What this automation does
This automation connects to your POS system, pulls sales data by menu item, combines it with food cost data, and uses AI to perform menu engineering analysis. It classifies each item as a star (high profit, high popularity), plow horse (low profit, high popularity), puzzle (high profit, low popularity), or dog (low profit, low popularity).
Based on this analysis, AI recommends specific actions: rewrite descriptions for puzzles to increase orders, raise prices on stars, reposition plow horses on the menu, and consider removing dogs. It also rewrites menu descriptions using persuasive food copywriting techniques proven to increase order rates by 10-15%.
Tools you need
- POS system with data export: Square, Toast, Clover, or similar — needs sales data by item for the past 3-6 months
- OpenAI API (GPT-4): Analyzes menu data and generates optimization recommendations ($0.05-0.10 per analysis run)
- Google Sheets + Apps Script: Data staging and AI integration — free and works with any POS data export
How to set it up
Step 1: Export sales data from your POS for the last 3-6 months. You need: item name, category, number sold, revenue, and ideally food cost per item. Import this into a Google Sheet with consistent column headers.
Step 2: Add a food cost column if your POS doesn't track it. Calculate the contribution margin for each item (selling price minus food cost). This is the foundation of menu engineering.
Step 3: Write a Google Apps Script that reads the menu data, formats it for AI analysis, and sends it to OpenAI. The prompt should request: menu engineering classification (star/plow horse/puzzle/dog), specific pricing recommendations with reasoning, rewritten descriptions for underperforming items, and menu positioning suggestions.
Step 4: Run the analysis monthly after closing your period. Compare actual results against AI recommendations from the previous month. Track which changes drove revenue increases and feed that back into the prompt.
Cost breakdown
| Item | Cost | Notes |
|---|---|---|
| Google Apps Script | $0 | Free, runs natively in Google Sheets |
| OpenAI API | $5-$15/mo | ~$0.08 per analysis — run monthly or weekly |
| POS data export | $0 | Most POS systems include data export |
| Setup time | 45-60 min | One-time, including data preparation |
| Total monthly | $5-$15/mo | Typically generates 10-15% revenue increase |
Frequently asked questions
At least 3 months of data for meaningful analysis, ideally 6-12 months to account for seasonal variations. If you're a newer restaurant, start with whatever data you have — the AI can still identify patterns with as little as one month of data, but recommendations improve with more history.
Yes. GPT-4 is trained on decades of food writing, copywriting, and menu psychology research. It uses techniques like sensory language, origin stories, and scarcity framing that have been shown to increase order rates. The key is providing your restaurant's style and brand voice in the prompt so descriptions feel authentic.