Difference between revisions of "ANLY482 AY2017-18T2 Group13 Analysis & Findings Final"
| Line 174: | Line 174: | ||
==='''Employee Training Dashboard Implementation'''=== | ==='''Employee Training Dashboard Implementation'''=== | ||
| + | |||
| + | Training Placement | ||
| + | As the sponsor company records each attendance of an employee to a training session as training placement, identical training course names can be provided in multiple sessions, especially regular orientations conducted for new employees. Thus, a new unique identifier column Training Placement was created in a format of “Course Name__StartDate__EndDate”. | ||
| + | |||
| + | Training Course Grouping | ||
| + | The training courses are categorized into different groups according to their purpose. For example, the courses designed to educate employees with safety standards and environmental compliances are grouped into Course Type HSEQ (Health, Safety, Environment & Quality), and those to provide operational guidance are under specific operation group such as Packaging and Warehouse. HSEQ course type showed the highest number of training placement since it consists of basic and essential courses usually undergone by new employees. | ||
| + | |||
| + | Training Expenditure and Budget | ||
| + | Each training placement incurs certain amount of course fee depending on whether it was provided internally or externally. | ||
| + | |||
| + | |||
| + | Fig 9. Snapshot of Course fee plotted against course budget | ||
| + | |||
| + | Total course fee for the current and previous year was plotted as a simplified bullet chart, based on the current system date. Using a marker, total course expenditure to date can be compared to planned budget. | ||
| + | |||
| + | Distribution of Training Hours by Employee | ||
| + | |||
| + | Fig 10. Snapshot of Distribution of training hours by employee Histogram | ||
| + | |||
| + | A histogram was used to provide with an overview of training distribution, which shows the number of hours that the employees spent for their training. Total training hours in the current year (Sum of Hours per employee) was calculated for each employee. Using bins of 20 hours, the height of each bar represented the frequency of employees falling into each bin of training hours. As such, given the maximum total hours per employee is 600, we can observe that a majority of employees undergo a total of 1 to 30 hours of training. | ||
| + | |||
| + | Given the limitations of Power BI, a histogram was utilized instead of a density plot to describe this distribution of training hours. | ||
| + | |||
==='''Target Performance'''=== | ==='''Target Performance'''=== | ||
Revision as of 14:45, 13 April 2018
| Interim | Final |
|---|
Contents
Abstract
Effective monitoring of employee retention, training cost and predicting future training demands are some of the main challenges the Human Resource Departments face. This is especially evident in a labour-intensive environment present in logistics and manufacturing industries. A solution that this paper offers is to build a strategic dashboard to aid the management in overcoming such challenges. As such the final paper aims to build a user friendly dashboard that meets the needs of a human resource department.
In-depth research on the selection process of visuals for the dashboard to best deliver the most salient information is covered in this paper. In addition, the use of interactive dashboard features such as bullet graphs, mosaic plot, slicers and drilldown filters are explored and reviewed. Justifications for the use of Power BI are elaborated upon before proceeding into the development of a case specific dashboard for a logistics firm in Singapore, covering 3 main areas - Employees, Training and Performance. One limitation encountered is the lack of staff level performance tracking, hence the study seeks to recommend performance tracking at an individual staff level to establish meaningful data collection for a more accurate assessment of training effectiveness.
Section I
As the use of business intelligence expands across businesses, business intelligence (BI) systems such as visual dashboards have become a critical component for companies in optimizing business operations[1]. Visual dashboards offer a concise visual representation of key indicators of the business with up-to-date performance measures. They serve as a powerful tool for employees to have a quick overview of valuable information critical in making key management decisions.
Our research stems from the need of research material in building interactive dashboards compatible for companies unfamiliar with business intelligence systems and seeking to implement visualization dashboards at a management level.
i. User-friendly Design The first objective of the dashboard is functionality. A good dashboard design provides a good user experience by maximising usability and ease of navigation. An uncomplicated design should capture the necessary top-level information while equipping the user with friendly navigation tools to drill down further to access more details[2].
ii. Close monitoring The second objective is to develop a dashboard that aids close monitoring. With the use of interactive visualization dashboards companies can now progressively monitor these key indicators monthly or quarterly as new data is being collected. This would enhance firm’s responsiveness to changing business needs and conditions. With a shorter response time, the management can stay ahead and anticipate business problems pre-emptively before they aggravate and translate into problematic complications or major losses to the company[1].
iii. Future Planning Thirdly, the objective of the dashboard is to explore the functionality of predicting business needs. Forward planning based on current circumstances and field-specific knowledge can help the firm strategize and set aside resources to meet future demand.
iv. Cross Department Analysis and Investigation Lastly, this paper aims to explore dashboard designs that enables the observation of relationships of inter-department data. For example, human resource training data may be combined with performance reports from the operations department to identify useful relationships that would otherwise remain isolated with the respective departments. Abnormalies in data may also prompt the management to perform further investigations.
In a data-driven business landscape, there is a need to go beyond statistical data and into effective communication of data for greater analytical results[3]. Visualization dashboards have become key tools used in modern enterprises as they provide analysts with critical business metrics that reflect the performance of the business[4]. By centralizing different data, dashboards make data easily accessible to the management where it can be drilled down to make cross department analysis and target comparisons. Users can thus monitor key risk and performance in a single screen to make decisions and actions that mitigate risk and improve performance[5].
The phenomenon of a surge in year-end spending is commonly observed and such a malpractice results in a lower quality of resource utilization compared to spending during other times of the year[6]. In response, tight supervision over resources are often required to helps firms follow their budget plans effectively and efficiently. Visualization techniques can help management make periodic adjustments regularly to alleviate such problems. For example, remedial actions can be undertaken should the dashboard signal warnings when the realized expenditure exceeds or fall below the ideal amount.
The utilization of such business intelligence extends to individual departments such as human resources. Some of the key areas of human resource departments that should be measured include recruitment, retention, capabilities and training expenditure planning[7]. In a case study , CAPPE08, the team also proposed to organize visualization by levels for managers to view main indicators first before drilling down into having quantitative or qualitative focus. For example, rate of employee turnover can be a main indicator while turnover by department would be part of the quantitative drill-down. Qualitative drill-down, on the other hand, provides managers with more information of a particular first level indicator[7] .
Graphical representation is key to visualization dashboard designs. Common dashboard visualizations include gauge charts, pie charts and histograms, They aid in visualizing target performance, part to whole relationships and distribution respectively[8]. However, such graphical representations have their drawbacks. Graphs in a circular shape, such as pie and gauge charts, do not fully utilize the dashboard space resulting large. Furthermore, interpreting the angles of such graphs require higher cognitive effort than interpreting lengths or visualizations with parallel positions relative to a common baseline[9].
Removing Information Clutter Beyond mere aesthetics, an intuitive dashboard is key in avoiding misinterpretation of data. Cognitive barriers that may hinder understandability should be eliminated. One example are 3D charts that may potentially distort the perspective of the chart and lead to a wrong emphasis of values. Briggs suggests that 2 to 3 charts should be created from the same set of data to display different comparisons[10]. This decomposes the information into digestible bite-size. Choice of Colour, Size and Shape Distracting colour schemes may confuse the user hence colour choice is important. Additionally, the interpretation of colours varies across cultures[11]. For example, red generally indicates stop, bad or attention is required in the Western culture. Sizes should be used with caution as the size of a circle may suggest the relative volume. Moreover, the gradient shading of a shape may suggest the use of a different measure. Colours also serve as state indicators which provide visual cues to users representing a categorical data (For example – Poor, Satisfactory, Good).
Section II
1. Bullet Graphs [Insert bullet graph picture] As compared to the gauge which can only hold 1 measure, bullet graphs is able to feature 3 variables within a constrained linear space, thus providing a richer display of data in a smaller space[13]. The bullet graph displays a single quantitative value in the form of a bar chart. Different background fill regions allow users to evaluate the indicative range (poor, satisfactory, good) that the current measure falls in. Furthermore, a symbol marker can be included on the bar graph to include a comparative measure. Additionally, users can compare across measurements quickly as bullet graph are in a parallel position.
2. Sparklines and Line Graph [Insert spark line picture] This display format is useful for displaying trends over time[11]. Such data-rich, word-size graphic enables the user to grasp the historical context of the information. Maximum and minimum values can also be highlighted at a glance while hiding the complexity of the details from the user unless requested in the form of a drill down[12]. For more details, full line graphs would be more suitable for comparisons and can be offered in a drill down function.
3. mosaic Plot [insert mosaic plot] The mosaic plot features a part-to-whole relationship, representing the proportions of a dataset in accordance to 2 variables. As compared to bar graphs which is restricted by a univariate plot, a mosaic plot can display multivariate categorical data[14]. The object which depicts the whole contains smaller parts to provide a high dimension visualization within a constrained space. Furthermore, with a regular shape, the mosaic plot is more space efficient compared to other part-to-whole graphs like pie charts. However, when featuring many categories within each variable in a mosaic plot, smaller parts to be less visible[15].
The datasets used to demonstrate the use of the dashboard covers three areas – training, employees and performance. The master staff list records that employees’ information including the employee identification number, designation, join date and leave date. The training records keep track of the training details of the trainee, trainer, and the training courses. Performance is measured by the key performance indicators of the firm (TnD) that are focused in productivity, quality and safety.
Employee data provided were tracked on a year to year basis in separate documents. In preparation of the dashboard, employees that have been present from 2013 to 2017 in the raw staff list provided were compiled to form a master staff list. An additional column to record the date left was included. This suggested format would help to make the computation of employee statuses easier.
The training records used were extracted from the raw data of the 2016 and 2017 training records. As yearly budget was not available, a sample budget was used instead to visualize expenditure monitoring.
Due to the limited access to monthly performance data, performance data was extrapolated from the 2014 monthly TnD raw data and relabelled as 2018. A sample 2017 TnD data was randomly generated with reference to 2014’s actual data. Additionally, to keep performance tracking consistent, the original cumulative Safety and Quality targets and actual values have been converted to monthly averages and monthly absolutes respectively.
Power BI software is chosen for the following reasons. 1. Compatibility with Excel Microsoft has developed this agile analytical tool closely connected to other Microsoft related products. The prevalence of Microsoft excel, especially in the logistics industry makes adopting Power BI ideal because it can be easily adapted to the existing raw data from Excel that the company keeps. 2. Affordability The basic version is freely available for download online. Where cost is a concern, Power BI is a top choice with a low acquisition barrier in terms of price-performance trade-off. An upgraded power user version is also affordably priced at $99.99 per user, per month[16].
Existing build-in charts were used with the addition of Bullet Chart custom visualization[17] which offers comparisons of actual and target values in qualitative performance ranges that can be further colour coded for the ease of readability. Qualitative performance ranges may also be flexibility adjusted manually.
Existing build-in charts were used with the addition of Bullet Chart custom visualization[17] which offers comparisons of actual and target values in qualitative performance ranges that can be further colour coded for the ease of readability. Qualitative performance ranges may also be flexibility adjusted manually.
A visualization dashboard should closely relate to the taxonomy of data as suggested by Shneiderman, B[17]. This allows users to have an overview of the data and filter accordingly to get details on demand. At the topmost level, the dashboard gives users a macro view of company performance over the past year using simple and clear visualization of key figures in 3 summary tabs which is elaborated in detail in Section 6.3. Subsequent tabs then provide graphical representations of data zoomed at a micro-level for users to conduct focused analysis on various business functions such as employee training and target performance indicators. By arranging dashboard pages from the company overview to individual business areas, the dashboard emulates a presentation style to facilitate user interaction with the dashboard. Fig 4. Snapshot of drill down feature The dashboard also includes drill down features in certain graphs to add depth in the information provided by the same graph. With this, users can drill into a certain time period or category in hierarchy data types. By having a user interface design catered to the needs of the users, the dashboard facilitates the logical thought processing essential in conducting cross department analysis and investigative work when a pattern of trend is observed.
Section III
Strategic and analytical dashboards are critical in helping logistics companies make key decisions at a management level. Companies on a medium to multinational scale often face challenges in scaling the business while monitoring various business function such as human resources. Furthermore, logistics companies are exposed to varying quality standards and targets that may be specific to individual customers.
Visualization dashboards should not be underestimated in this field. Data examined in this case study is from a local branch of a international logistic processing company that has 2 business units. Hence, the following case study aims to present the application of the proposed dashboard in a real life scenario and the effectiveness of it in conducting close monitoring of the performance of its business units, examining employee retentions and evaluating effectiveness of training.
[Display 1 - overview] [Display 2 - Employee] [Display 3 - Training] [Display 4 - Performance] [Display 5 - Performance in detail]
Employee Dashboard Implementation
Employee Status Breakdown Given the join and leave date of employees, we are able to segment the employees according to 2 variables - Existing or New employees and Stay or Left.
Fig 6. Mosaic Plot for Employee Status
We use a mosaic plot to compare the proportion of employees in these 4 different components. While there is evidently a small percentage of new hires, we can observe that the company has an employee turnover rate close to 50% for both existing and new employees.
Yearly count of employees who left and joined
Fig 7. Line graph of the employees who left
The graph depicting those who left from 2011 to 2016 shows a declining trend in the number of employees leaving from 2012 to 2013. Dropdown lines also serve as visual cues to guide users in making accurate interpretations. Drilldown features is included for users to zoom into the months and days of the week in which employees left within each to examine seasonality patterns at a micro-level.
Employee Training Dashboard Implementation
Training Placement As the sponsor company records each attendance of an employee to a training session as training placement, identical training course names can be provided in multiple sessions, especially regular orientations conducted for new employees. Thus, a new unique identifier column Training Placement was created in a format of “Course Name__StartDate__EndDate”.
Training Course Grouping The training courses are categorized into different groups according to their purpose. For example, the courses designed to educate employees with safety standards and environmental compliances are grouped into Course Type HSEQ (Health, Safety, Environment & Quality), and those to provide operational guidance are under specific operation group such as Packaging and Warehouse. HSEQ course type showed the highest number of training placement since it consists of basic and essential courses usually undergone by new employees.
Training Expenditure and Budget Each training placement incurs certain amount of course fee depending on whether it was provided internally or externally.
Fig 9. Snapshot of Course fee plotted against course budget
Total course fee for the current and previous year was plotted as a simplified bullet chart, based on the current system date. Using a marker, total course expenditure to date can be compared to planned budget.
Distribution of Training Hours by Employee
Fig 10. Snapshot of Distribution of training hours by employee Histogram
A histogram was used to provide with an overview of training distribution, which shows the number of hours that the employees spent for their training. Total training hours in the current year (Sum of Hours per employee) was calculated for each employee. Using bins of 20 hours, the height of each bar represented the frequency of employees falling into each bin of training hours. As such, given the maximum total hours per employee is 600, we can observe that a majority of employees undergo a total of 1 to 30 hours of training.
Given the limitations of Power BI, a histogram was utilized instead of a density plot to describe this distribution of training hours.
Target Performance
References
[1]
