One of the things where we waste more time planning staff and tasks is not so much the initial generation of staff, but the adjustment of employee preferences that forces us to make changes, including many are made as employee requests because they are not managed by managers. These changes are often not produced by the burden of the controls to be planned. However, not doing them implies less satisfaction and less performance of workers in their work. An example might be a worker who enjoys taking his son to English classes on Tuesday afternoons. This level of detail is not recorded, usually only legal restrictions are set and the supervisor does not have time to adjust all these details. Thus, the worker leaves his post earlier, has less productivity and increases his stress so as not to move the Tuesday afternoon shift. Although they are not saved, what can be done is to analyze the data for consideration in future planning in a transparent way for managers.
What data is used to know the preferences if they are not saved? In this initial phase, an overall preference detection process was followed to be planned, validating which data may affect preferences. The relationship between initial and final planning is analyzed and the variables that most affected or had more weight to affect planning. The most weighted data that we will use to estimate preferences are:
• Initial workers’ plans or pre-established employer
• Modifications of the worker’s schedule by the managers
• Planning changes brought about by employee requests
• Employee leave requests approved
• Canceled employee leave requests
• Employee absences
• Incidents with presence control
The neural network receives all of these inputs, and when a schedule is done, it identifies if this mismatch on that day for that employee is an issue with the preferences of the same during all so historic. It’s one neural network per employee, which complicates development and this is the current phase where we are in development. To be able to develop a network per worker and which learns only from this worker or to be able to create a more global neural network where we can use the learning of certain workers for others. Both scenarios are undoubtedly interesting and have advantages and disadvantages.
The neural network helps us to define which planning constraints are desirable in planning and can be taken into account. That is, in teams we have work restrictions such as 12 hours off or not exceeding the weekly hours by contract which are mandatory, but there are other desirable ones which can be considered in the planning. flexibly and helping to integrate into working life. and personal.
In the virtual event that will take place on Thursday, January 14, we will talk about the technological advances that we have made throughout 2020 as well as those that we will make during 2021. But we will not only see the technological aspects of the changes, but also the implementation process, the worker’s experience with our tool, the strategic partnership program, etc.