crumbs

Local insights for better business decisions

• GitHub repo
• GitHub showcase
• Video on Google Drive
• Slides
• Demo video

Crumbs helps small food businesses turn local customer feedback into practical business improvements. We’re building for independent restaurants, cafés, and bakeries whose owners are busy managing daily operations and don’t have time to research nearby competitors. Useful information is scattered across reviews, menus, and business listings. An owner might know their overall rating but still struggle to answer: What do nearby businesses do better? What keeps customers coming back? Is there a product people want that we could offer? Crumbs brings this information together in an easy-to-understand view, helping owners decide what to investigate and improve. We built a web-app prototype with two connected experiences: - Business owners get an Insights dashboard, comparisons with similar nearby businesses, and alerts. The goal is to turn local evidence into specific actions: investigate recurring wait-time complaints, protect a strength customers praise, or test a product receiving positive attention nearby. - Customers get a neighborhood map, cuisine filters, review summaries, and side-by-side comparisons based on attributes such as food quality, value, and wait time. They can find places that fit their preferences, save them, and open directions. The customer experience supports the business mission. It helps people discover local businesses and could give owners additional insight into which alternatives attract interest. In-app selections indicate interest; they do not prove visits, purchases, or why someone chose a competitor. Our research code uses Tavily to collect accessible local information, with specialized agents working in parallel across sources. Claude is intended to interpret the evidence and produce clear summaries and recommendations. We have tested connected Tavily collection and separate product-opportunity analysis; the complete app-to-Claude workflow still needs end-to-end verification. To get started, an owner selects “I run a spot” and enters their business name, address, and cuisine. The team hosts the technical services so the intended setup requires no coding from the owner. An optional Square connection is being prototyped for authorized inventory information. The prototype currently includes sample scores, alerts, and customer-selection activity. Live integrations and deployment need more work, and accessible reviews provide incomplete coverage. Recommendations should be treated as ideas to test, supported by source evidence and the owner’s knowledge. Our next step is a small business pilot to measure whether Crumbs reduces research effort and helps owners make useful changes to their menus, service, and operations.