From manual route planning to continuous optimization: the different types of route optimization

[REPORT]
The different types of route optimization

Route optimization is a strategic concern for most organizations in the goods delivery industry, as well as for companies providing home services or technical interventions. It means finding the least expensive solution to visit all the locations while following certain constraints (the most well-known being time, vehicle capacities, or technician skills, but there are thousands more, some of them very specific to an industry, an activity or even a company).

In practice, route optimization is most often carried out by a planner (sometimes called a coordinator, dispatcher or scheduler), either manually or using a dedicated software. In order to help planners find the best routes among an (almost) infinite number of routes, sometimes taking into account several hundreds of variables and constraints, the solutions dedicated to route optimization nowadays integrate powerful algorithms capable of creating, almost flawlessly and in a few minutes, what humans struggle to achieve in several hours, days or weeks (depending on the complexity of the problem and the desired level of optimization).

But not all solutions are equal. Depending on the quality of your data, the organizational processes you already have in place, or the degree of flexibility of your business, a static approach may not be the most relevant. From manual planning to continuous optimization, this article explains the different types of route optimization, their advantages and drawbacks.

The 3 types of “conventional” route optimization

Companies that need to optimize their routes have an internal team dedicated to allocating activities (goods, parcels, services, waste collection, etc.) to resources (drivers, technicians, garbage collectors, etc.). In this section we will review the different types of optimization processes implemented by organizations.

Manual optimization by experts

Manual route optimization by a planner is the most widely used approach across all industries. An expert, who has often worked in the field, is in charge of assigning tasks (deliveries, interventions, etc.) to drivers. Thanks to their experience, they create routes based on information that have proved to be effective over time. Organizations have often adapted themselves to make the planners’ work easier.

For example, in last-mile delivery companies, there is often a pre-existing allocation of areas (often defined by zip codes) to loading bays (and thus to drivers). This simplifies parcel sorting and the planner’s job. Of course, this process is clearly not the most efficient, as assigning zip codes to drivers is not necessarily the most efficient system. Moreover, it does not guarantee that all constraints, such as delivery slots, will be met. In practice, scheduling within rounds is regularly done by drivers themselves during the round.

In the case of home visits, schedules are often defined only once a week or even once a month. Coordinators have to manage a large number of constraints, most of which are difficult to define formally, such as preferentially assigning the same professional to a client or, on the contrary, avoiding professionals and clients that don’t get along together. In practice, this schedule, which is supposed to be defined beforehand, must in fact be adapted every day, mainly because of the availability or not of some professionals, as the profession has a high level of turnover and absence. Thus, planners’ spend much of their time making last-minute adjustments to the schedules rather than optimizing them to achieve maximum performance.

In spite of a lack of optimization in their approach, planners have a great capacity to manage the unexpected by adapting their decisions to the context. They are able to take into consideration hard to define parameters such as “with the delays we recently had, which clients can I afford to disappoint if I can’t do otherwise?” or “this client is dissatisfied with yesterday’s cleaning service, should I send a nearby housekeeper to finish the job and risk making him late for other interventions?

Humans can manage business priorities in real time, make subjective and context-dependent decisions when an event occurs, and use information that is not in any database: these are their greatest strengths.

Benefits:

  • High ability to manage business priorities
  • Ability to make subjective choices based on the context
  • Ability to take into account information that is not in the database
  • Process that allows on-the-go planning, which is more realistic in many contexts where orders arrive progressively

Drawbacks:

  • Limitation in computing power, weakening the efficiency of the rounds (costs and service quality).

Static optimization*

Static optimization involves the most commonly found algorithms and softwares on the market. They provide schedulers with powerful algorithms to automate the optimized dispatch of tasks to individuals, as well as their scheduling, before the rounds start. This is called “static” optimization because it must be launched before a deadline, when the user considers that all the data the algorithm needs to calculate the routes is known and that this data is perfectly valid and will not be reconsidered afterwards. Most often, these algorithms can be configured to optimize several objectives, such as reducing the number of kilometers, reducing the resources used, etc. They can also be designed to optimize the number of routes as well as handling a wide range of situations, such as limited time slots, capacities, skills required to complete a task, etc.

For a parcel delivery company, for example, the list of deliveries to be made and the available vehicles are sent to the software at a given time in order to generate an optimized schedule that limits the number of kilometers to travel to do all the deliveries. This schedule can no longer be adapted during the day. Only manual changes are possible, but most of the time, decisions are made between the scheduler and the drivers in real time (or by the driver alone) and the changes are not communicated to the tool, which therefore does not take them into account in its activity analysis and reporting.

Benefits:

  • The algorithms calculate optimal theoretical routes (in terms of costs and constraints provided)

Drawbacks:

  • Inability to adapt the planning to any information that arrives after the calculation has begun
  • Loss of performance of the route plan in the field
  • No follow-up or reporting on the results
  • Fixed planning process requiring all the necessary data before launching the calculations

*Note that the term “static” is also commonly used by Gartner to define weekly or monthly route plans. Gartner contrasts “static” routes with “dynamic” routes, which are routes that are recalculated on the go.

Optimization with real-time tracking and adjustments

Some recent software providers offer the technological ability to combine real-time route tracking with route optimization. This allows the user to check in real time how the routes are evolving in the field, to follow their progress and to adjust to any arising changes.

It is thanks to this type of solution that it is now possible to (automatically) warn a customer a few minutes in advance that they will be delivered late.

The “adjustable” optimization is similar to a “static” optimization, except that it provides users with additional features to help them adapt the schedule to certain events before and sometimes during operations. The main scenarios are the following:

  • Adding a new task (collection, interventions) to be performed: the tool provides the user with several possible rounds in which to add the new task and then inserts the task in the best way possible to the selected round
  • Resource cancellation: the tasks initially planned for this resource are manually distributed by the user to other resources available in the tool
  • Delayed rounds: the scheduler is notified and can cancel the tasks that will not be able to take place and manually insert them in another round for a same-day pickup

In these situations, it is important to emphasize that the software does not make suggestions “on its own”, it is up to users to react to the problems that they can identify from the data that is sent back to them and to manually adjust the routes in the tool. The latter takes into account the modifications and communicates them to the on-field workers, but there is no global investigation of the route plan to anticipate opportunities for improvement that may have arisen resulting from changes in the context.

It is therefore impossible to consider task transfers between two rounds to free up the schedule of one of the drivers and add an additional order. It is also impossible to automatically anticipate delays and compensate for them by reorganizing the rounds to ensure that the time slots are met.

Benefits:

  • The algorithms calculate optimal theoretical routes (in terms of costs and constraints provided)
  • Ability to monitor real-time performance
  • Possibility to inform customers in the event of problems
  • Ability to manually fix certain problems caused by field hazards

Drawbacks:

  • The algorithms do not leverage the optimization opportunities that become available over time
  • Fixed planning process requiring all the necessary data before launching the calculations

The future of route optimization: continuous optimization, known as "Always On”

As previously mentioned, the limitation of optimization to deliver value in the field is that the human decision process is by nature a continuous process. Conventional route optimization, on the other hand, always assumes that at a given moment it will be possible to have all the data necessary to define the optimization problem, and that it will remain unchanged (or only slightly).

However, in practice, knowledge of the field is purely theoretical: it never really happens (or never for long). That is why, in order to remove the last obstacles to the implementation of route optimization, it must work alongside and at the pace of those running the logistics. In other words, it must be continuous.

If optimization remains fully alert throughout the entire logistics process (during route planning and execution) and is fed with real-time data from the field, then it is able to adjust the planning on the go or use new optimization opportunities to remain efficient at all times.

This new approach is explained in detail in our article dedicated to continuous route optimization.

Continuous Route Optimization combines the advantages of mathematical approaches (route performance) while allowing business experts to interact with the routes and provide their expertise. This approach allows for real-time correction of inaccurate data (without having to restart the entire calculation process). The ability to use data that is only known to the operational staff (and does not exist in the database) and real-time re-optimization, taking into account any unforeseen events, allows for better and more relevant routes to be designed once in the field.

types of route optimization

What happens in the field, depending on the type of optimization

In manual planning, the routes are planned by an expert ahead of the driver’s departure. The performance of the routes, in terms of costs (number of vehicles used, number of kilometers travelled) and quality of service (meeting customer commitments, time slots etc.), depends very much on the experience and skills of the human expert, but will always remain below that of a powerful algorithm can provide.

types of route optimization

Once the drivers are on the road, each incident has an impact on the performance of the routes and, without monitoring information from the field, the loss of performance cannot be offset. Field hazards cause the same problems to “static” optimization, which does not allow for offsetting performance losses.

If we now look at route optimization with monitoring and adjustments, we can see that hazards also have an impact on route performance, but not all in the same way: some hazards, which are not managed by the solution, have the same impact as with static optimization. Some other hazards can be adapted for and still others can be partly compensated for by manual adjustments made by the user, who, thanks to the tool, has a clear view of the progress of the ongoing routes.

Finally, in the case of continuous route optimization, hazards have an impact on the routes, but the algorithms are constantly looking for better solutions, which allows them to identify new optimization opportunities based on what is happening in the field. This is how routes can sometimes recover in performance thanks to smart choices that the algorithm provides to users in real-time.

Continuous route optimization solution