.75 flat for any topping | 9% |
| Sauce drizzle, 25 ml | $0.14 | $0.75 flat for any topping | 19% |
| Chopped nuts, 15 g | $0.21 | $0.75 flat for any topping | 28% |
A flat charge for any topping is easier to ring up and explain than three prices, and the worst case, nuts, still lands under the 30% ceiling. Giving sprinkles away is a defensible courtesy at 7 cents; if 540 serves a month carry them, that is about $38. Giving nuts away at 21 cents would cost about $113 a month on the same volume, which is why the rule should be written down and not left to whoever is on shift.
Gross profit per minute at the counter
Cost share says what an item costs. It does not say what the item takes to serve, and in a summer rush the counter is the limit. Time ten of each item with a phone stopwatch. Illustrative timings: a single takes 45 seconds, a double 55, a sundae two minutes, a packed pint a minute.
- Single cone: $3.15 gross profit in 45 seconds, about $4.20 a minute.
- Double cone: $4.95 in 55 seconds, about $5.40 a minute.
- Sundae at $9.50: $6.48 in two minutes, about $3.24 a minute. At $10.50 it is $7.48, about $3.74 a minute.
- Pint at $12.00: $8.47 in a minute, the most of any item.
The sundae fails twice: it is over the cost ceiling and it earns the least per minute of counter time. The pint is over the ceiling at the old price but earns the most per minute, which is why an item above the target is not automatically a problem. I would lead the rush-hour board with cones and pints and keep the sundae as the sit-down order.
Ask AI to stress-test the sheet
Once the sheet is right, the assistant earns its place by running scenarios faster than you would by hand. Give it the table and be explicit about what changes.
Below is my costing sheet (item, cost, price). Run three scenarios separately and show the new cost share and gross profit per serve for each item:
1. Ice cream tub prices rise 8%. Apply the rise to the ice cream lines only, not to cones, sauces or packaging.
2. Scoop drift: the yield falls from 55 to 49 scoops per tub.
3. I raise the single cone from $4.00 to $4.25. Tell me how far single-cone sales can fall before I earn less gross profit than now.
Show every calculation.
[paste the table]A common first answer applies the 8% to every ingredient, cones and cups included, which overstates the damage; the instruction about the ice cream lines only is there to stop that. With it, the illustrative output looks like this:
| Scenario | Single cone | Double cone | Sundae (at $10.50) | Pint (at $12.00) |
|---|---|---|---|---|
| Today | $0.85 (21.3%) | $1.55 (23.8%) | $3.02 (28.8%) | $3.53 (29.4%) |
| Tubs up 8% | $0.91 (22.7%) | $1.66 (25.5%) | $3.19 (30.3%) | $3.77 (31.4%) |
| 49 scoops a tub | $0.94 (23.5%) | $1.72 (26.4%) | $3.27 (31.2%) | not affected by scoop size (check pack weight) |
Read it as a plan. An 8% rise in tub prices pushes the sundae and the pint back over 30%. Getting back under takes $10.75 for the sundae (29.6%), still inside the neighbours' range, and about $12.60 for the pint (29.9%), just above it; given what the pint earns per minute, leaving it a little over the ceiling is a defensible choice. Scoop drift hurts the sundae most, because it carries three scoops. Both results point to the same habit: re-run the sheet when an invoice changes, not at year end.
For the third scenario the assistant should return a break-even sales drop of 7.4%. The sum: the single cone's gross profit per sale is $3.15 now and $3.40 at $4.25, so profit stays level until sales fall to 3.15 divided by 3.40, which is 92.6% of today's. A 25-cent rise earns more unless more than about 7% of single-cone customers walk away. A drop of that size is visible in the till, so the check is yours to make: compare single-cone sales for the two weeks after the change with the two weeks before.
A week to get from invoices to a new price board
Allow four to five hours of your own time spread across a week, because the longest job, the yield test, happens while you serve.
- Day one, about 90 minutes: collect the latest invoice from each supplier, run the unit-cost prompt, and fix any line that fails the multiply-back or the plausibility check.
- Days one to five, no extra time: tally scoops on three tubs per tier during normal service.
- Day five, about an hour: build the sheet with the layout prompt and enter the single cone first.
- Day six, about an hour: pull overhead shares from three months of accounts and set the cost ceiling.
- Day seven, about 45 minutes: photograph three neighbours' boards, fill in the price table, and pick the changes.
Change two or three prices at a time, not the whole board, so the till can tell you what each move did. If the board feels crowded with changes, start with the item furthest over the ceiling.
A month-end check that the sheet matches the till
A costing sheet is a claim about how the shop behaves, and the till is the evidence. Once a month, compare scoops available with scoops sold. Tubs finished, leaving out any used to pack pints, multiplied by 55 gives what the sheet expects; the till gives what was rung up, counting a double as two scoops and a sundae as three.
Illustrative month: 38 tubs finished, so 2,090 scoops available, against 1,980 scoops sold through the till. The gap is 110 scoops, 5.3%. The yield test already includes tasting spoons and drops, so the gap points at drift, staff scoops or free ones given away. My starting rule is to look into portioning when the gap sits above about 5% for two months running, then adjust the rule to your own history.
If your gap is small but your margins still feel thin, the sheet is probably missing a cost: a sauce, a lid, a delivery charge. Anything sold on delivery apps needs its own price, because commission comes off the top; the delivery app markup tutorial works through that. And once costs are settled, the sensible next job is ordering: the tub forecast by flavour and weather uses your yields to say how many tubs to buy.
The cost per cup tutorial applies the same sheet to drinks, if the counter sells both. Whatever you sell, the order stays the same: measure, cost, compare with the street, then change prices one item at a time and watch the till.
Questions about costing ice cream
How many scoops should I get from one tub?
Divide the tub volume by the volume of your scoop to get the ceiling, then expect fewer. Heaped scoops, the layer left on the tub walls, tasting spoons and drops all take a share. Serve three tubs from each flavour family through normal trading, tally every scoop, and cost from the average rather than the arithmetic.
Should every flavour cost the same per scoop?
No. Tub prices differ, so cost per scoop does too. Group flavours into tiers, price the standard menu from a blended cost weighted by what sells, and check the premium tier separately. If pistachio costs far more than vanilla, a small flavour surcharge puts that extra cost on the customer who picks it.
Can AI choose my menu prices for me?
It can do the sums, spot items above your cost target and test what a cost rise does. It cannot see what the shop down the road charges or what your customers will accept. Those two inputs, the ceiling on price, come from you walking round and photographing menu boards.
How often should I redo the costing?
Update the unit costs whenever a supplier invoice shows a different price, and redo the yield test when you change scoop, supplier or recipe. Compare the sheet with the till once a month. A full review every quarter is usually enough if those triggers are handled as they happen.
Further reads
- Restaurant Labour Cost Percentage: Test Rota Changes With AI First — Wages are the biggest overhead block in the margin target; test rota changes.
- Ice Cream Shop Marketing Ideas: Weather-Triggered Posts With AI — Bring customers in on warm days once the prices are settled.
- Food Truck Pricing: Set Menu Prices After Fees, Fuel and Pitch Costs — The same costing logic with pitch fees and fuel added.
- How Can a Coffee Shop Use AI? Start With Supplier Prices — Start from supplier prices in a cafe, not a scoop shop.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Wikipedia, Disher (portion-scoop sizes by number). Every shop figure in the worked example is illustrative; replace it with your own invoices, yield tests and accounts.