UPS Capital, the financial services and insurance arm of the major carrier, has introduced an AI-powered solution to combat the scourge of porch piracy by creating a score indicating the likelihood of a package being stolen from an address, so it can be diverted to a safer pickup or delivery location.
Called DeliveryDefense, the solution uses AI and machine learning to do a two-year lookback at data from 11 billion UPS deliveries to 130 million addresses. It then creates a confidence score ranging from 100 (most likely to be stolen) to 1,000 (safe for delivery). Packages can be routed to one of thousands of UPS access point locations for pickup, which includes UPS stores and local businesses.
DeliveryDefense has been prototyped and beta tested since January, and the API was released in July. A web version was made available this week to customers of UPS Capital’s InsureShield product. It allows them to bulk upload and search 3,000 addresses at a time. A general market version will be released in late October on a subscription basis, allowing shippers to search 100 addresses at a time.
A 2022 study by Security.org found 49 million Americans had at least one package stolen in the prior 12 months. The total take of stolen goods was $2.4 billion, with a median package value of $50.
Ashley Hillman, director of digital channel marketing for UPS Capital, said increases in computational power, processing speed and storage capacity with Google Cloud is what enabled UPS Capital developers and data scientists to come up with DeliveryDefense.
“Historically you couldn’t look at the address level, querying the database for all the different experiences at this address, every shipping member and all tracking numbers,” Hillman said. “We built our proprietary model on top of (Google Cloud), how to identify loss patterns and frequency with machine learning to come up with a confidence score. That tells us, if you ship to that address, this is the likelihood of a successful outcome.”
Asked why DeliveryDefense was the product of UPS Capital and not the parent company, Hillman said the smaller unit acted more like an agile startup, using its cross-functional teams to develop it a lot quicker.
“The product is more of a fit for us, as we’re responsible for empowering the merchant’s customer experience and post purchase,” she said. “We provide insurance and a financial backstop, plus risk mitigation.”
Hillman said UPS shippers can use DeliveryDefense to create a value estimator, plotting shipping data across the range of confidence scores. They can set rules based on peaks and valleys to determine what action to take at different score levels. She said generally a score of 100-450 indicates low confidence, 450-800 is medium confidence, and 800-1,000 is high confidence.
In testing thus far, Hillman said, UPS Capital found 2% of addresses fall within the range of low confidence, but drive about 10% of all losses.