Many of us have felt the pain taking those first steps to start a career. At the beginning of the journey, you find yourself full of hope and positive vibes. This can soon start to fade once you realize that after hundreds of applications you’re still empty-handed.
One of the most useful features of LinkedIn is the simplicity with which you can apply for hundreds and thousands of jobs. But, keeping track of all your job applications is anything but straightforward! Monitoring them precisely not only allows you to unveil interesting patterns but also improves your job hunting skills. That's why we decided to build a browser-based data app that visualizes insights about your LinkedIn job data in a meaningful, interactive way.
Today, we want to walk through the workflow behind the data app, show you how to request your LinkedIn data and use the data app with your own data!
Try out the data app using some sample data here. The gif below shows what to expect.
Use the Data App with Your Own LinkedIn Data
The first page of the data app gives you the option to upload your own LinkedIn data. You have to request this file from LinkedIn, which you can do by following these steps:
Open LinkedIn and click the profile icon (top-right).
Go to “Settings & Privacy” and then “Data Privacy”.
Under the section "How LinkedIn uses your data", click “Get a copy of your data”.
Select the “Download larger data archive, including connections, contacts, account history, and information we infer about you based on your profile and activity” option.
Congrats! LinkedIn will now start preparing your data. You will receive your LinkedIn archive within 24 hours.
The files of interest are those referring to the connections (Connections.csv) and the job applications (Job Application.csv).
Please note that the files concerning the job applications will be stored in a dedicated folder. If you’ve applied for multiple jobs, you’ll have multiple files.
How We Built the LinkedIn Data App
Our LinkedIn data app was built with the goal of enabling the user understand how their application process is going at a glance. We also wanted the data app to be customizable, interactive, easy to share, and consumable as a web application.
KNIME data apps are built using KNIME Analytics Platform, where we can collect, process and visualize data by drag-and-dropping and connecting nodes from the Node Repository (Fig. 3). We can define the desired level of customization, interactivity, and easily export our workflow to share it with others on the KNIME Hub. To make our KNIME data app available as a web application we'll deploy it to the KNIME Business Hub.
Enable Data App Users to Upload LinkedIn Files
When we use a browser-based data app, each “page” of the application is defined by the components containing nodes with an Interactive View. You can set a layout for each element in the page, with each component composite view organized using the Node Usage and Layout window in KNIME Analytics Platform.
The first page of the data app prompts the user to upload their LinkedIn personal files.
Our main goal here was to ensure workflow portability and generalizability, avoiding an instance-based workflow that would not work with other data. To this end, the workflow control features in KNIME, such as flow variables in combination with the switch and widgets nodes were crucial. Flow variables gave us flexibility. Instead of using predefined values, they change dynamically together with the data. And using switch nodes it was possible to create parallel branches depending on the flow variable’s value, introducing a little error handling logic.
Users can upload their personal files using the File Upload Widget node wrapped inside the File Upload component. The result is a simple web page to upload all the files you have in just a few clicks.
Some of the uploaded files contain a disclaimer at the top, which can make auto-detection of the right file format troublesome. However, exploiting the configuration options provided by nodes in the KNIME’s file handling framework, we can easily customize the behavior of the CSV Reader node to allow for a graceful handling of unnecessary rows.
Each job application file can contain at most 200 observations. We know, however, that job-seekers often end up applying to more than 200 positions before they get hired. To account for that, we allow the user to upload up to 5 different job application files. Next, using the Concatenate node, we can create a concatenated table of all job applications for further processing.
Engineer Features, Mine Texts, and Wrangle Dates
After importing the data and solving data file readability issues, we process the data to extract insights job-seekers may look for. LinkedIn does not provide a comprehensive analysis of the job application history, especially when it comes to the questionnaire that most companies require for the job applications.
The data processing and preparation of the workflow addresses this issue. To discover hidden information in the data, we adopted a heterogeneous approach, using operations from text analysis, feature engineering, and date&time manipulation. The complexity of the workflow was then wrapped into multiple metanodes to ensure the workflow is easy to follow.
Starting from the job position information, we used feature engineering techniques to extract a “detail score” out of the position title.
This score is built starting from the job position name’s length: if the position’s length is longer than the average, then a positive value normalized between 0 and 1 is assigned. The detail score helps applicants understand how broad or specific job positions on LinkedIn are adjusting their expectations. Additionally, an “urgency score” was also created filtering applications based on whether or not the urgency of the company to fill the job position is mentioned during the questionnaire.
As we mentioned above, the extraction of insights also involved applying text mining techniques to the applicant’s questions and answers. We focused especially on the most frequently asked questions and/or topics revolving around the term “salary”, computing the word frequency of that term and matching the question in which the term appeared (see Fig. 10).
We visualized the retrieved information in a word cloud for ease of interpretation (read more about it in the next section).
With the preprocessing steps implemented above, we can now get deeper and more informative insights into what the companies are interested in while hiring new employees.
Provide Connections and Job Application Insights at a Glance
The core of the data app is represented by two dashboards visualized as web pages:
The first one elaborates a few insights about connections and job application data. The output is a short compact version of the visualizations;
The second page dives deeper into the job application and connections data, including additional options for interaction and granularity.
The first page contains a “snapshot” visualization of your job application and connections data. This is extremely useful for visualizing at a glance some of the most insightful categories of the data provided by LinkedIn.
On the right side, you will find the visualization about the job applications while, on the left side, you can visualize the connections charts. Below there’s a short description of the meaning and usefulness of this dashboard:
Job positions of your connections. Having a network of connections working in a similar field to the one you are interested in might help you find a job more easily, or receive the right suggestions. Thus, it is useful to realize what are the most frequent job positions of your connections.
Time trend of connections. It is important to monitor the always expanding network of people around your LinkedIn profile. A more dense and spread network will help you in your professional life.
Job positions with more applications. While applying for many jobs during a short span of time, you may not be completely aware of all the job positions you applied to and with which frequency. This visualization allows you to keep track of the job positions of the application, letting you know both the variety of your applications and the availability of job positions.
Get a Deeper Dive into LinkedIn Data
The last and main page of the data app capitalizes extensively on the interactivity of KNIME dashboards.
The data app user can change the settings and play with different visualizations. Below you can find some examples of the data app, with a special focus on the interactive parts.
The gif below shows how to interact with the data app to change the number of questions you want to visualize in the most frequently asked questions shown in the table.
The next gif shows how you can interact with the data app to select a time range, for example the number of job applications in the last month, year, or week.
This next gif shows how you can use the data app to visualize the most urgent job applications. The workflow finds those applications that mentioned the urgency of the application in the questionnaire.
A more articulated visualization is the one about the network of connections, which displays the network of all your connections that work in the top 3 companies by number of employees (Fig. 16). This lets you visualize where the majority of your connections work. This is possible using just a few nodes thanks to the KNIME Network Mining extension.
If KNIME is one of the three companies represented in the network, all your connections working at KNIME will be visualized with the yellow-KNIME color instead of the default blue.
Last but not least, if you want to make your data app look even fancier and customizable, you can use the Data App Flowchart. It’s a KNIME verified component and adds a flowchart header to all your web pages, improving the aesthetics of your data app for a better user experience.
Try these Tips & Tricks in Your Data App
Include a high-level of interactivity: The design of the LinkedIn data app relies on a combination of charts, tables and text, where the user can enjoy a high level of interactivity, and the freedom to customize their visualizations, for instance changing the type of plot or the variables in it.
Ensure a flexible design: Without diving deep into all the sections, a great example of the capabilities of KNIME Analytics Platform can be found in Fig. 17. In this section of the workflow, we used widgets nodes to let the user pick the number of questions that they want to visualize (Refresh Button and Integer Widgets). Then, based on the number of rows the table has, two different tables are displayed. This describes perfectly the process of designing the data app in such a way that it would fit all different data instances.
In addition to the above, we list below some useful, advanced tricks to improve the final rendering of your dashboards.
Tag Clouds customization: The default look is already good enough, but playing around with the display options unveils a lot of features to create a tailored tag cloud. For instance, the font family of the words, the orientation of the words, and the word colors are all features that can be customized.
Learn how to assign colors to bars in a bar chart.
Color assignment: It is possible to change the color of all charts using the Color Manager node. It may be a little tricky at the beginning but the result is definitely worth the effort. Take also into consideration the usage of loops (Chunks Loop nodes) to make the color assigning smoother and more flexible (Fig. 18).
CSS styling: Sometimes the default styling options in your dashboard may require that extra kick to look as you envisioned them.. If this is your case, you can overwrite the styling options of graphs using a customizable CSS sheet included in a dedicated node (CSS Editor node). The customized CSS will be propagated to the view nodes via a flow variable.
Widgets: The widget nodes allow you to do a lot of things that otherwise wouldn’t be possible: (i) providing input parameters to be used in downstream nodes; (ii) enhancing interactivity giving the user the options to input custom settings; (iii) including text outputs that can be customized using HTML tags (Fig. 19).
Learn more about KNIME Widget nodes.
The Power of Visualization for Quick Data Insight
In this article, we discussed how to obtain beautiful data visualizations starting from simple raw data that anyone with a LinkedIn account can request. We went through all the steps build a publicly accessible, multi-page, interactive data app.
Creating a data app with KNIME is not only fun but also useful. Each visualization contains valuable information, either implicit or explicit. We hope you have fun using it in your job finding process.
Download and explore the workflow behind the data from the KNIME Community Hub.
