Our evaluation process combines AI-powered analysis with rigorous quality assurance to identify hidden fees, pricing risks, and negotiation opportunities in contracts, invoices, estimates, bills, and service agreements. Here is exactly how it works.
DetectHiddenFees evaluates documents through a structured, multi-stage process designed to maximize the accuracy and usefulness of our analysis. Our process integrates AI pattern recognition with industry-specific knowledge and quality assurance to provide consumers with actionable insights about hidden fees in their documents. Every document undergoes the same rigorous evaluation, ensuring consistent results regardless of industry, document type, or complexity.
The evaluation process begins when a user uploads a document through our secure platform. The system first verifies the document can be processed: checking file format, file size, and basic readability. Supported formats include PDF, image files (JPG, PNG, TIFF), and common document formats (DOCX, TXT). If a document is a scanned image, our OCR system extracts text content before analysis begins. Documents that fail processing requirements — such as corrupted files or heavily damaged scans — are flagged for user attention with guidance on how to resubmit.
The AI screening stage is the core of our evaluation process. The system analyzes the document at multiple levels: (1) Structural: How is the document organized? Where are pricing, terms, and fees located? (2) Linguistic: What language is used to describe fees and charges? Are there vague or suspicious terms? (3) Numerical: What are the pricing patterns? Are percentages, flat fees, or compound charges present? (4) Comparative: How do charges compare to industry benchmarks and known fee patterns?
After initial screening, the AI applies pattern recognition models trained on over 100,000 labeled documents. These models identify fee patterns associated with common hidden fee tactics: administrative fee padding, documentation fee inflation, duplicate line items, percentage-based surcharges, escalation clause structures, and vague terminology that often masks undisclosed charges. Each identified pattern is assigned a risk score based on its similarity to known fee manipulation tactics.
Each identified fee or charge is classified into one of our standardized categories: Transaction Fees, Contract Fees, Service Fees, Financial Fees, Subscription Fees, or Consumer Fees. This classification ensures consistent reporting across all documents and allows users to understand exactly what type of fee they are dealing with. The classification is based on the fee's structure, context, and typical application rather than just the name used in the document.
Our evaluation process incorporates industry-specific knowledge to provide contextually accurate analysis. The system maintains fee benchmarks for major industries including automotive, healthcare, banking, home improvement, telecommunications, insurance, real estate, and subscription services. When evaluating a document, the AI considers industry norms, typical fee ranges, and known pricing practices specific to that industry.
Quality assurance is integrated throughout the evaluation process. The system performs automated consistency checks on all analyses, verifying that risk scores are internally consistent and that flagged items are supported by document evidence. A random sample of analyses is reviewed by trained quality assurance analysts weekly. User feedback is collected and analyzed to identify patterns that may indicate analysis errors or missed fee types.
While our AI handles the primary analysis, human review plays an important role in quality assurance and continuous improvement. A team of trained analysts reviews a subset of analyses to validate accuracy and identify areas for model improvement. Analysts are trained in consumer finance, contract analysis, and billing practices across multiple industries. Their feedback directly informs model updates and classification improvements.
The final output of our evaluation process is a comprehensive analysis report. Each report includes: (1) Overall risk rating for the document, (2) Category-specific risk scores, (3) Flagged fee items with explanations, (4) Questions to ask the provider, (5) Negotiation strategies based on identified issues, (6) Links to relevant educational resources. Reports are designed to be actionable — each item includes specific guidance on what to do next.
No evaluation process is perfect. Our system has limitations that users should understand: (1) The AI cannot detect fees that are disclosed verbally rather than in writing, (2) Accuracy varies by document quality and complexity, (3) Novel fee structures may not be recognized immediately, (4) Industry benchmarks may not apply to all geographic regions or market conditions, (5) Analysis is not legal advice. Users should treat our evaluation as a starting point for investigation, not a definitive determination.
Q: How long does the evaluation take?
A: Most documents are evaluated in 30-60 seconds. Complex documents with hundreds of pages may take up to 2 minutes.
Q: Is every document reviewed by a human?
A: A random sample of analyses is reviewed by human analysts for quality assurance. Most analyses are fully automated.
Q: Can I request a human review of my analysis?
A: Yes. Contact us through our support channels and we will arrange a human review of your analysis results.
Q: How do you ensure evaluation consistency?
A: Our AI applies the same analytical criteria to every document. Quality assurance checks verify consistency across analyses.
Q: Are industry benchmarks updated regularly?
A: Yes. Fee benchmarks are updated quarterly based on new data from regulatory filings, industry reports, and document analysis.