Data Relationships of Remote Density and Company Evaluation
The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Data Science Background
- Data is scraped in this project. Recently, a lot of companies have started posting job openings as remote. There are many benefits to allowing work at home, including fewer interruptions, a quieter room, less (or more efficient) meetings, as well as less commuting time for employees that makes them more likely to work longer. These benefits would lead to higher productivity.
- In business, productivity can be calculated by dividing outputs by inputs where inputs may be labor and capital, and outputs may be products and services. Therefore, minimizing the essential work at the office and expanding remote work will decrease the input cost and increase the output results.
Data Science Objective
- Even though there are many positive perspectives of remote work, it's not easy to know how remote employees evaluate the job or the company. The goal of this project was to bring to light any relationship between the remote density and the company's evaluation. The remote density was calculated based on the job openings in a job search engine, indeed.com, as (remote openings)/(total openings), and the company's evaluation was scraped from the company pages that indeed.com shared.
figure 1: search result example from indeed.com
- As shown in figure 1, the job opening result requires a job title and a location. Some popular remote jobs such as 'data scientist', 'web developer', and 'accountant' and the top 10 biggest population cities were used in the project.
Data
- This project used the module BeautifulSoup in Python for web scraping to gather data.
-
figure 2: collected data from top: company's webpage, bottom: search result
- Figure 2 shows the raw data of scraping results from the company webpage and the search result pages. At the company evaluation, it shows the name, overall rate with the count of each star, and the salary satisfaction percentage. The job posting result shows the job title, the company name, overall rate, the location, remote availability, company web address. Since the data have many improper values, some preprocessing methods, such as dropping NaN value, changing data type, and replacing unnecessary symbols, were applied before checking the relationship.
Result
figure 3: scatter plots for top: remote density vs overall rate, mid: remote density vs salary satisfaction, bottom: salary satisfaction vs overall rate
- In figure 3, the plots show the relationship within variables of 'Remote density', 'Overall rate', and 'Salary satisfaction'. 'Remote density vs Overall rate' and 'Remote density vs Salary satisfaction' don't have a strong relationship within the variables, while 'Salary satisfaction vs Overall rate' indicates a positive relationship. There are many 0 or 1 values in remote density, which means that none of the job postings from a company are remote or all of the postings are remote.
Conclusion
- It was difficult to know the actual remote density in a company since the job data were only gathered from job openings in a job search engine. That means the data represented the whole company while it's only a part of the whole job status. This lack of job opening data caused poor results. On top of that, remote density was calculated for the future job openings and the overall rate, salary satisfaction were evaluated from the current or previous employees. Therefore the remote density couldn't have a good prediction with other variables when overall rate and salary satisfaction have a relationship.
Future
- It is needed to collect the data from each company's current remote status. However, it's difficult to know the whole remote ratio from a company. The future project can focus on a specific branches within a couple of departments. For example, the parameters may be limited for a location to New York, and job as data scientist related only. Then, it's possible to compute the remote density for data scientists and compare the rate for the data department of the company. Having these data for all companies in New York will return an improved result on the relationship between the remote density and the company evaluation.
About Author
Jungu Kang
View all posts by Jungu Kang >Leave a Comment
No comments found.