ANLY482 AY2017-18T2 Group27 : Project Findings / Final
| Interim | Final |
Contents
0.0 Abstract
Visualisations serve as an important tool in effectively analysing data and gathering meaningful insights. The ease of connecting multiple diagrams together in a dashboard sequence, allows users to see multiple connections between different variables to identify plausible trends quickly. Often, managements utilise visualisations to identify valuable insights on their business functions. Thus, dashboards offer a striking solution to an organization's need for information at a glance. Organisations commonly believe that traditional charts such as bar graphs and pie charts are able to effectively convey information. However, this is not sufficient in providing insights and a seamless user experience. To further explore this, a research study is conducted to demonstrate the use of a common organisational visualisation tool, Power BI in developing effective and sustainable dashboard for the air freight industry, and to overcome Power BI’s current technical limitations.
The case study is an in-depth analysis on a leading logistics company and its air freight shipping division. It will utilise transactional data relating to a client of the logistics company over 3 years. Also, the case specifically looks into elements such as rates, density and shipping patterns and tries to develop links that can be further analysed and transformed into actionable business insights. The study introduces the business and primary research motivations. Following this, it looks into suitable literature on PowerBI tools as well as air freight shipping. Next, analysis will be applied to the case and relevant discussions will be presented. Subsequently, designs and final applications will be exhibited. A comprehensive summary on the results will be provided along with recommendations to other organisations as well as Power BI technical limitations.
1.0 Research Contribution & Objectives
Due to better pricing, rapid and continuous improvements, PowerBI has gained popularity among many companies. Our research efforts will focus on PowerBI as a common visualisation software and better visualisation. The objective is to provide logistics company account managers with effective business visualisation tools on PowerBI that will allow them to assess commercial aspects across time. A three-pronged approach will be utilized to fulfill the following analytics requirements:
- To investigate PowerBI’s effectiveness and limitations as a visualisation software; and methods to overcome limitations
- To explore the use of visualisation tools – Quadrant Matrix and Chord Diagram, in analyzing the varying patterns across data.
- To build interactive visualization dashboards that can depict 3 key commercial aspects – Rate per Kg (RpK), Density and Ship to Profile (STP), and how they are used
2.0 Literature Review
2.1 Dashboard Design
Common issues in designing dashboards include exceeding the boundary of display screen, inadequate context for the data, poor data arrangement, misusing or overusing colour, displaying too much details and choosing inappropriate display media. [1]
Exceeding the Boundary of Display Screen
Dashboards that exceed the display screen boundary sometimes require scrolling. According to Few, an experienced and influential dashboard designer, many users do not scroll to view the other data. They assume data viewed at first glance to be utmost importance. [1]
Inadequate Context for the Data
Measures in businesses usually require a comparison to produce meaningful insights. For example, knowing the density of one customer’s shipment is not enough. It must be compared with others. Also, in using median to describe the average of a skewed transaction data, such as weight of shipment by customers, a distribution of the data should be displayed. Viewers can then expect and investigate possible outliers. [1]
Poor Data Arrangement
As dashboard space is limited, data should be arranged according to their emphasis. Few suggested that dashboards should be arranged in the following manner:
Figure 1
Primarily due to western reading conventions, the top left-hand corner is the most important, followed by the top right, bottom left and bottom right. [1] However, the center of the screen is also a region of strong emphasis, due to a more fundamental inclination of visual perception. Few discovered that information in the center is only emphasized when there is a white space surrounding it, allowing it to stand out. [1]
Misusing or Overusing Colour
To effectively highlight data, colours should be used only when necessary. Otherwise, grey or muted hues should be used. As visual perception is highly sensitive to differences, a contrast from the norm can attract viewers’ attention to it. Methods to attract viewers’ attention includes intensity. Also, as visual perception is sensitive to associated meanings, colours have different meanings. Red may mean urgency in certain cultures but not in Chinese. In Chinese culture, red represents prosperity and joy. [1]
Data-ink Ratio
According to Tufte, data-ink ratio should be maximised. Ink on dashboards should be purposeful. Data-ink ratio is defined as 1 ‐ proportion of a graphic that can be erased without loss of data‐information. [1]
Displaying excessive details
Too much details can slow down viewers’ analysis. Also, these excessive details do not value add analysis. [1]
Inappropriate Display Media
An inappropriate display media can confuse viewers and obscure data. [1] To effectively visualize the key measures in our case study, we propose the following:
2.2 Chord Diagram
WTCHG
WTCHG, total air freight for the shipment, appears to follow a similar distribution to WGT, where both regular and non-regular BUs’ mean is higher than its median. This means that their histogram is skewed to the right. This is as expected as air freight shares a positive relationship with chargeable weight.
Interesting to note, in WTCHG, Aviation is more than non-regular BUs, the opposite of the trend in WGT. This may be due to extra shipment during peak seasons or extra urgent shipment requested by Aviation.
RPK
In the overall summary statistic of RPK, those in the 0.05% percentile are between $404.33735/kg and $21986.66/kg.
Only 3 of the 180 transactions are dangerous goods. These dangerous goods are requested by Aviation, specifically Aircraft Engines. As they are from Aircraft Engines, these dangerous goods may be flammable. Their unusually high RPK can be attributed to the additional service or special aircraft provided by Company X to ship the dangerous goods.
The other 177 shipments are non dangerous goods. More investigation for details have to be conducted to reason this occurrence.
8.2 Density
EDA was performed in the following variables of Density's module: Volume of goods shipped in each shipment (VOL)
VOL
The box plot (not shown due to data privacy) reflects a rather distributed and varied field for Aviation customer as compared to the other customers. This is further supported when a closer look at the average volume being shipped is tabulated.
There is also a possibility of an outlier for this customer. A closer look at this (analysis on the proportion of customers) graph indicates that the dataset contains large transactional data relating to Aviation and Healthcare. Healthcare, when compared against the its respective box plot indicates a smaller box plot in comparison with Aviation. This shows that GE Healthcare does not ship heavy density shipments as compared to GE Aviation
To supplement this, we also considered the average volume by customer. From a bar chart of average volume (not shown due to data privacy), we note that there is a lot of higher range volumes which affects the average volume and skew per customer.
8.3 STP
EDA was performed in the following variables of STP's module: Chargeable weight (WGT), Freight Charges (WTCHG) and Rate per Kg (RPK).
BU Contracts
Firstly, we started by comparing the actual shipped weight against the total annual projected weight for each BU.
In general, we see that Healthcare has the highest contractual amount awarded and also, the highest actual amount shipped. It was not able to match up to its projected total amount.
On the other hand, Aviation, Energy Management, and Oil and Gas actual shipped amount exceeded the projected amount.
Considering these insights, the Company X can consider zooming in into these business units to further investigate which country and station lanes are causing these discrepancies.
Region Lanes
Another broad category to look at would be Region Lanes. Region Lanes track the origin region and the destination region. There are 3 main regions – ASPA, EMEA and AMER.
ASPA stands for Asia-Pacific, EMEA represents the European region while AMER represents the American continent.
Figure 15 highlights the actual shipped amount versus the Projected amount for each region lane. Here, we see that for Region Lane Pairs AMER > ASPA, ASPA >AMER, ASPA > ASPA, the actual amount shipped never matches up to the projected amount.
However, in the case of ASPA > EMEA, the actual amount shipped almost matched the projected amount. Additionally, for the region pair EMEA > ASPA, the actual amount shipped outweighed the projected amount.
Region Lanes and BU
Regions Lanes and Business Units are 2 broad categories. By combining these 2 filters, we will be able to gain more actionable insights as compared to looking at each aspect individually.
By using these 2 aspects as legends, the possible combinations have increase as each BU is paired with a Region Lane.
More details will be in our report due to data privacy concerns.
