Security Verification

For security purposes, a password is required to analyze and compare product images.

Please enter the following password:

SOB-0c866f112a1ab6e2


Same One Back

AI-Powered Return Verification for PayPal Sellers

Verify the Item. Protect the Seller. Keep Refunds Fair.

Inspiration

Imagine carefully packing a gold necklace, photographing every detail, and shipping it to a customer.

A few days later, the buyer opens a PayPal dispute claiming the item was not as described. You accept the return and wait for the package to arrive.

But when you open it, something is wrong.

The necklace is there, but the chain has been replaced.

You know you shipped the original item. However, how can you prove that the item returned is not the same one?

This is a difficult problem for online sellers. Existing dispute processes can verify transactions, shipping records, and product conditions, but identifying whether a returned physical item is the exact same item originally shipped remains challenging.

We built Same One Back to solve this missing piece of the return process.

The Problem

In 2024, U.S. merchandise returns reached approximately $685 billion. According to Appriss Retail, approximately 15% of returns involved fraud or abuse, representing an estimated $103 billion.

Return fraud can take many forms:

  • Item substitution: Returning a different item that looks almost identical to the original.
  • Component replacement: Replacing valuable components, such as a jewelry chain or accessory.
  • Missing components: Removing parts before returning a product.
  • New damage: Returning the original item in a worse condition than when it was shipped.

For merchants, these cases can result in losing both the product and the refund amount.

Traditional dispute processes often rely on shipping records, written explanations, and customer-submitted photographs.

But these records cannot reliably answer one critical question:

Is the physical item being returned actually the same item that was originally shipped?

What We Built

Same One Back is an AI-powered return verification solution designed for online merchants accepting PayPal payments.

It creates a visual identity record of each product before shipment and uses AI to compare the original item against the returned item.

Our system identifies four possible outcomes:

  1. Original Item: The returned product matches the original shipment record.
  2. Item Substitution: The returned product appears to be a different physical item.
  3. Missing Components: One or more original components are missing.
  4. New Damage: The original item has acquired additional visible damage.

Instead of simply generating a similarity score, Same One Back highlights exactly where the differences occur and provides enlarged side-by-side comparison images.

The seller reviews and confirms the findings before submitting the evidence through the supported PayPal dispute workflow.

How It Works

1. Register the Original Item

Before shipping, the seller selects the corresponding PayPal order and photographs the product using our app.

The seller captures important identifying details, including:

  • Serial numbers and engravings.
  • Unique scratches and surface patterns.
  • Product-specific components.
  • Distinctive physical characteristics.

These images are stored and linked to the original transaction, creating a visual shipment record.

2. Detect PayPal Disputes

When a buyer opens an "Item Not as Described" dispute through PayPal, Same One Back retrieves the available dispute information.

The seller receives a notification in the app, including the dispute status and response deadline.

The seller can review the case and accept the return through the supported workflow.

Buyers continue using PayPal's existing return process without needing to click unfamiliar external links.

3. Capture the Returned Item

When the returned package arrives, the seller photographs the product again using the app.

The new images are automatically associated with the original order and shipment record.

4. AI-Powered Physical Item Verification

This is where our AI makes the difference.

A simple image comparison is not enough.

Two products may look almost identical, while the differences that matter are only a few pixels wide.

Comparing two full-resolution images at once can also cause important details to disappear when the images are resized.

Same One Back uses region-based visual analysis:

  1. Region Detection: Identifies important product areas, such as serial numbers, engravings, clasps, labels, and unique surface details.
  2. High-Resolution Cropping: Extracts corresponding regions from the original and returned photographs.
  3. Visual Feature Comparison: Uses computer vision and multimodal AI to examine similarities and inconsistencies.
  4. Difference Localization: Highlights the exact areas where potential discrepancies occur.
  5. Classification: Categorizes the returned item as original, substituted, missing components, or newly damaged.

The result is an explainable visual report with enlarged comparison images, allowing sellers to understand and review the AI's findings.

5. Seller Review and PayPal Evidence Submission

Before submitting anything, the seller reviews and confirms the AI-generated findings.

Same One Back then prepares a structured return verification report with:

  • Original shipment images.
  • Returned item images.
  • Identified differences.
  • Enlarged side-by-side comparisons.
  • AI verification results.

The seller can use this information to support the return confirmation or report that a different or altered item was received, depending on the supported PayPal dispute workflow.

6. Dispute Resolution

PayPal remains responsible for the final dispute decision.

In our sandbox demonstration, the simulated dispute decision determines the final outcome.

The goal is simple: provide better physical evidence so both buyers and sellers can receive fairer dispute outcomes.

How We Use AI

Our core innovation is not simply recognizing whether two photographs contain the same type of product.

It is determining whether they represent the same physical object.

For example, two gold necklaces may have the same design, color, and overall appearance, but different chains or clasps.

A conventional full-image comparison may overlook these small but critical differences.

Our region-based approach focuses on the product's identifying details rather than relying on one overall similarity score.

By examining individual regions at higher resolution, the system can identify potential substitutions, missing components, and new damage.

We also include human-in-the-loop verification: AI findings must be reviewed and confirmed by the seller before being submitted as dispute evidence.

This makes the system an evidence-generation tool rather than an automated judge.

Why PayPal?

Same One Back is designed around the PayPal dispute and return workflow.

The key opportunity is the returned-item confirmation stage.

Sellers need a way to report what they actually received, especially when the returned product does not match the original shipment.

Same One Back adds a physical item verification layer to this process.

Rather than replacing PayPal's existing dispute resolution system, we aim to strengthen it with structured visual evidence.

This is also why our initial focus is on PayPal merchants rather than building a generic image comparison application.

Who Is It For?

Our primary target customers are small and medium-sized online merchants accepting PayPal payments.

We initially focus on product categories where physical details are important and individual returns can create significant financial losses.

Examples include:

  • Jewelry and Accessories: Unique engravings, clasps, chains, and identifying marks.
  • Trading Cards and Collectibles: Graded cards, rare collectibles, and distinctive surface details.
  • Electronics: Serial numbers, identifying marks, and removable components.
  • High-Value Specialty Goods: Products where a single fraudulent return can cause substantial losses.

Our initial MVP targets WooCommerce merchants using PayPal, with plans to expand to Shopify, BigCommerce, and other e-commerce platforms.

What Makes Us Different?

Most existing return management solutions focus on transaction records, shipping information, customer communication, or damage documentation.

Same One Back focuses on a different question:

Is the physical item coming back the same item that was originally shipped?

Our differentiation combines three capabilities:

  1. Pre-shipment visual identity registration.
  2. AI-powered region-level physical item comparison.
  3. Integration with the payment dispute workflow.

Together, these capabilities turn ordinary product photographs into structured evidence that can support return dispute resolution.

Technology

Our prototype combines:

  • AI-powered visual analysis.
  • Region-based image comparison.
  • High-resolution image cropping.
  • Product detail detection.
  • Automated discrepancy classification.
  • Visual evidence report generation.
  • PayPal transaction and dispute workflow integration.

The application connects original shipment records with returned item images and presents the verification results through a merchant-facing dashboard.

Challenges We Faced

One of our biggest technical challenges was identifying differences between two nearly identical products.

A standard image similarity score is not sufficient for this task. A necklace with a replaced chain may still have extremely high overall visual similarity.

We therefore designed a region-based comparison workflow that focuses on identifying and examining smaller product details individually.

Another challenge is ensuring that AI findings are reliable enough to support a dispute.

Differences in lighting, camera angles, and image quality can affect visual comparison. This is why our solution includes seller review and confirmation rather than automatically making final decisions.

We also need to validate which returned-item confirmation and evidence-submission actions are available through PayPal's APIs and permissions. Unsupported actions are represented through clearly labeled sandbox simulations in our demonstration.

Future Plans

Our roadmap includes:

  • Improving visual verification accuracy across different product categories.
  • Expanding product-specific image comparison models.
  • Launching WooCommerce integration.
  • Supporting Shopify and BigCommerce.
  • Adding merchant analytics for return patterns and verification results.
  • Expanding to additional payment and e-commerce ecosystems.

Our long-term vision is to make physical item verification a standard part of the online return process.

Impact

Same One Back is designed to protect honest sellers without compromising the rights of honest buyers.

For sellers, it provides a reliable visual record of what was originally shipped and helps reduce the time and uncertainty involved in return disputes.

For buyers, it supports more transparent and evidence-based decisions instead of relying solely on conflicting descriptions.

We believe returns should be fair to both sides.

Same One Back — The missing link in return protection.

Verify the Item. Protect the Seller. Keep Refunds Fair.

Built With

  • ai
  • automation
  • claude
  • computervision
  • imageanalysis
  • paypal
  • python
  • riskassessment
Share this project:

Updates

Submission history