Food Costing & Recipe Digitization

Two adjacent problems on one page. Food costing is about menu margins and recipe-cost analysis. Recipe digitization is about getting your grandmother's index cards into a searchable database. For commercial-kitchen / restaurant inventory above this layer see Inventory & Asset Management and Point of Sale; for OCR engines see OCR & Vision; for self-host document management generally see Self-Hosted Document Management; for the recipe importers feeding digitized recipes see Recipe Import & Scrapers; for self-host file / archive backup see Self-Hosted Files & Cloud (cloud / file sync general).

Restaurant / commercial food costing

Hosted / paid

• ★ PlateIQ — paid; AP automation + recipe costing + inventory; bigger operators / chains.
• ★ Galley Solutions — paid; recipe management + nutrition + costing for foodservice; FoodService brand-strong; allergens / nutrition compliance.
• MarketMan — paid; restaurant inventory + invoice OCR + COGS / costing; mid-market.
• xtraCHEF (Toast) — paid; bundled with Toast POS; invoice OCR + COGS.
• MarginEdge — paid; inventory + invoice automation + costing.
• TraceGains — paid; supplier compliance + costing for processors.
• BlueCart — paid; ordering + supplier; tangentially costing.
• Recipal — paid; nutrition labels + recipe costing for CPG.
• MenuQuest / MenuPro — paid; menu engineering.

OSS / FOSS

• ★ ERPNext Restaurant — full ERPNext set up for restaurants; recipe BOM = costing automatically; GPLv3. The serious OSS pick. See ERP Systems and Point of Sale.
• Odoo Restaurant — Community + Enterprise; varies; costing depends on configured BOMs.
• A spreadsheet — what the majority of small restaurants actually do; honest answer.

Recipe-costing math (the actual unlock)

• Theoretical food cost — sum of ingredient costs at recipe yield ÷ menu price. Target: 28–35% in most full-service restaurants.
• Ideal vs actual — the gap is your waste / theft / portion drift.
• Recipe BOMs change with supplier prices. Without automation, you're updating quarterly at best.
• Allergens / nutrition disclosure — increasingly mandatory; Galley / Recipal handle; FOSS doesn't really.

Recipe digitization (paper → searchable)

The OCR + organize stack

• ★ Paperless-ngx + recipe tags — the right OSS pick; OCR every page of an old cookbook; tag with recipe + cuisine + meal-type. Search-grade across thousands of pages. AGPLv3. See Self-Hosted Document Management.
  • https://github.com/paperless-ngx/paperless-ngx
• Tesseract + a custom pipeline — for messier handwritten cards; LLM cleanup on top. See OCR & Vision.
• Mealie + photo upload + OpenAI Vision — newer Mealie versions do "snap a recipe card, get a structured recipe out." Pay-per-recipe but cheap.
• Tandoor + GPT-readable extraction — same shape; lower polish.
• Adobe Acrobat OCR — paid; reliable on printed pages; weak on handwriting.
• ABBYY FineReader — paid; excellent OCR; especially on tables.
• Whisper-style speech-to-text for spoken recipes — see Transcription.

Hosted "recipe-card scanner" apps

• Paprika 3 — has a one-shot photo OCR for printed recipes; works fine. See Hosted Recipe Apps.
• ReciMe — paid; AI-driven; handles handwriting better than most.
• Cookbook (BigOven) — paid; OCR limited.

Handwritten recipes (the hard problem)

• ★ Take photos with good light first. Square framing; even illumination; minimal shadow.
• GPT-4o / Claude Vision — astonishingly good at handwritten cards in 2026. ~$0.005 per card. Beats every pre-2024 OCR.
• Google Lens — free; surprisingly capable for printed; weaker for cursive.
• Apple Notes Live Text — free on iOS; printed only.
• Trainable OCR (Calamari, Tesseract custom training) — open-source; only worthwhile for huge volumes of one-handwriting style.

Family / heirloom workflow (the right pattern)

1. Photograph every card / page in good light.
2. Run through Mealie's photo importer or a Vision API; capture structured output to JSON.
3. Hand-correct the obvious errors; original photo always linked alongside the parsed text.
4. Tag generously: cuisine, person-it-came-from, occasion.
5. Backup the originals + JSON to Self-Hosted Files & Cloud or print-on-demand the cleaned set as a family cookbook (see Family Recipe Books & Cookbook Compilation).

Photographing for archival / costing

• See Photo Mobile Editing for phone-side cleanup; for the underlying RAW workflow see the wider photo section.
• Lighting matters more than camera. Daylight + a sheet of white paper as a bounce board beats a $1k camera in bad light.
• Square / orthographic framing matters for OCR — phone-document-scan apps (Apple Notes, Google Lens, Adobe Scan) do auto-deskew well.

Practical guidance

• ★ For commercial food costing, ERPNext or a spreadsheet are honest. PlateIQ / Galley pay off above ~$2M revenue.
• ★ For family recipes, photo + Vision API + hand-clean is the modern workflow. Beats anything from 2020.
• Don't toss the originals. Heirloom value beats convenience; scan and store both.
• Costing is a habit, not a tool. A weekly invoice review beats the best AP automation in a kitchen that doesn't review.
• Allergen labels are legally regulated; verify your tool's database before you trust auto-generated labels for a CPG product.

Pricing landscape (mid-2026)

Pick this if…

• Restaurant costing, going pro: PlateIQ or MarketMan.
• Restaurant costing, OSS, will commit: ERPNext.
• Restaurant costing, just starting: spreadsheet.
• Family / heirloom recipe digitization: Mealie photo import + GPT-4o Vision + Paperless-ngx archive.
• Printed cookbook OCR at scale: Paperless-ngx + Tesseract.
• Single recipe card, casual: Paprika OCR or just photograph + retype.