IS428 AY2019-20T2 Assign KANG HUI YUN

Overview

Background

In the interests of faculty, students and staff of Singapore Management University (SMU), an all-encompassing survey is conducted by SMU Libraries every 2 years to learn about their perception of the libraries (namely Li Ka Shing Library and Kwa Geok Choo Law Library), with particular focus on the importance and satisfaction level of different library-related services. This piece of information serves as a valuable input for the department to calibrate the standards of the facilities by identifying areas where users are underserved and those that exceed expectations but are relatively less important.

Objective & Tasks

After closure of the survey, a Library Survey Report is generated for consumption. However, this traditional method of analysis comprising huge number of tables and ineffective charts can be frustrating and difficult for users to grasp. Hence, the objective here is to make use of existing tools and technologies to develop an interactive visualisation of the survey data (Feb 2018), such that it is intuitive, easy to comprehend and telling at the same time. By cutting down on the report's "verbosity", the people in charge can save time deciphering trends and underlying patterns of the data with high degree of conciseness and accuracy.

The primary task is to reveal the level of services provided by the 2 SMU libraries as perceived by the following four groups: the undergraduates, postgraduates, faculty and staff. This mainly includes the concept of the Importance-Performance (or Importance-Satisfaction) rating as well as the Net Promoter Score (NPS).

Survey Data

About the Data

The data for the 2018 Library Survey Report consists of:

• 2639 records (rows)
• 89 fields (columns), which includes
  • A new column for ResponseID (unique to each respondent)
  • All other 88 columns related to the survey questions

The survey is predominantly made up of importance and performance rating (26 such questions, represented by 78 fields in the dataset). The 78 fields are spread across 3 main categories: importance, performance and N/A (hence 26x3, giving us 78). Respondents who picked 'N/A' for the question need not select a rating for both importance and performance. Prior to these questions, there are 4 basic background and demographic questions that are all non-quantitative. Also, there are 3 questions on the frequency (of visits and access), 2 rating scales on overall satisfaction and likelihood of recommendation, and finally one textual input for comments and suggestions. For the most part, the survey questions are based on Likert scale (questions with suffixes between 1-27, for both Importance and Performance). Worthy of mention is yet another rating scale—the Net Promoter Score (NPS), which, when calculated can tell us the overall (i.e. net) satisfaction/loyalty of library users.

Data Cleaning & Preparation

Before jumping into data visualisation, it is important to make sure the data is clean, rightfully organised and structured. This will facilitate the visual development using tools such as Tableau, allowing for categorisation and sorting. In the cleaning process, we shall engage in steps such as removal and/or creation of new fields, field renaming, and pivoting. Finally, we will piece up the bit-sized information into a single dataset via merging (e.g. join/union). Ultimately, we would like our data structure to be thin and tall (i.e. less columns, more rows) so we can analyse certain dimensions easily as we wish.

Initially, I have used Tableau Prep to clean the data. However, I quickly switched to using R (on RStudio) mainly because recreation of the same process to generate the final tidy dataset is tedious—it requires a lot of manual work, especially clicking and entering values. Writing a single R script allows me to automate the execution of the desired actions. Nonetheless, I have successfully created the same dataset using both methods.