AI-powered reasoning engine for fraud detection in HP’s affordability offers

1. Introduction & problem statement

Affordability offers are a key lever for HP to expand the consumer market and win new customers. HP spends a significantly large sum to support “Affordability offers” for accelerating their consumer business. Retailers are claiming benefits linked to “no-cost EMI” and other affordability schemes even when not authorized or mapped to specific HP products. This leads to an increased incentive payout and impacts on the profitability of HP’s consumer business as well as creating a non-level playing field amongst partners.

HP wants to control any such financial leakage due to fraudulent or misrepresented financing practices at the retailer level.

With a high volume of affordability-linked transactions across thousands of retailers and multiple financial partners, manual verification of each scheme-product-retailer mapping is not scalable. This presents an opportunity to leverage AI and Machine Learning to flag anomalies, prevent fraud, and optimize financial efficiency.

2. Solution overview:

CORESight – AI-powered platform for anomaly detection in HP’s “Affordability offers” data.

Key capabilities include:

• AI-driven fraud detection: Leverages ML models to detect anomalies in retailer behavior and finance patterns.
• Product-scheme-retailer mapping: Automatically validates transactions against authorized mappings.
• Transaction behavior analysis: Flags sudden spikes, off-hour activity, or unusual frequency.
• Dashboard & alerting: Real-time alerts for HP finance and operations team.
• Continuous learning: Human-in-loop system for model improvement using confirmed fraud examples.

This solution enables HP to gain visibility into fraudulent activity and take corrective action proactively.

3. Core modules and technical features

• Anomaly detection engine: Unsupervised learning – Isolation forest, Autoencoders, Supervised learning – time-series based change-point detection models, tree-based models.
• Retailer risk profiling: Dynamic scoring of retailers based on historical claim behavior.
• Pattern analysis: Flags mismatches in EMI schemes, cashback misapplications, claim-to-sales ratio spikes.
• Excel-based report for review
• Feedback loop: Integrated feedback module for inserting flagged fraudulent transactions back into the algorithm for improved accuracy.

4. Requirements from HP

a. Data (last 6–12 months preferred)

• Sales transactions: Product ID, Date, Channel, Retailer ID, Pricing.
• Affordability claims: EMI type, tenure, finance partner, interest value, claim amount, payout status.
• Retailer master data: Authorized schemes by retailer, eligible product mappings.
• Claim logs: Interest payouts, approvals, rejections.
• Known fraud cases (optional): Blacklisted IDs, flagged cases.

b. Support

• SPOC from finance/operations.
• Scheme rulebook and eligibility policy explanation.
• Walkthrough of affordability claim workflow.

5. Deliverable to HP

• CORESight : AI solution for anomoly detection in “Affordability offers” data.
• Secure access via browser.
• Portal to upload transaction data.
• Automated weekly report and monthly review.

6. Commercial terms

• One-time set-up fees (includes model set-up, vector database creation, report format set-up): INR 12,00,000/-.
• Monthly fees: INR 4,30,000/- (includes weekly model enrichment).
• GST extra.

7. Implementation and system configuration period:

• One time set-up: 3-4 weeks from PO.
• Accuracy will improve as system gets enriched with data and feedback.