Simple Group: using dynamic route planning to add 10% more orders to courier's route

  1. Challenge
    Speed up route planning and optimize the current zone-based routing of deliveries to retail outlets
  2. Solution
    Develop an optimal routing system that factors in drivers' experience in familiar areas and maximizes the strengths of dynamic routing
  3. Results
    10% more delivery stops per vehicle, 9% increase in utilization rate by weight, significantly faster route planning, and accurate arrival time forecasts
About the partner
In this case study, we worked with Simple Group – one of Russia's leading importers of wine, spirits, and nonalcoholic beverages, and a nationwide distributor with its own retail network. The company receives about 1000 orders daily in the low season, and up to 2500–3000 per day in the high season. 

They used to plan all their routing manually, but when the logistics team reached maximum performance, they looked for scalable ways to continue growing.
Step 1. Study the current state of affairs
Like many companies, Simple Group was using territory-based planning. It’s a classic method of route planning, where drivers are assigned specific zones and deliver only within their designated territories.

Zone-based routing has a serious weak spot: it has a hard time adapting to changes like spikes or dips in demand, seasonality, and even new drivers. Another common problem is when loads are distributed unevenly: say, one zone has 10 deliveries per vehicle and another zone has 19. 

Simple Group carried out this manual planning routine throughout the entire day. As new orders came in, the dispatcher added them to existing routes, gradually forming a final path. The bulk of new orders came in the afternoon and evening – that’s when restaurants determine their product needs for the next day.
Simple Group's old map of delivery zones. Each sector was assigned 2-3 vehicles
Step 2. Replace static planning with dynamic routing
An alternative to rigid delivery zoning is dynamic planning, which is carried out without regard to zones. The main advantage is that routes are optimized for vehicle utilization. This results in denser routes, with fewer vehicles required to deliver the same number of orders. 

We took Simple Group's existing routes and came up with a model to show how dynamic planning could affect delivery times and overall efficiency.

Metrics, average day


Routes

RouteQ planning

Changes

Total weight of the cargo, kg.

150 891

150 891

Orders, pcs.
1421
1421
The point of delivery, pcs
986
986
Total travel time, hours
257,26
206,25
-19,8%
The total time of the routes, hours
1077,43
836,82
-22,33%
Total mileage on the routes, km.
8352
6391
-23,4%
Machines involved
78
62
-20,8%
Orders not delivered
0
0
Orders with violations of arrival window
352
3
-99%
Comparing key delivery metrics in static and dynamic planning
It turned out that with dynamic planning, we could increase the number of orders per route from 18 to 22 while using fewer vehicles – 62 instead of 78. The financial benefit of dynamic planning is in the daily savings from not running those 16 extra vehicles. The total travel time also dropped, and there were now zero incidences of late deliveries. 

But after analyzing the new routes in more detail and conducting a few pilot runs, we realized we had to refine the model and incorporate Simple Group's existing delivery practices. 
An example of how routes look with dynamic planning
Step 3. Tailor the solution: mixing things up
Dynamic planning produced extended routes that often crossed each other. At the same time, Simple Group wanted to retain the expertise their drivers had accumulated over the years of working in the same delivery zones.

Our job was to find a way to combine the benefits of dynamic planning without cutting the drivers off from their usual zones. Our solution was to make the zones larger in size. This gave the routing algorithm more flexibility to combine and search for optimal delivery solutions while keeping drivers on familiar routes. 

The main challenge here was making the borders of these extended zones flexible. For example, an order in a borderline area might be better served by a vehicle from a neighboring sector. We made some adjustments so that the algorithm would decide which of the vehicles it would be more cost-efficient to send.

This way, we kept drivers in their familiar zones while giving the algorithm the scope to build optimal routes. At the same time, we solved the problem of cross-border orders by letting the algorithm choose the vehicle that would be most cost-efficient for a delivery. 

The delivery map divided into larger sectors
Implementing RouteQ algorithms: results and figures
It took us about four months to help Simple Group make the transition from static zone-based routing to combined planning with flexible sectors.

Instead of continually planning throughout the day, Simple Group now builds routes only twice a day. The basic structure of routes is laid out in the first round, and then final routes are set at 1 a.m. This saves the logistician's time and frees them up to focus on other tasks.
+10%
number of orders served by courier
+9%
utilization rate by weight
5 times
faster route planning
As a result, in the first months of using the new system, Simple Group saw improved vehicle utilization, a 10% increase in the number of points on each route, and a 9% boost to the utilization rate by weight. A good result – especially given the broader context of the pandemic and economic crisis combined with the low season. The company is waiting for the high season, when the number of orders should double, to test the new route planning scheme at maximum load. The current algorithms are configured in such a way that higher volumes will give them even more opportunities for optimization. 
Any questions?
Feel free to fill the form and we'll get back to you  shortly
Fri Jan 14 2022 16:56:49 GMT+0300 (Moscow Standard Time)