By
Hunter Yoakum
Solutions Engineer
Direct mail remains one of the most effective marketing channels – but it hasn’t always been the most efficient. From navigating complex postage pricing to managing undeliverable addresses, marketers face a range of operational hurdles that can drive up costs and dilute impact.
Machine learning (ML) is changing that.
At Lob, we’re reimagining what’s possible in direct mail optimization by applying machine learning to reduce costs, improve deliverability, and unlock new levels of campaign performance. In this post, we’ll walk through two powerful ML use cases we are currently testing: optimizing postage class selection and predicting Return-to-Sender (RTS) mail risk.
Choosing the right postage class – USPS First Class vs. USPS Standard – has always been a trade-off between cost and speed. First Class mail is faster, but more expensive; Standard is cheaper, but less predictable. Historically, marketers had little data to inform this decision at scale. That’s where ML comes in.
Lob’s Postage Class Optimization model uses historical mail delivery data to predict how quickly mailpieces will reach their destinations via different USPS classes. By training a model on millions of past delivery timelines, we’re able to forecast delivery speed by geography, time of year, and mail type – and make intelligent, automated recommendations for the ideal postage class.
The result: optimized spend without sacrificing delivery performance.
Undeliverable mail isn’t just a nuisance, it’s expensive. Between wasted postage, printing, and operational costs, RTS mail can erode ROI and delay time-sensitive campaigns.
Our Return-to-Sender (RTS) Prediction Model solves this with proactive intelligence.
By combining real-time address verification data with historical delivery outcomes, the model flags high-risk addresses before mail is ever sent. It evaluates dozens of signals, like DPV codes, address formatting, past RTS events, and geographic patterns, to generate a predictive risk score for every mailpiece.
Hypothetically, marketers could then suppress or validate these addresses before production, reducing waste and improving deliverability across the board.
Machine learning is transforming direct mail from a static, one-size-fits-all channel into a dynamic, data-driven engine for growth. Whether it’s smarter postage selection or reducing waste from RTS mail, ML empowers marketers to optimize every step of the process.
These models are just the beginning.
At Lob, we’re investing in the next generation of predictive and prescriptive analytics to help our customers unlock more value from every campaign.
