restaurant-menu-benchmarking

Restaurant menu benchmarking

Build a matched menu comparison for specified branches, dishes, and order modes. The output is a benchmark with evidence and unmatched items, not an order.

Setup and API contract

Set CRAWLORA_API_KEY to your key from crawlora.net. Run the bundled scripts/crawlora.sh from this skill directory or by absolute path. It sends x-api-key to https://api.crawlora.net/api/v1; keep the key in the environment. Read reference/endpoints.md for the selected endpoints, required parameters, limits, and response behavior. Check the application code as well as HTTP status; successful payloads are inside data. Stop on 401/403, back off on 429, and retry a transient 5xx once. A failed or partial fetch is not an empty market or catalog. Bound requests to the user's scope and credit budget. For repeated collection, save the query, source IDs, pagination progress, and retrieval timestamps with the results so interrupted work can resume; do not create monitors implicitly.

Collect comparable menus

1. Define the city or coordinates, restaurant branches, meal/daypart, currency, collection window, and order mode. Compare the same branch across channels before attributing a difference to a platform. A chain name alone is not a branch match: verify address and location using store metadata.
2. Discover DoorDash stores with /doordash/search and Uber Eats stores with /ubereats/search. DoorDash search is a pickup catalog; do not label its prices as delivery quotes. Its store and menu calls both need the same latitude and longitude. Uber Eats IDs are storeUuid values from search; they are not DoorDash numeric IDs. Keep platform IDs separate.
3. Use /chipotle/restaurants to resolve a nearby restaurant_number, then /chipotle/restaurant/menu for its priced menu. Preserve its dine-in and delivery price pair, including customization charges. /chipotle/menu is national taxonomy and customization context only: omitted prices are unknown, never zero. include_unavailable=true can distinguish menu presence from current availability on the restaurant endpoint.
4. For national menu breadth, discover McDonald's categories for one country and pass that same country and returned category slug to /mcdonalds/menu. This surface has no prices. Compare its item coverage, not invented prices. National catalogs and branch menus describe different scopes.

scripts/crawlora.sh /chipotle/restaurants \
  latitude=37.7749 longitude=-122.4194 page_size=3
# Take restaurant_number from discovery before fetching /chipotle/restaurant/menu.
scripts/crawlora.sh /mcdonalds/categories country=us
scripts/crawlora.sh /doordash/search \
  query=Chipotle latitude=37.7749 longitude=-122.4194

Match and calculate

• Retain original names and descriptions. Match dish, protein/filling, size, item count, included sides/drink, and required modifiers. Mark exact matches, comparable alternatives, and unmatched items separately; do not force a combo meal and a standalone entree into one price comparison.
• Preserve displayed currency and price units. Inspect actual payload fields before converting minor units; an integer alone does not prove cents. A missing price is null. A free modifier is not evidence that the base dish is free.
• For exact matches, report absolute difference and (comparison - baseline) / baseline * 100, only with a positive priced baseline. Use per-item or per-size prices only when quantities and units are known and comparable.
• Separate regular, promotional, membership, and starting prices. Menu prices exclude any fees, taxes, tips, and minimum-order effects not explicitly returned. Do not call their sum a final checkout total. Preserve availability flags.
• Count unique menu items separately from modifiers and bundles. Deduplicate items repeated in several menu sections without collapsing genuinely different sizes.

Deliverable

Return a table with branch/address, source and collection time, order mode, original item, matched item, portion/bundle contents, currency, price, availability, match confidence, and price difference. Include a coverage summary with fetched branches/sections and missing sources. Summarize price positioning and menu gaps as hypotheses; popularity, margins, sales volume, and customer demand are not established by menu presence. Link original URLs when returned and retain IDs and API request paths when no public item URL is supplied.