Scoupy Receipt Scanner

Enhancing User Experience for Hassle-Free Cashback

Overview

At Scoupy, cashback only works if scanning receipts feels effortless. We redesigned the receipt scanner to remove friction, increase trust, and dramatically improve scan success. By combining user-centered design with AI-driven recognition, we turned a frustrating moment into a confident, repeatable action.

Challenge

The existing receipt scanner struggled with real-world behavior. Receipts came in all shapes, sizes, lighting conditions, and levels of crumpling. Users were unsure how to position their receipt and received little feedback during scanning.
The result: a scan success rate of just 67%, high frustration, and many abandoned attempts. The scanner felt unreliable, even when the technology underneath was capable.

Research & Insights

User research showed clear patterns:

  • Users did not understand why a scan failed
  • There was no guidance during scanning, only feedback after failure
  • Users wanted control when automation failed, not a dead end

Solution

We redesigned the scanner as part of a broader validation flow instead of a standalone feature.
After scanning a receipt, detected purchases are automatically matched with eligible coupons. When matches are uncertain or multiple options apply, the system transitions into a guided selection step where users can confirm or adjust the result.


Key improvements:

  • A step-by-step scanner interface with clear visual framing and positioning cues
  • Real-time feedback that confirms progress while scanning
  • A fallback validation flow that allows users to confirm or manually adjust detected matches
  • Machine-learning–based receipt recognition trained on a wide variety of receipt formats

Automation where it works. User control where it matters.

From recognition to validation
After scanning the receipt, recognized purchases are linked to available coupons. When confidence is high, matches are applied automatically.
When ambiguity remains, users are guided through a clear selection step to confirm or correct the outcome.

Role of Machine Learning

Machine learning was used to improve recognition accuracy across different receipt layouts, fonts, and print qualities. The system relied on trained pattern recognition models rather than fully autonomous AI.

Instead of hiding uncertainty, the interface exposed recognition confidence and smoothly handed control back to the user when reliability dropped. This human-in-the-loop approach increased trust and reduced frustration, ensuring users stayed in control when automation was not sufficient.

Results

The impact was immediate and measurable:

  • Scan success rate increased from 67% to 93%
  • Active receipt scanners grew by 60
  • Redeemed discounts increased by 30%

More importantly, users stopped blaming themselves when something went wrong. The system felt helpful instead of brittle.