Labor demand planning involves placing the right people at the right time. This saves warehouses money. Predicting demand spikes and scheduling shifts to reduce costly overtime. Improving operational efficiency yields indirect gains. Working late frees up docking space, reducing congestion.
It's all about getting the right people with the right skills to the right place at the right time when it comes to labor demand planning. In warehouses, getting this right can save a lot of money. For example, choosing shifts that end sooner, which are usually less expensive than night shifts, or anticipating demand spikes and scheduling shifts accordingly to reduce overtime hours, which are often more expensive than regular working hours. These are costs that can be cut directly. Gains can also be made in a more indirect manner by increasing operational efficiency. Working as late as possible reduces the amount of occupied space in docking areas, resulting in less congestion.
These decisions can include determining how many full-time employees are required versus how much variable work is assigned to more expensive part-time or last-minute shifts (also called flex shifts). We must also consider the availability of certain resources (for example, forklift trucks) that are required by some workers. Some people have specific skills that are linked to limited resources (e.g., forklift drivers).
We must keep in mind that later shifts (e.g., at night) are often more expensive than shifts during normal working hours when deciding to plan certain tasks later to reduce congestion in the docking area.
All of these factors are difficult to consider separately, and even more difficult to consider together. When warehouses have hundreds of employees and demand fluctuates dramatically from week to week, this problem becomes so complex that manual planning is no longer an option. This results in a significant amount of operational efficiency being left on the table in warehouses all over the world.
We use machine learning at Solvice to forecast seasonal, day-to-day, and hour-to-hour demand patterns. We use our shift creation algorithm to assign the most cost-effective shifts while taking into account all of the constraints mentioned above and more.
In many warehouses, shifts are manually scheduled, usually in response to recent trends or reactive responses to problems with the previous week's schedule. As a result, the shifts are mostly static or only get better over time. Labor demand forecasting is usually very simple when done by hand. Even on the calmest days, fixed shifts are scheduled to meet an expected base demand, which is the bare minimum of work that must be completed. Flex shifts are used to cover demand spikes that are expected. To compensate for unexpected spikes, overtime is used..
When people are present at the right times during the week, the entire warehouse operation becomes more efficient. This ensures that spikes in workload are met by increased worker presence when needed, that key areas are less congested by completing work later, and that management is less burdened by reducing the number of shift start-times.
In a variety of ways, employee satisfaction can be greatly improved. Reducing overtime makes warehouse employees' workweeks more predictable, resulting in noticeable quality-of-life improvements. Workers can choose from a variety of shifts created by algorithms, giving them control over how they want to structure their week. Some people prefer 35-hour work weeks, while others prefer 50-hour work weeks.
Shift costs can be reduced directly by choosing cheaper shifts, as previously mentioned. A variety of factors influence shift costs. In most cases, it depends on the time of day the shift occurs and whether it occurs during the week or on the weekend. Cheaper shifts usually come with some drawbacks. To balance the benefits and drawbacks of different shifts, we must first define how to assess the value of choosing cheaper shifts at the expense of operational efficiency.
There are two issues that need to be addressed in this situation. We'll need a precise forecast of future demand for a specific time period, and we'll use that forecast to create an optimal schedule for the next few weeks, determining who we need in the warehouse when.
Demand prediction and shift creation are two aspects of the overall labor planning process. The forecasting of demand is the first stage. We use our demand forecasting solution, Delphi, to predict incoming work before we start creating shifts.
We use this prediction as the basis for creating the shifts after predicting the demands for the coming week(s). The Shift Creation Solver, one of the products in our OnShift API, is responsible for the second phase. The Shift Creation Solver uses the Delphi output directly as input, along with additional information about shifts, resources, skills, and so on. After that, the Shift Creation Solver generates the final output.
In addition to predicting the volume of work, we must also forecast short-term trends. Let's say a large number of orders must be completed by 4 p.m. on Thursday. Over the course of Monday and Tuesday, these orders arrived. These are patterns that may differ from one warehouse to the next. Only if we can predict this trend ahead of time will we be able to schedule productive shifts and avoid missing our deadlines on Thursday (or hiring flex workers ad hoc).
The majority of the issues with assigning shifts based on anticipated demand have already been addressed. We want to emphasize that the main challenge is to balance all of our various considerations against one another in order to arrive at a great solution. Each of the aforementioned constraints (scheduling work as late as possible, choosing cheaper shifts) is simple in and of itself, but not so simple when they must all be considered at the same time. The following is a major difference between how algorithms and human planners handle these considerations.
When a minor detail of the current situation changes, human planners will frequently make minor adjustments to their schedules to accommodate these minor changes. Algorithms will begin from the beginning each time. This means that, if desired, an algorithm can create completely different schedules for each situation. Of course, completely rearranging people's schedules every week isn't always desirable, and our creation solver has features to prevent this if necessary.
