As retail operations grow increasingly complex, delivery planning becomes a crucial part of the entire logistics supply chain.
While many companies have implemented transportation-management systems, daily route planning and scheduling are for the most part done manually. These repetitive processes consume long hours, and frustrate efforts to reduce inefficiencies.
A common retail delivery operation involves store fulfillment from multiple distribution centers, using a mixture of private fleets and third-party logistics providers. Several variables make optimal route planning for retail distribution different from those of wholesale distribution or delivery:
Large fleet-based retailers usually serve hundreds of shops from between three and 10 DCs, using fleets of more than 30 trucks. Each truck does three to 10 stops per trip, depending on each shop's distance and order size.
Even though delivery locations are stable, the demand for these locations can vary substantially from day to day. Differing service times and quantities make it hard to maintain the same route plan, as well as delivery sequencing from DCs to shops. As a result, planners are required to spend long hours placing orders in trucks, and figuring out the number of stops required to meet delivery windows. In addition to the long planning hours and potential for late deliveries, companies often drive unnecessary and excessive miles, as planners usually don’t have tracking capability to generate appropriate key performance indicators.
Which variables should planners consider, when the goal is achieving an optimal route plan for a given day of orders?
Delivery duration. As previously mentioned, DCs serve larger retail distribution areas compared with city dispatch, and can take several days to complete deliveries. Planners need to maintain accurate driving times based on start times and routes. They also need to consider the mandatory hours of service requirements for single drivers and teams. Even if planners can more or less predict driving times between different stops, it's hard to combine these with varying service times, and keep both properly organized via manual operations when alternating between stop sequences.
Delivery windows. Most stores have strict delivery windows which determine when trucks must unload. An average window for a given store varies from one to three hours, and might overlap with others. As trips can take several days, any variation in stops creates problems for planners, when needing to fit new information into delivery windows.
Service times. Average service time at each shop can vary for multiple reasons. Some stores are busier than others, or might have different degrees of efficiency in their receiving operations. Secondly, service times will be different because of variable order sizes. According to general practice, service times for a given store typically vary from 30 to 90 minutes. Tight delivery windows and extended service times also make it harder to alternate the number and sequence of stops.
Thus, planners should balance all these variables to get executable routes and loads. The problem is that there are hundreds or even thousands of potential combinations of orders, with multiple trucks and alternating stops. Done manually, this process takes too much time.
Planners must select the best solution to fit all stops into delivery windows. Significant inefficiencies are likely to occur due to the increasing risk of late deliveries or excessive mileage, because the human brain simply can’t juggle so many variables to produce the most optimal plan.
Vehicle-routing problems may be easy to solve in simple cases, but incorporating all of the above-mentioned constraints makes it hard to solve them in real life. One of the most difficult is the capacitated vehicle routing problem with time windows. The main goal of this algorithm is to get optimal mileage or driving time out of a plan, while meeting all delivery windows and truck-capacity constraints. Our brains spend too much time trying to solve for large data loads, and individuals will likely get non-optimal results by planning for excess mileage or using too many trucks.
There are a number of route-planning software applications that solve the problem for large numbers of deliveries. While many might seem to possess similar functionality, retailers need to be careful in picking the one that considers all requirements, and best fits targets for volume deliveries. Multiple real-life situations aren’t covered in academic papers; hence some application builders fail to account for planning realities when building heuristic algorithms for their software. It would be wise to do tests on several weeks of data, to see if a given solution is the best fit for a company's operation.
Vardan Markosyan is CEO at Less Platform.
Timely, incisive articles delivered directly to your inbox.