Job posting sites are great for finding opportunities but they do little to help job seekers evaluate potential employers.
Introducing Visual Job Search, a tool that uses data visualizations and employee ratings such as company culture, pay and benefits, and work/life balance to help job seekers curate and apply to high quality opportunities.
Over the course of 10 weeks, my team delivered research insights, interactive prototypes, a dataset of ~82,000 job postings and company ratings, and a live job search tool.
Working with two researchers and another designer, I flexed my data science skills in addition to my UX tookit. I was responsible for:
While existing job search tools are great for finding available jobs, they do little to assist in evaluating opportunities by keeping company reviews and job postings separate or minimally integrated. The wealth of employer ratings and reviews online is helpful but combing through to gather insights is a time-consuming process.
Since the human brain is much more adept at interpreting and drawing insights from visual information than text or numbers, my team set out to create a tool that would assist job seekers in evaluating potential employers through the use of data visualizations.
We conducted 6 hours of interviews and surveyed 22 people who had recently landed a job or were actively looking for work to learn about their job search strategies and decision factors.
As expected, all of our interviewees and survey respondents had searched for work and looked at company reviews online.
Job search factors varied widely among our participants, and included considerations such as pay and benefits, company structure, work style, company culture/image, leadership, team size, and diversity. Each participant seemed to have a slightly different combination of factors that they cared about most.
Based on our survey and interviews, we created a representative persona, Tania. Keeping Tania’s in mind helped ground our design and reminded us to design in a way that was flexible enough to support job seekers with different priorities.
To build out an initial dataset, I wrote up some Python code to request data from the Glassdoor Companies API. After cleaning this dataset I was left with 29 data points for each of 9,394 companies, including ratings of various aspects of each company (i.e. compensation, work/life balance, culture and values), company reviews by former employees, information about the CEO and senior leadership, and the industry segment. We used this dataset for our initial prototype.
Glassdoor does not make job postings available through its public API, so for the second stage of data collection I used company names from the initial dataset to query Indeed’s Web Services API, collecting job title, posting URL, city and state, and geolocation data.
After hitting the API request limit and merging and cleaning the two datasets, I arrived at an 85 MB CSV file consisting of 35 columns, with roughly 83,000 unique job postings across 4,200 companies, at 98% overall completion. This dataset was used from our second prototype onwards.
We kicked off our design process by sketching out some possibilities for different ways to visualize and search through the reviews dataset we’d gathered.
Analyzing the initial dataset, we realized there was a potential to create some meaningful visualizations by aggregating metrics related to specific themes.
Based on the priorities expressed by our research participants, we created a “career related’ visualization and a ‘culture related’ visualization.
Our five participants immediately understood that companies in the upper-right quadrant of each scatter plot had the best ratings:
Despite the ease of interpreting the position of each data point, participants had difficulty gleaning insights from the color and size encodings.
Taking the lessons from our first prototype we began rapid iterative testing and evaluation (RITE), producing and testing six prototypes within a week.
Incorporating the updated dataset with live jobs data during this stage enabled us to develop and refine different workflows for job search and filtering, customizable linked dynamic visualizations, a color scheme for effective data interpretation, and supporting copy for the interface. We got pretty good at using Tableau too.
Backed by an extensive database of job postings and company ratings, our final Tableau prototype used data visualization best practices to enable job seekers to uncover insights and evaluate job opportunities before they apply.
A customizable scatter plot shows an overview of the distribution of relevant opportunities based on search terms and user-selected rating dimensions.
Users can narrow down results by company, job title, industry, and location using the filters at the top of the application. By interacting with the data visualizations and sliders users can filter by ratings such as company culture, pay and benefits, and work/life balance from actual employees.
A customizable parallel coordinates plot shows ratings across all metrics for top companies. Users can also hover or click on a specific company to see the associated ratings values and job title.
In the future, I’d like to add some additional functionality to Visual Job Search such as a walkthrough explaining the visualizations, qualitative information from relevant reviews, social network integration that connects job seekers to current employees, and an interactive map of available opportunities.
Although Visual Job Search was originally built in Tableau, I'm currently developing a web application using D3.js and will be posting updates as they come.
Want to learn more about this project?