Food Costing & Plate Cost Analysis
Plate cost is the most important number on every dish you serve and the most consistently fudged. Too many restaurants set a price once, never re-cost, and discover six months later that beef shoulder went up 40% and their signature dish is now losing money on every fire. AI lets you re-cost any plate in under 90 seconds โ provided you give it real invoice prices.
What You'll Learn
- The food cost equation and how to use AI as a calculator
- A reusable plate-cost prompt for any dish
- How to factor yields, trim, and waste into a true cost
- How to use AI to do menu engineering (Stars, Plowhorses, Puzzles, Dogs)
- A monthly re-costing cadence chefs actually keep
The Food Cost Equation
Plate cost = sum of (ingredient as-purchased cost ร portion size รท yield) for every component on the plate, plus a small allowance for "below-the-line" items like sauce demis, garnish trim, salt, oil, and herb.
Food cost percentage = plate cost รท menu price.
Industry-standard targets vary: 28-32% is typical for casual dining; 22-28% for fine dining (where labor is higher and food cost has more room); 35%+ for steakhouses where protein dominates and prices justify it.
AI can run the math. You must supply the prices.
The Reusable Plate-Cost Prompt
Act as my food cost analyst.
Dish: Pan-seared scallops, cauliflower puree, brown butter, hazelnut
crumble, sherry reduction.
Per-portion build (weight in grams):
- U-10 dry scallops: 120g (3 each)
- Cauliflower puree: 90g (made from cauliflower 0.4 yield to puree, plus
butter, cream, salt)
- Brown butter: 15g
- Hazelnut crumble: 12g
- Sherry reduction: 10g
- Micro herb garnish: 2g
- Plate sauce, finish oil, plating salt: assume 0.50 total
Current invoice prices (my actual costs this week):
- U-10 dry scallops: $24.00/lb
- Whole cauliflower: $2.10/lb (yield to puree ~40%)
- Whole butter unsalted: $3.80/lb
- Heavy cream 40%: $4.50/qt
- Hazelnut, raw: $9.50/lb
- Panko: $2.20/lb
- Sherry, dry: $14/750ml
- Demi glace (made in-house): assume $4/qt for cost purposes
- Micro herbs: $25/clamshell, ~80g usable
Calculate:
1. Cost of each component per plate
2. Total plate cost
3. Food cost % at our $42 menu price
4. The menu price needed for a 28% food cost
5. The menu price needed for a 30% food cost
Show your math. Note any assumption I should sanity-check.
The output is a clean cost table you can drop into your recipe-card system or your POS.
Yields and Trim Matter More Than You Think
If you buy whole cauliflower at $2.10/lb and your puree yield is 40%, your true cauliflower cost in the puree is $5.25/lb. The unaware chef costs the dish at $2.10/lb and quietly loses 60% of expected margin on every plate. AI handles this if you supply the yield โ which is why the prompt above includes it.
Common yields to know (verify with your own butcher tests):
- Whole salmon, head-on, gutted: ~55% edible
- Whole chicken to cut parts: ~70% edible
- Cauliflower to floret: ~55%; to fine puree: ~40%
- Strawberry, hulled: ~85%
- Whole pineapple, peeled and cored: ~50%
- Spinach, raw to sauteed: ~25% (you lose 75% of weight)
- Mushroom, raw to sauteed: ~50%
Ask AI to build you a yield reference table for your common items, then run your own butcher tests to confirm.
Menu Engineering with AI
Once you have plate costs for every dish, classic menu engineering uses your POS data to put each item in one of four boxes:
- Stars โ high margin, high popularity. Protect and promote.
- Plowhorses โ low margin, high popularity. Re-engineer cost or modestly raise price.
- Puzzles โ high margin, low popularity. Reposition on the menu, ask servers to recommend, or rework the description.
- Dogs โ low margin, low popularity. Cut.
The AI prompt:
Below is my menu with plate cost, menu price, contribution margin per
dish, and units sold over the last 30 days from POS.
[paste data: dish, plate cost, menu price, margin, units sold]
Classify each dish into Star, Plowhorse, Puzzle, or Dog using popularity
above/below the menu median and margin above/below the menu median.
For each dish, write a 1-sentence action recommendation:
- Stars: protect (sometimes a small price increase)
- Plowhorses: cost re-engineer or +$1-2 menu price
- Puzzles: reposition on menu, server upsell, rewrite description
- Dogs: cut or replace next menu refresh
Output as a markdown table.
Run this monthly. The pattern of which categories your menu drifts toward tells you a lot โ too many Puzzles means your menu is over-creative for your audience; too many Plowhorses means your prices haven't kept up with cost.
The Monthly Re-Costing Cadence
Pick the first Monday of every month. Block 90 minutes. Re-cost the top 10 dishes by sales volume using current invoice prices. If any have moved more than 1.5 percentage points in food cost, make a decision โ re-engineer, raise price, swap component, or accept the new margin.
This cadence is what separates chefs who hit their numbers from chefs who don't. AI takes the math from 90 minutes of arithmetic to 30 minutes of judgment.
A Note on Below-the-Line Costs
Most plate-cost calculations under-count because they ignore "below the line" โ salt, herbs, oils, sauce demis, plating finishes. A reasonable allowance is 3-5% of plate cost added on top. Build this into your prompt so you don't surprise yourself.
Add a flat 4% below-the-line allowance to every plate cost calculation
above to account for plating salt, finish oils, herbs, garnish trim, and
sauce demi components I haven't itemized.
Key Takeaways
- AI runs the food-cost math; you supply real invoice prices
- Always include yields โ buying whole cauliflower at $2.10/lb means $5.25/lb in your puree
- Use the reusable plate-cost prompt for any dish, and ask AI to show the math
- Run menu engineering monthly using POS data โ Stars, Plowhorses, Puzzles, Dogs
- Keep a monthly 90-minute re-costing cadence; it's the difference between hitting and missing your numbers
- Add a 3-5% below-the-line allowance so you don't under-count

