Retailer deductions cost suppliers billions annually — most go undisputed because no one owns the process end-to-end.
MarginRecover identifies invalid chargebacks, prioritizes disputes, and automates recovery workflows.
- Identifies invalid deductions across major retailers (Amazon, Walmart, etc.)
- Prioritizes disputes based on recovery value and likelihood of success
- Automates claim documentation → reduces manual workload
- Provides visibility into revenue leakage across operations
📈 Example (sample dataset):
- $24,700 total deductions analyzed
- $15,790 identified as recoverable
- 4 high-priority disputes surfaced automatically
- Rule-based deduction validation
- Evidence-readiness scoring
- Dispute prioritization pipeline
- Automated dispute letter generation
- Revenue impact simulation
Supports optional LLM-powered dispute generation (Anthropic API), with deterministic fallback logic.
Retailer OTIF and compliance deductions often leak revenue because:
- Finance sees the short pay
- Operations sees the shipment
- No single owner drives dispute execution
MarginRecover closes this gap by combining:
- Structured validation logic
- Automated documentation
- Decision support for recovery
- Python
- Streamlit (frontend UI)
- Pandas (data processing)
- Rule-based validation engine
- Modular architecture (data, rules, metrics, AI layers)
pip install -r requirements.txt
streamlit run app.pyIf you want live AI summaries and dispute letters:
export ANTHROPIC_API_KEY="your_key_here"Without an API key, the app still works using fallback logic.
marginrecover/
├── app.py
├── ai.py
├── charts.py
├── components.py
├── config.py
├── data.py
├── metrics.py
├── models.py
├── rules.py
├── requirements.txt
├── sample_data.csv
└── images/
├── dashboard.png
├── disputes.png
└── letter.png
Giorgi Svanidze
Chemical Engineering + Supply Chain @ Case Western Reserve University