The Shipping Architect: What Justin Cramer Learned From 300 Enterprise Implementations

Justin Cramer has spent 25 years on one problem: how does an enterprise shipper move data through its systems fast enough to execute shipping at scale? He started as a nuclear machinist mate in the Navy, then watched bad execution at a legacy shipping software company drive engineers away. One too many whiskies at an Irish pub later, he had started his own company: ProShip, founded in 2001. In this episode, he talks with Timur about what 300-plus implementations taught him about systems thinking, why adding carriers is not a strategy, and where AI actually belongs in shipping software.

From nuclear machinist mate to shipping software

Cramer’s systems thinking started in the nuclear Navy. On a power plant, somebody turns off a breaker and you have to know what the next 24 hours look like: what scrams, what boils off, what breaks. He brought that discipline to shipping software, where the same questions apply: where is the data coming from, how many sources of truth are there, and what happens when one updates and the other does not? The Navy gave him aptitude. The legacy shipping company gave him a lesson in what not to do. “Employees don’t quit companies, they quit bad managers,” he says, and execution was a big problem there. He and a few co-founders tried to split engineering from sales; when that failed, they started their own company with one concern: the quality of the deployment.

For integrators, by integrators

The founding idea was middleware: a front end that could move data between the systems in an enterprise shipping stack. For 25 years, Cramer says, that is all he has thought about. The early product was built by integrators, for integrators. The design scenario: you are stuck at a warehouse at 3:30 on a Friday afternoon on the West Coast, and support has left for the weekend. Can you, as an engineer, fix it with duct tape and bailing wire, get the customer shipping, and fly back in a week to do it right? That was the bar, back when end-of-day manifests went out over dial-up modems.

Speed, compliance, support

As the product matured, three pillars emerged: speed, compliance, and support. Speed came from writing their own platform engines instead of calling carrier APIs. An API is a messaging process between your system and the carrier’s, outside your control. Their engines run inside your infrastructure and evaluate carriers in under 100 milliseconds, versus 800 to 900 for a typical carrier API. On a conveyor that allows only 1,500 milliseconds per package, with the printer and network eating their share too, that difference is everything. Compliance is the price of writing your own engine: you have to keep it updated as carriers change their rules. The United States Post Office, he says, will return an entire trailer if it dislikes the color of the stock the labels were printed on. Shippers from the 2000s have the scars. Support starts with how the software is architected. Too often in an enterprise stack, you catch problems at red. Cramer wants to catch them at yellow: software that health-checks itself and flags trouble before shipping stops. His architecture guidance follows the same control logic. Ship-from-store works fine on carrier APIs: two seconds of latency means nothing to a store associate. Two seconds on a conveyor means boxes launching off the corners. Use APIs where the problem set fits, platform engines where it does not, sometimes both at the same customer.

The last shipping software you will ever need

That combination produced ProShip’s pitch: the last shipping software you will ever need, scaling from a few hundred packages a day to customers shipping more than a million a day across four continents. One decision made that supportable: since 2005, ProShip has shipped on a continuous integration and deployment cycle with essentially one version of the software. “What’s the easiest number of versions to support? One,” Cramer says. No customer runs a 2013 build. Everyone gets the updates, like your phone. When Timur asked whether he ever expected million-package-a-day shippers, Cramer answered with his favorite analogy: “I’d rather be an 800-pound tuna in the ocean than a 20-pound goldfish in a 10-gallon tank.” The goldfish runs the tank. The tuna gets real control, real power, and a much more interesting life.

Carrier diversification is not a strategy, rationalization is

The biggest shift Cramer sees is how shippers think about carriers. When he started, being “multi-carrier” meant UPS, FedEx, and the Post Office. Today, if you only have the big three, you know you are leaving money on the table. Regionals can incrementally change the cost-to-expectation equation in certain regions for certain customers. They will never fully replace the nationals, but they are real tools. His warning: adding carriers is diversification, and diversification without judgment is just a stock picker buying random letters. The discipline is rationalization. Does this carrier back you up in that region? Does it lower cost for a similar service? Does it offer unique features at a decent price? Rationalize how the carrier fits your mix; do not just see how many you can cram on the bus.

Same-day delivery is marching, not viral

Cramer says he has felt like a broken record on same-day delivery, predicting it would take hold “any day now.” What he actually sees is a slow, deliberate rollout, past the proof-of-concept phase of the early 2020s. About 10 percent of ProShip’s customers now include same-day shipping through crowdsourced networks like DoorDash, Uber, and Roadie. Same-day means delivered the same day, not shipped. That requires inventory within roughly 30 to 50 miles and an order management system that knows where it is, plus a pickup time, not just a delivery time. The shipping software, as Cramer puts it, has been sitting there waiting to present the choice at checkout: same-day for this price, next-day for that price. Drones are the longer bet. Cramer is a longtime fan of Zipline, which spent nearly a decade delivering medical supplies across Africa. Drones ignore bridges, rivers, and the worst intersection in town, and he sees them taking the under-five-pound, under-a-cubic-foot shipments that DoorDash and Uber will have to fight to keep.

AI is for probabilities, not the math

Cramer’s framework for AI starts with a distinction: discrete computing versus probability. Discrete computing has one right answer. Rating, routing, labeling, cost: there is exactly one way to calculate the bill, and the carriers, especially USPS, do not accept rounding in the wrong place. Probability is where AI lives: guesses, powered by massive data sets. Hallucination, he says, is just a bad guess. In shipping, almost everything is discrete except one thing: how long the package will actually take to get there. Transit time is highly probabilistic, weather, ice storms, hurricanes, carrier network changes during peak, and that is where AI belongs. A shipping software company does not have to build the model, but it must ingest AI-generated transit times and execute against them in real time. He also expects carriers to open up to AI through MCP, the model context protocol: the robot picking up a hammer, setting it down, and reaching for scissors.

The problem you are told is rarely the real one

If Cramer could redesign enterprise shipping from scratch, he would standardize everything: rating, zoning, end-of-day communication, APIs. Differences in technical connection points, he says, sometimes override carriers’ operational merits. A common standard “would put me out of a job,” he admits, because ProShip exists precisely because the carriers made connecting to more than two of them too painful. From an engineering standpoint it is right; from a business standpoint, he says, it is never going to happen. His closing philosophy is one word: systems thinking. Understand the system as it sits, and understand the problem truly. “Often the problem that is presented to you is not the problem they need to solve,” he says. “It is a symptom of the problem that needs to be solved.” Refine the problem into a better one, then solve it. That applies well beyond parcel shipping.

---

This post is based on the Logistics Remix episode “The Shipping Architect” with Justin Cramer, Co-Founder of ProShip. The views are the guest’s and the host’s own.