For a labor-intensive industry, making a budget to determine the operating costs for the next fiscal month can often be simplified into the following formula:
成本= 员工数目 * 人均开销
Among them, per capita expenses include salary, training expenses and office expenses. These expenses are often relatively fixed or easy to calculate and can be regarded as quantitative.
Therefore, the remaining variable is the number of employees, also called HeadCount. The key to predicting costs lies in predicting the number of employees.
The number of employees depends on the workload. For example, in the after-sales industry, the work progress of employees will be tracked through a software record, so the labor hours (Labor Hours) of employees in the same period in history can be obtained. For example, there are 22 working days in a month, and after-sales service this month The service requires a total of 2200 working hours, so the number of employees required is approximately 2200/(8*22)=12.5, which is approximately equal to 13 people.
But in reality, it is impossible for employees to use up all 8 hours of working time. They will have meetings, training and leisure time. This leads to the concept of efficiency value. For example, according to statistics, an employee spent 5 hours on a certain day. hours of work, then his efficiency value is 5/8. So the formula for calculating the number of employees is as follows:
员工数 = 该月需要的工时/(8*工作日数目*效率值)According to this formula, if we determine the "working hours required for the month" and "efficiency value", then the number of employees can be determined, and then the cost can be determined and the budget completed.
Here we talk about how to use data analysis methods to confirm efficiency values. First, we need to clarify what factors are needed to predict monthly efficiency values. The factors I use include:
The month, the number of working days in this month, whether there is a Spring Festival holiday in this month, and how many users' service contracts have expired this month.
Generally, users will always find ways to suddenly consume the remaining service time of the service contract before the expiration of the service contract. The resulting service requests are often relatively simple and have an impact on the efficiency value.
After determining these factors, let's talk about how to predict the efficiency value using R language.
First of all, before processing a large amount of data, you can use a clustering algorithm (or artificial distinction) to divide the data into rough lines, and then predict the subsets obtained by each cluster. This is a little trick. For example, the after-sales service of a software company can It is divided into cloud service after-sales and traditional software after-sales. When forecasting the budget, because the labor costs in different countries are different, they can be forecasted separately.