The Delivery Theft and Fraud Issue

The scale of package fraud and theft in the US is staggering, with some estimates suggesting that over 1.7 million packages are stolen or lost every day. This is extremely painful for e-commerce merchants trying to provide a positive consumer experience. It is also costly to their margins as fulfillment issues such as delivery fraud and/or theft are expensive to rectify. 

There are limitations to current prevention methods for delivery theft and fraud. As outlined in previous blog posts (Alternatives to Shipping Insurance & Porch Piracy), reacting to issues after they’ve happened and dealing with the claims process is cumbersome, often futile, and costly. 

However, this problem doesn’t have to be a big problem without a solution for much longer thanks to AI and machine learning. 

Enter Machine Learning: A Powerful Solution

To combat this growing threat, companies are turning to advanced technologies, particularly machine learning models that leverage address data. Here's how these models can help:

1. Fraud detection: By analyzing patterns in shipping addresses, frequency of orders, and other data points, ML models can flag potentially fraudulent activities before items are shipped.
2. Risk assessment: Models can assign risk scores to deliveries based on various factors, allowing companies to take extra precautions for high-risk shipments.
3. Address verification: Incomplete addresses can be innocent mistakes or could indicate potential fraud activity. These models can help identify if something isn’t correct.
4. Predictive analytics: By identifying trends and hotspots for theft, these models can help companies allocate resources more effectively.
5. Real-time decision making: ML models can support split-second decisions during the shipping process, potentially rerouting packages to safer locations or requiring additional verification.

Deliverlitics’ Innovative Approach - The Risk Score

The Deliverlitics model uses diverse data sets from several sources across e-commerce and geospatial information to feed the risk scoring machine learning algorithm.   

By aggregating e-commerce merchant’s order and claims data, as well as geospatial and address specific information, the model is able to make a risk prediction for each order. 

Our model allows each e-commerce customer order to be categorized with an aggregated risk score. This score indicates the probability that there will be a delivery issue at a given address. 

After the order is placed, the model instantly scores the risk and provides merchants the ability to see the score and make an informed decision using data on how to proceed with the order. Deliverlitics allows for real-time order adjustments based on business rules put in place by the merchant and their individual business needs.

Conclusion

See Part II of this blog series for a more in depth look into the model and the impacts to the delivery theft and fraud problem. Deliverlitics is working to solve this underserved area of e-commerce to help merchants delight customers and grow sales.