Honey! Where's my Money?

Company: Honey! Where's my Money? (also referred to as "Where Did It Go")

Self-reported basis: The analysis is based entirely on the author-supplied project description from Devpost, including the tagline, write-up, and technical details. No independent verification or historical data is available.

The company appears to be a hackathon project that builds a local-first financial statement parser and categorizer, designed to work without bank login or account linking. It uses AI for transaction classification and includes features like an interactive dashboard, savings forecasting, and a learning loop that improves over time through user corrections. The app is built with Streamlit and integrates tools such as Groq (Llama 3.3 70B), natural language processing, and Plotly.

What Changed: This project was submitted to the OpenAI 2026 hackathon and represents a prototype solution to a common problem: interpreting messy bank statements in a way that is both private and actionable. It emphasizes local-first design, privacy, and user control over data.

Single Most Important Open Question:

Is there evidence of traction or commercial viability beyond the hackathon demo? The description makes no claims about revenue, customers, or adoption — only self-reported features and technical implementation.

The description states that the product is a local-first financial statement parser and categorizer. It allows users to upload bank CSVs or text-based PDF statements, which are then parsed, classified into categories (e.g., food, transport, bills), and visualized in an interactive dashboard.

Key components include:

• A parsing pipeline that handles different bank formats, including HDFC-specific layouts and a generic position-aware parser.
• An AI categorization engine, powered by Groq (Llama 3.3 70B), which classifies unseen merchants in batch mode to reduce cost and latency.
• A local learning loop: corrections made by the user are remembered permanently, improving future performance.
• An interactive dashboard built with Plotly.
• An AI-generated Markdown financial report summarizing spending patterns.
• A savings goal forecast based on income and expenses.

The app is described as being able to process bank statements without requiring a bank login or account linking — a key design constraint from the outset.

Inference: The product is a prototype tool, not yet deployed for general use. It is built with Streamlit and Python, using open-source or third-party AI models.

The description states that bank statements are technically complete but practically useless, and that existing tools either require linking bank accounts or offer generic visualizations that don’t explain spending behavior.

The company positions its solution as:

• Private (no bank login)
• Local-first (AI decisions cached locally, user corrections override AI permanently)
• Actionable (interactive dashboard, savings forecast, written report)
• Respectful of user data (no sensitive data sent to AI unless necessary)

Inference: The positioning is rooted in privacy and usability concerns, with a focus on empowering individuals to understand their spending without sacrificing control or security.

The description does not state a specific customer segment or ideal customer profile (ICP). It implies the tool is for individuals who struggle to make sense of their bank statements, particularly those who:

• Are frustrated with generic financial tools
• Value privacy and don’t want to link accounts
• Want to understand where their money goes over time

Inference: The target audience is likely early adopters or financially conscious individuals, but no explicit ICP is defined.

There is no evidence of a business model or pricing structure in the description. The project is described as a hackathon submission with no mention of monetization, subscriptions, or paid features.

Inference: No commercial model has been established or described.

The app is built using:

• Streamlit for UI
• Python
• Groq (Llama 3.3 70B) for AI classification
• Natural language processing
• Plotly for dashboards
• A position-aware PDF parser
• A local caching system to avoid repeated AI calls

Key technical features:

• Batched AI calls to reduce cost and latency
• Local memory layer for corrections
• No bank login required
• Support for multiple bank formats (HDFC-specific and generic)
• Manual correction overrides AI permanently

Inference: The app is built with a focus on performance, privacy, and usability. It avoids reliance on external APIs where possible.

The description makes no claims about traction or adoption. It is explicitly described as a hackathon project, not a product in production.

Inference: No evidence of revenue, customers, or usage metrics exists beyond the authors’ own account.

The description does not mention any competitors. However, it implies that existing tools either:

• Require bank linking
• Offer generic visualizations (e.g., pie charts)
• Fail to explain spending behavior clearly

Inference: The project is positioned as a niche alternative to mainstream financial tools, but no competitive landscape is described.

1. No commercial traction or revenue: The product is a hackathon submission with no evidence of adoption.
2. Limited scope: It only supports CSV and text-based PDFs; image-only statements are flagged rather than processed.
3. No scalability claims: The app is built for local use, not enterprise or high-volume deployment.
4. No monetization strategy: No pricing, subscriptions, or business model described.
5. No long-term roadmap beyond hackathon: Features like OCR and encrypted storage are listed as “next steps” but not implemented.

Inference: The project is a proof-of-concept with no indication of commercial viability or scalability.

1. What is the intended path from this prototype to a product that could be used at scale?
2. Are there any plans for monetization or business model development beyond the hackathon?
3. How do you plan to handle OCR for scanned statements, which are flagged in the current version?
4. Has the team considered how to expand beyond the current local-first approach without compromising privacy?
5. What is the expected user base or market size for this tool?