As part of my Specialist Diploma coursework, I had to design a Data Visualization dashboard using Qlikview. The module was titled “Data Visualization Fundamentals”, and the key focus was on the visualization of data using QlikView, instead of using the dashboard to perform data analytics/sense-making.
Note: this is an assignment for grading purposes. Data sets provided/used do not accurately depict real world situation
The backstory is as follows: Due to the recent increase in the number of consumer complaints in the financial products and services sector, the U.S. government has tasked its consultant vendors to analyze and identify key insights from the relevant data sets.
The dashboard aims to help Business Analysts analyze complaints data from the financial products and services sector in the years 2011 to 2017. It allows for Informative Reporting, Historical Trend and Distribution Analysis, Pareto Analysis, What-If Analysis, and Geographic Distribution Analysis.
The following provides a brief description of each sections in the dashboard:
- Cover Page – An introduction of the dashboard, comprising of the title, and recommended screen dimension for optimal experience
- Executive Summary – Provides key statistics for critical decisions, such as the total number of complaints, average unemployment rate, and top complaint issues/companies
- Summary Dashboard – Allows for slicing and dicing of complaints data, with multiple filters, such as by product, date, company, and complaint issue. Key charts include historical trend analysis, and Pareto analysis.
- What-If Analysis – Projects complaint data through the use of intuitive sliders to adjust key metrics to perform “What-If Analysis”
- Geographic Distribution – Shows the geographical distribution of the complaints across the U.S. using an interactive heat map
The data sets in their original form were not perfectly aligned, and several data pre-processing steps were required to ensure that no synthetic keys and/or intermediate tables were created. Additional steps were required to extract the year, month, and day fields from the dates. The following diagram illustrates the relationship between the data sets:
Dashboard Design Considerations
In order to accommodate the most number of displays, the dashboard has been designed to fit a screen resolution of 1024×768. In addition, a dark background with light text is used to reduce user eye strain when interacting at prolonged duration.
To ensure consistency and intuitiveness, the visualizations use the same color scheme to represent the same data across multiple charts. Easy to understand charts were selected so that the user does not need to spend a lot of time trying to learn how to read and interact with them.
Additional functionality is hidden so as not to clutter the screen and distract the user from the analysis. However, they can be easily brought onto the dashboard with just one click.
Features that are not relevant to the current analysis are disabled so as not to confuse the user. For example, when drilling down into the Product and Sub Product, the Pareto Chart button is disabled if only one product and sub product is selected.
The Dashboard was designed with two groups of audiences – policy makers, and Business Analysts.
For the first group, the “Executive Summary” sheet allows for the reporting of important information that are critical for policy making (i.e. the total number of complaints, the average unemployment rate, and the Consumer Price Index / Inflation). It also allows for the slicing and dicing of key metrics (i.e. Complaint Issues, and Companies) that span across the entire sector.
For the second group, the “Summary Dashboard”, “What-If Analysis”, and “Geographic Distribution” will help identify Consumer Complaint trends, correlation between Consumer Complaints against other factors, and the top complaints that are responsible for 80% of total complaint cases (Pareto Analysis).
- Cover page omitted as it only comprises of title of dashboard, and course/module code.
- Note: this is an assignment for grading purposes. Data sets provided/used do not accurately depict real world situation