[REPORT]
Data in route optimization: state of play, issues and best practices
In route optimization, data is at the very heart of algorithms. We mentioned this in our article on route optimization data: the more accurate and reliable the data, the more relevant the optimized routes will be. Indeed, poor data quality has a strong impact on the accuracy of the calculated routes, which quickly become irrelevant.
Our experts have observed that the relevance of routes is very often strongly impacted by changing, incorrect or unknown data. Unfortunately, it is still common to see these problems in the logistics and transport industry, mainly because:
- Logistics operators still often enter data by hand, which increases the risk of error.
- They don’t always have control over the quality of the data collected by their customers.
- Their work is becoming increasingly complex and to handle unexpected events, it is becoming ever more essential to take into account new real-time data (new orders, traffic jams, accidents, etc.).
Kardinal has decided to tackle these issues by developing an innovative and fundamentally different approach to route optimization: the “Always-On” approach.
The main challenges of route optimization data
1 - Changing data
On a daily basis, professionals have to deal with many uncertainties: new orders that need to be planned in a hurry, a driver that warns minutes before starting his shift that he will be late, changing traffic conditions, etc. The data provided to the optimization algorithm is only valid at a given moment and is likely to change. As field reality can change significantly, it is essential to be proactive in the event of a change.
The majority of route optimization solutions on the market have a “static” approach, whereby once created, the route plan is not rearranged by the algorithms. This approach is not suitable given the changing nature of the field: if the data has changed, the result of the optimization is no longer valid. To include the changes in the routes, planners must either rerun the optimization calculations if they have time or adjust in real time, manually.
2 - Incorrect data
When analyzing the databases of logistics companies, we often find the same mistakes: an incomplete address, mixed fields, wrong schedules, incorrect weights and volumes… there are different examples. However, poor data quality can have a significant impact on routes to be carried out:
- A wrong address will complicate geocoding and hinder distance calculations. It will also make it difficult for the delivery driver to find the delivery location, taking more time.
- Mixed fields will lead to errors and inconsistencies in the optimization depending on the fields concerned.
- Incorrectly filled-in schedules can lead to a driver going to the delivery location when it is closed. Not only will this result in delaying the route, but it will also require a second trip.
- Incorrect weights and volumes may require vehicles to be rebalanced before leaving for the delivery route because their loading capacity has been reached, as the packages are much larger or heavier than initially planned.
All these examples show that data quality is critical to properly optimize last mile deliveries. Logistics professionals are making increasing efforts to improve data collection, but perfection is difficult to achieve.
In the context of a static optimization, the processing of incorrect data can be long and time-consuming. Indeed, once the optimization calculations have been made, if an incorrect data must be corrected because it impacts the calculations, the planner has no choice but to rerun the entire optimization of the route or to fix it manually. In practice, this often leads organizations with imperfect data to not use the optimization tool to its full potential, some even using it only as an interface in which they plan almost everything manually.
3 - Unknown data
Our experts often find a lot of missing data in the databases of logistics companies. This information is generally in the heads of business experts who know the field: this important customer prefers to be delivered by a specific driver whom he prefers — on the other hand, another driver has had a conflict with a customer who no longer wishes to see him. Some professionals may or may not have the skills to carry out a given task, some goods may be loaded together but others not, etc.
These constraints are very rarely specified in databases: they are more about know-how, business knowledge and human sensitivity. Yet, they are very important to maximize the quality of the service provided and customer satisfaction. By ignoring these constraints, optimization algorithms cannot calculate the most relevant routes for operational teams.
With static optimization, we are often inclined to run huge data projects aimed at collecting the information in the heads of the operational teams to integrate it into the optimizer. Although admirable, these projects often require a lot of time and effort from teams. The more databases there are, the greater the effort to keep them up to date. This can lead to team burnout and eventually to a stop in using the database altogether.
Kardinal's solution to these data challenges: continuous optimization
At Kardinal, our mathematicians thought of a solution that would meet these three data challenges and created the ARO (Always-On Route Optimization) solution. As suggested by its name, this route optimization solution works continuously: it never stops optimizing! Thus, the user can change, add or delete any data at any time without having to restart the optimization. A real innovation for planners who used static optimization softwares until now!
Continuous route optimization solution
For all three of the above problems, changes are made as new data arises: new orders to be scheduled, a driver late for the warehouse in the morning, wrong addresses, etc. These additions or data modifications by the planners can be instantly added to Kardinal’s tool, adapting the optimization with ease. Changes in road traffic are natively taken into account in the solution and update the routes as they take place. Thanks to this real-time optimization, planners can rely on the best possible routes at any given time and make decisions when they need to.
In the event of missing data, it is not always required to create huge repositories to collect information known only by some operators. Our ARO solution gives them the possibility to add this data on the go for the algorithms to take into account in their calculations. By keeping control of the tool, they can monitor and adjust the proposed optimizations.
Thanks to this approach, a sort of conversation is created between, on the one hand, a user who has a better field experience, and on the other hand, a machine whose computing power is far more powerful and faster. Such a cooperation allows for maximum Human-Machine synergy, which our ARO solution will take advantage of to go beyond the limits of conventional route optimizers.