Around half of all last-mile related costs are expended in the last few hundred feet because of challenges in finding the exact delivery location, according to Nitin Gupta, Founder at Beans.ai, during a presentation at the Home Delivery World Conference in Philadelphia, Sept. 1, titled “Case Study: Reducing Delivery Failure Rate with AI.” Worse, an uncountable number of deliveries fail altogether because the person (or drone or robot) making the delivery cannot find the exact right location.
Beans.ai is a location intelligence company that creates location data where it previously didn't exist, and builds tools to use this data to improve completed first-time delivery rates, among a host of other benefits.
Beans.ai was not the only tech firm at Home Delivery World highlighting this problem. Others included SimpliRoute, which tackles the issue as part of its routing software and logistics intelligence technology. But Beans.ai has a mission to use artificial intelligence to dig into it at an unprecedented level.
The problem is both overlooked and urgent. With e-commerce retailers finding their margins trimmed tighter and tighter, failed or time-consuming deliveries can mean the difference between profit and loss. With social media-based reviews driving competitive edge, too, they can also mean a customer enthusiastically satisfied, or lost, discouraging others in their wake. Furthermore, delivery drivers are harder to come by and even harder to retain. Sending them on a wild goose chase around an apartment complex or a business facility, looking for an elusive delivery point when the clock is ticking on future deliveries is likely to lead to them under-performing or quitting.
It’s a classic case of the ship sinking for the sake of a half-penny worth of tar. “Whether it’s a human, an autonomous car, a robot, or a drone, if you don’t know where the address is, you can’t make it,” said Gupta.
Part of the problem is that most Americans don’t live or work in an Addams-Family-style mansion with a giant stoop and an obvious front door. According to Beans.ai, 30% of addresses in the U.S. have a secondary address, such as an apartment number that doesn’t immediately reveal which floor it’s on, or an office that might be in any number of buildings in a commercial complex. These secondary addresses are rarely standardized (and a surprising number of online order forms don’t even standardize the first line of addresses), and multiple, critical details remain uncaptured. These include the answers to questions such as: What kind of place is this going to? Is the address on a higher floor? Will the package fit inside the elevator? Is there an elevator? How will the driver access the address? Is there an access code? Is there a doorman? Are there business hours?
Although it’s the customer who feels the immediate impact of a late or missed delivery, it is rarely their fault, said Gupta. He said retailers do not provide the tools for consumers to enter reliable, detailed, accurate addresses. Delivery carriers then do not perform due diligence on the addresses they receive before they commit to the delivery. Then there are poor operational practices at the warehouse, including dispatch systems that don’t ascertain exact locations of intended deliveries on the routes they’re scheduling.
Added to this is the tyranny of map apps, such as Google Maps, which are often just flat wrong. Gupta said more than 30 million residences in the U.S. are off by more than 300 feet on pretty much every platform. Anyone who’s waited in vain for an Uber ride to show up at a multi-entranced airport terminal or hotel will appreciate the frequency with which this leads to a dismal outcome. (Although it remains mysterious why Amazon, UPS and other drivers can’t use a mobile device to call your number when they can’t find you.)
What appeared to be a brilliant solution, launched a few years ago, What3words, which assigns a unique, three-word combination that identifies every single 3 square-meter area on the face of the earth, has failed to be the killer app, Gupta regrets. “What3words is actually a partner — and it works really well for places with no address, such as oceans, forests and trails,” he said. But persuading consumers to use three words instead of their address is a different battle that has gained zero traction in both the retail and logistics space. The other problem, Gupta said, is that what3words is pointing to one particular place, and consumers don't know if they should give the three words for their home location, entrance, driveway, or some other place on the property. Lastly, it doesn't really work in a multi-dwelling unit, since you now need more words to specify a unit number or floor. “This breaks the entire promise of using just three words,” Gupta said.
Back to the drawing board. The only way to fix it is to go after address data accuracy, specifically. This would have been an impossibly huge task before the ready availability of artificial intelligence, which is what Beans.ai is using to re-jig address data so that it’s super-accurate in a way that’s useful for deliveries. “We are solving the fundamental problem in routing, and that is data,” he said. “Your routing is only as good as your data.”
Gupta said the company’s primary offering is simply to make it easier for a delivery driver to find where he or she needs to go, which makes for much faster deliveries. The myriad delivery routing software systems on offer today promise much in terms of efficiency, but they haven’t put a pin in this particular problem, Gupta argues. “We think that 99% of them are the same, if you take out the design. In fact, all routing apps are potential customers of our data, and we actually are working with a couple of them,” he said. “We're also very driver-centric. Most features in our app are for drivers rather than managers, which is different from most routing apps.”
Gupta reckons 80% of delivery failures could actually be headed off before dispatch through better identification of delivery destinations. Other benefits are significant, too. Based on data Beans.ai has gathered from existing customers, using its technology saves two to eight minutes per delivery, means 73% fewer calls to customer support, and garners 14% higher tips for drivers.
Timely, incisive articles delivered directly to your inbox.