C.H. Robinson examines impact of predictive ETAs and related benefits for retail shippers

Minneapolis, Minn.-based global logistics services provider and freight forwarder C.H. Robinson is keeping a close eye on the current inventory situation, for the 2023 holiday season, in terms of how it has changed the flow of retail freight and subsequently led to retailers relying more on predictive ETAs (estimated time of arrival).

A C.H. Robinson research paper made the case that having real-time freight visibility is key for retailers during the holiday season, with the company’s data science team, digital carrier network, and data on three million shipping lanes sharply focused on predictive ETAs. This was driven, the company said, by the impact of the post-pandemic inventory glut retailers dealt with, which, in turn, made them cautious about over-ordering product for the 2023 holiday season. What’s more, it made the case that retailers are instead making orders for smaller amounts, for product they need as needed.

C.H. Robinson has hit 98.2% accuracy in predicting when a truckload shipment will arrive within an appointment window, as well as 92% accuracy in predicting that a less-than-truckload shipment will be on time.

LM Group News Editor interviewed Noah Hoffman, C.H. Robinson Head of Retail Logistics, about predictive ETAs and the benefits they provide for retailers during the busiest time of the year.

Logistics Management (LM): How does C.H. Robinson define predictive ETAs and how do you utilize them to help shippers with freight visibility and inventory management efforts?

Noah Hoffman: Using data science and artificial intelligence, we predict whether a shipment will arrive on time. Shippers like having real-time visibility into their freight at all times, but knowing whether the truck will arrive during the scheduled appointment window is what shippers and receivers plan for. For retailers, the predictive ETA affects staffing and throughput at their warehouses, their distribution and fulfilment centers, and their stores. Especially this holiday season, when many retailers are trying to maximize holiday sales with less inventory and less staff, they need to know with more precision where their inventory is.

LM: What are the key steps shippers need to take in order to know with more precision where their inventory is?

Hoffman: It starts with choosing the tools that are the right fit for your business. You can spend a lot of money and devote a lot of staff time to real-time visibility without necessarily improving on-time delivery. That’s because real-time visibility tools are only as good as the data they have access to. A retailer can put themselves in the best position if they’re tracking inventory at rest together with inventory in motion, in case they need to redirect merchandise. Then once inventory is in motion, a shipper shouldn’t have to take any steps if they’re getting accurate ETAs, getting meaningful notifications, and working with a logistics provider that can intervene when needed.

LM: What are the main steps, or measures, C.H. Robinson takes to make sure customers receive an accurate prediction of whether their merchandise will be on time or not?

Hoffman: We’ve now hit 98.2% accuracy in predicting that a truckload shipment will arrive on time and 92% for less-than-truckload shipments. To get there, we took a different approach than other companies. We determined which tracking data produces the most accurate ETAs. We stopped collecting data that’s unreliable but typically required of carriers across the industry, and instead we asked our carrier network to provide status updates at more relevant points in time.

LM: What tracking data produces more accurate ETAs?

Hoffman: It’s a little counterintuitive, but just getting more status updates or more frequent status updates on a shipment doesn’t necessarily equate to knowing whether it will arrive on time. A shipment could be late despite 150 status updates in a row that seemed to indicate it’ll be on time. Besides, most shippers don’t want to wade through all that noise. What we homed in on is the most critical tracking information and getting it at the critical points: on the way to pick-up, when the driver leaves the pickup appointment and the final two hours before delivery.

LM: What are some examples of unreliable data?

Hoffman: Like most logistics companies still do, we used to ask carriers to tell us when their truck was empty after delivering their last load. Seems logical to know when and where a carrier finished their last load, right? We didn’t always get that information and even when we got it, it actually lessened the accuracy of the predictive ETAs. So, we stopped collecting it.  

LM: How did carriers marking themselves empty previously lessen the accuracy of predictive ETAs?

Hoffman: Just because the truck is empty doesn’t mean it’s coming. It could get a flat. It could hit traffic. The driver might need to take a break. No widely used technology can detect those things today, so carriers have to manually report being empty. Any manually reported update is harder to collect as consistently as an automated update. Also, the truck could be marked empty a day or more before their next pickup. Too much can happen in the meantime for that to be a good predictor. 

LM: What are some examples of the new rules you instituted for carriers to provide tracking data at different, more meaningful points in time?

Hoffman: In April, we began requiring carriers to provide automated updates starting at four hours and again at two hours before the pickup appointment. The appointment window is heavily weighted in our data-science model. If a truck is four hours late for pickup, the model knows the truck can’t go 900 mph for the rest of the trip to make it up. That influences the predictive ETA along with other data like how long it takes for the truck to be loaded, when they leave the pickup and other in-transit data. The new real-time data gets fed into a model that has already learned from our historical data on 20 million shipments a year across 3 million shipping lanes.  

LM: How quickly were improvements seen in ETAs when C.H. Robinson stopped asking carriers to report when their truck was empty and started the new rules?

Hoffman: Our data scientists tell me it's been gradual progress. The machine learning is continually taking advantage of each new data point flowing in.

LM: Are you seeing an uptick in shippers looking for this type of help in the 2023 Peak Season, at a time when inventories for the most part remain low, especially for this time of year? 

Hoffman: At the beginning of the year, 59% of the customers we surveyed said that improving real-time visibility into their freight would be critical to their success. That’s clearly been the case for Peak Season. On the whole, the 7,500 retailers we serve have been careful not to over-order. Many started the holidays pulling from existing inventory and have been replenishing their hot sellers only when they need to. They’re ordering small amounts of new inventory for specific locations or shifting merchandise between locations. When your inventory management is that dynamic, you have to rely much more heavily on predictive ETAs to make sure the right product gets to the right place at the right time.