World-Class Fraud Prevention, Receipt Processing and Campaign Measurement on Fetch

At Fetch, accuracy and integrity are at the heart of our platform, and we’ve built a tech stack from the ground up that allows us to deliver best-in-class experiences for brands and consumers. 

With 12 million weekly active users, Fetch processes billions of receipts from a variety of retailers and partners. Ensuring that fraudulent activity is flagged and addressed quickly and accurately is not just an operational necessity—it’s a cornerstone of our commitment to both our users and brand partners.

A Tech-Driven, Multi-Layer Approach 

The impact of fraud on the Fetch platform is minimal, accounting for around 2% of total activity. And when fraud does occur, it’s identified swiftly and scrubbed from partner billing – around 80% of fraud within the first 30 days and 95% within 60 days. 

Fetch employs a tech-driven, multi-layered approach to fighting fraud. Our team has developed sophisticated machine learning models that work together to identify and prevent fraudulent activity. Each model is designed to target a specific type of fraud, which ensures that our system is comprehensive, precise and constantly improving. 

As new types of fraud emerge, these machine learning models are continuously updated and refined, ensuring they remain effective against evolving threats. The Fetch team continuously iterates on these systems through a process that reviews and evaluates potential fraud while safeguarding the experience of legitimate users.

Sometimes, well-meaning users may unknowingly break the rules. When a user’s behavior raises red flags, we conduct a comprehensive review of their activity. In some cases, we provide educational messages rather than penalizing the user. This ensures that Fetch remains user-friendly while still holding firm against bad actors. By quietly flagging the fraud, we protect our brand partners by removing the points earned from these receipts without disrupting the user experience.

Beyond Fighting Fraud: Accuracy in Receipt Processing and Campaign Measurement

At its core, the power of the Fetch platform lies in its ability to accurately understand receipt data and consumer behavior – in real time and at scale – through artificial intelligence and machine learning. 

Receipts are notoriously complex documents, with no standard for product abbreviations and no consistent formatting across the diverse retail landscape. On top of that, there’s also a high degree of variability when it comes to issues like crinkled paper or smudged ink. 

That’s why Fetch developed its own receipt scanning technology to ensure unmatched accuracy, speed and scalability in processing millions of receipts daily. This proprietary tech, powered by AI and machine learning, allows Fetch to capture and verify real-time purchase data, enabling precise targeting and personalized rewards.

To train an AI model to reliably and accurately interpret receipts required more than just data – it demanded a monumental “boots on the ground” effort, with the Fetch team visiting stores around the country to purchase items and manually catalog product codes while training the AI models. Over the course of two years, this process provided a deep understanding of every product and receipt code imaginable, and it’s why we can claim that Fetch’s receipt-reading technology is the best in the world.

Since launch, Fetch’s next-generation computer vision and product intelligence engine has improved accuracy in text detection and token labeling by 200%, reduced latency by 50%, and enhanced scalability to support 80 million receipt scans per week. This translates to better intelligence for brand partners and a faster, more relevant experience for users.

World-Class, Real-Time Campaign Measurement for Brands

Fetch’s ability to deliver precise, actionable insights through campaign measurement is why it’s trusted by some of the world’s largest brands. Fetch’s proprietary measurement methodology goes beyond traditional measurement approaches like return on ad spend (ROAS) by verifying campaign impact through real-time receipt data at the SKU level. This allows brands to measure the true incremental effect of their advertising spend with confidence.

Our approach has earned recognition from leading organizations such as Fast Company, KPMG, and Digiday, AdExchanger and more, validating our commitment to delivering value to both brands and consumers  

Through our continued focus on innovation, accuracy and platform integrity, Fetch is committed to staying at the forefront of the adtech and mobile app space, creating unparalleled experiences for consumers and offering brands a trusted, data-driven platform to engage with their audiences.

Working with NielsenIQ to Provide Independent Verification of Product Assignments 

To instill even more confidence in campaign results, Fetch has collaborated with NielsenIQ (NIQ) to offer independent verification of the product assignments that trigger rewards. Fetch’s receipt-reading technology translates details scanned from receipts into standard product definitions that are used for targeting and rewards. The NIQ collaboration enables an independent assessment of these assignments. Using the item-level receipt details that Fetch scans from receipts, NIQ is able to leverage its extensive reference data and industry knowledge to independently evaluate if the receipt details represent the product that Fetch determines. 

Results from the campaigns processed to date show that NIQ is able to verify 91% – 95% of transactions for most campaigns with an overall verification rate of 93%. This means that, for the assessed campaigns, 93% of Fetch product assignments have been verified by NIQ. NIQ does not focus on verification of transactions with receipt details that do not exist in their reference data, which drives the majority of the unverified transactions.  

Being able to provide this unbiased, independent verification solidifies the rigor, accuracy, and integrity of Fetch’s receipt processing models.