This post has been republished via RSS; it originally appeared at: Microsoft Tech Community - Latest Blogs - .
Everyone is familiar with the movie trope where a rag-tag group of people with unique skills and backgrounds are thrown together to overcome a certain challenge, and succeed. That has been somewhat our experience with Radboud University’s DARBS (Data-driven Analytics for Responsible Business Solutions) course from beginning to end. Although being business administration students, our backgrounds diverged widely. Our backgrounds ranged from entrepreneurship, to organizational design, to system dynamics. Some of us had more work experience, some less. Add to this different cultural backgrounds (think of Vietnam, Romania, Germany and the Netherlands). However, we were similar in our knowledge of data analytics. Most of us had no previous experience with Power BI or the likes (except for superficial knowledge of Excel). We want to share our story with you, in which we will showcase how a hands-on approach to data-analytics software could be introduced in MBA programs.
But before we jump into the details of this course and final project, we would like to introduce our team and our motivations for choosing the DARBS course exploring the possibilities of Power BI and data analytics.
Our Team
Product Owner:@DaanvBrenk
I am finishing my Master’s degree in Organisational Design and Development. Maybe because of my background in the Humanities, I started noticing that more and more people in my environment are dissatisfied with their work nowadays. I think this could be improved, which is why my goal is to create higher quality jobs for people. If I look back on my experience with the DARBS-course, I think the use of Power BI could be complementary to this goal. It could be a powerful wake-up call for people when presented with facts on their way of working, rather than telling them directly. Aside from that, I liked that the course had a hands-on approach to teaching Power BI, which also motivated me to enroll in the course. https://www.linkedin.com/in/daan-van-brenk/
Data Analyst: @Julian_Vriens
I am currently busy finishing my master’s degree in Strategic Management. Having already had experience and classes in data analytics/management, such as with Excel, Power BI, SPSS and Microsoft Access, I knew that following a course to deepen my Power BI skills and learn how to analyze data ethically would be a good fit to further expand my skill set. The DARBS course has allowed me to reach these goals, and I look forward to using these new skills in a professional environment to analyze data ethically. https://www.linkedin.com/in/julian-vriens-810785181/
Data Analyst: @CristianCrk
I am studying Innovation and Entrepreneurship. I have always been interested in having knowledge and working with data, and for this reason I’ve enrolled in the DARBS course. It taught me a big deal about what is possible using Power BI, data collection and data security. I am also passionate about humanity and I believe that with responsible use, data can benefit society by offering knowledge about societal tendencies and human needs, along with ways to reach them. https://www.linkedin.com/in/cristian-ceausescu/
Data Visualization Consultant: @JanicRa
I am in the process of completing my Master’s degree in Organisational Design and Development. Throughout my studies, I have become interested in many topics such as technological development. Therefore choosing an elective exploring Data Analytics with Power BI was certainly the right choice as it not only introduced a great tool for analyzing data and displaying it in a digestible way but also presented the possibility of using AI to predict and think about what data analytics will contribute to society in the future. https://www.linkedin.com/in/janic-rabe/
Data Visualization Consultant: @Thanhnguyen96
I am doing the Master's Business Analysis and modeling. I have over five years of experience in brand management, working for leading consumer goods companies, such as Reckitt and Nestle. I find it important to use the power of data-driven analytics in decision-making, which led me to the DARBS course. The skills I gained out of this project, I could apply in my future job to get consumer insights, and to facilitate insightfully groundbreaking marketing campaigns. https://www.linkedin.com/in/cong-thanh-nguyen/
Project Overview
The goal of the project was to deliver a final report and video blog for a fictitious company called VenturaGear. The goal was to allow VenturaGear to stay competitive through, as you might have guessed, data-driven business solutions. We would develop our skills in and knowledge of Power BI in the process. For the report, we were asked to provide managerial recommendations to Venturagear. Additionally, our findings had to be backed up by academic sources, and we were also required to reflect on ethical considerations regarding data-use. We had roughly a month to complete this task and upload a video blog on our findings. While the goal was not necessarily to make us into data experts, the course provided us with a strong basis for further skill development. In other words, the course allowed students to work hands-on with Power BI software in seminars. The learning process was also facilitated by weekly assignments that were not necessarily linked to the case, but provided stepping stones in the form of Microsoft modules to engage with the software on a deeper level.
Project Journey
The course had us focused on getting some basic skills with Power BI for the first half of the course duration. After that, all Teams received a “SnapShot” of VenturaGear’s database. Within this database, we were free to explore data between 2011-2014. The overarching goal of our project was to improve the customer experience for VenturaGear while considering and presenting ethical considerations concerning data analytics.
With our newly acquired knowledge on Power BI basics, we first explored the provided data by feeding “interesting” data sheets into PowerQuery and merging some, while forming new relationships between the datasets. Through this rough first exploration, we also discovered data that would not be useful for our sales-focus approach to the project. Additionally, we also came across data that to our understanding could be considered ethically questionable, such as customer skin color.
The next step had us further narrow down the data until we effectively had one Excel file. By compiling all important data into one file, we were able to build dashboards that allow us to show the lack of special offers in the B2C sector. Furthermore, this Excel file would be the basis for our machine learning in which we tried to predict the amount of sales from special offers versus sales without special offers.
The final step was to combine all found insights into a report and present everything in a video blog, which you can watch on this page.
Project Result
We looked at the year 2013 to spot a pattern in monthly sales and found out that there are dips in the months of April, August, and November which go below the average sales. Monthly sales are also lower at the start of the year compared to the rest of the year.
Graph 1: Sales 2013 (Revenue and Quantity)
We have also looked at the special offers and found that the majority is B2B and directed towards retailers, in the form of a volume discount. A new product offer has been implemented for only two models of Touring bikes in May, June and July and a seasonal discount has been implemented for the Sport Helmet in May. As you can see in Graph 2.
Graph 2: Special Offers 2013
We have also found a difference in the products bought throughout different regions. In Europe the order in terms of quantity for the first 3 categories of products is Jerseys, Touring Bikes, Road Bikes. In North America it is Road Bikes, Mountain Bikes, Jerseys while in the Pacific region it is Tires and Tubes, Touring Bikes and Road Bikes. This shows a difference between the customer preference across regions.
Graph 3: Top products per region
By running a machine learning model on the sales data, we tried to understand if the discount of a product and its price determine whether customers buy more of this product. We have found that there is an effect and that products up to $496, and between $978 and $1466 sell better than the rest, and a discount of 8 -12% is also correlated to higher sales. The model displayed an accuracy of 62%, which shows we are not very confident in our prediction, but it signals potential for further investigation and gives an idea for additional special offers.
Graph 4: Machine Learning Discount and Price
Unfortunately, within the VenturaGear data set, we found almost no data on possible B2C discounts which could help in predicting the Sum of OrderQty. In the future maybe after VenturaGear has explored the impact of B2C discounts and collected some data on it, a proper machine learning model can be implemented.
Table 1: Current Discounts
Regarding ethics, we have found the following table in the data, which contains details about individual customers. Here, information about their ethnicity and skin color was gathered, along with income level. We decided not to use this data further in our analysis, as we wanted to avoid any possible ethical dilemmas. Including this data in our machine learning, could have for example predicted that we should focus more on people with light skin color and less on people with dark skin color, obviously discriminating against one group over the other. Another reason why we did not include this information is because data analysts have an obligation to make sure that data is handled responsibly and collected with consent. As we do not know whether VenturaGear asked their customers if it was okay to include such data, our general rule was to leave questionable variables out of analyses.
Table 2: Sensitive Data
Future Development
Further ideas and improvements for our project are as follows. A machine learning model would be required so that the price sensitivity of each product can be seen. In this way the special offers can be applied to certain products in the right proportion. Improving the accuracy of our developed machine learning model predicting orders based on discount and price should be improved in order to display a higher accuracy. If no improvement can be made, investigations of whether the current one can be accepted are necessary. A new measurement of return rate per product is useful to learn about customers, as it can show if they are dissatisfied. Although returns are normal and can have a reason of incompatibility which wasn’t known by buying online, there can be underlying motives that say something about the product itself.
A further analysis into the top 10 product categories bought in different regions can offer further insight on the type of customer demand there is and help determine special offers. A step further can be analyzing the countries and the demand for each country, for specific targeted offers.
Having teams focus on each of the three regions and creating a filtered dataset for each can be helpful to seeing a more detailed picture, and work on a targeted performance solution.
Our next step is studying and obtaining the PL-300 certification and continue learning about data and A.I., also obtaining other certifications. The idea of working within the data analytics community after graduation has certainly been kindled throughout the duration of this course.
Lessons Learned
- Data analytics
Before the start of this course, data analytics seemed intimidating to some of us. We were thinking of huge Excel sheets that required an advanced understanding of data analytics to get to a result. Through Power BI, some of us discovered that understanding data analytics was actually manageable and even fun at times through the ability to create understandable and intuitive dashboards from Excel sheets transformed and cleaned in Power Query. For those of us with some experience, it allowed us to gain a deeper understanding of the software and try out additional/ new things.
- Teamwork
Due to the scope and general freedom of the project we had to work closely as a team not only to achieve multiple perspectives on data. It also required us to set boundaries for the project as there was rather limited time to complete it. This forced us to really develop as a team and complement each other with our personal insights and skills in order to help each other successfully complete the project. In this light, we also discovered a very helpful online community, from Microsoft blogs to YouTube explaining for example how to design certain graphics through the use of DAX and providing general tips for better designs, etc.
- Ethical considerations
Lastly, one of the big lessons we learned is how to analyze and manage data ethically. For example, by only collecting the data that is absolutely necessary (data minimization) and implementing measures to prevent the data from getting leaked/lost (data loss protection). This course and project allowed us to understand that ethical data analytics goes much deeper than we previously expected, and that unexpected harm can be done to people if data is not handled ethically.
Teamwork
The final project of DARBS required that each member take one out of three roles and contribute to the project accordingly. The 3 roles were: Product Owner, Data Analyst, and Data Visualisation Specialist. Considering the scope of this project and our limited experience with Power BI from the beginning we were aware that this would have to be a rather extensive teamwork effort in order to produce a satisfying result. Daan (product owner) made sure to not let the scope of the project get out of hand while always considering ethical implications that we will have to be aware of. The data analysts Cris and Julian had the main task of siphoning through the provided data set and determining, for example, possible problems that might hinder better sales. The data visualization specialists Thanh and Janic ensured that the findings of the data analysts could and would be presented in easily understandable dashboards in addition to creating a video blog. One drawback to the set-up of the course is that even though specific roles were assigned to us, the fact that we all had very limited experience with Power BI meant that we were all learning along the way and discovering new ways of analyzing data throughout the project. This led to our team members supporting each other and taking on multiple roles. In turn, this meant that the roles were mostly fluent throughout our project.
Overall, we believe that teamwork was absolutely essential in completing the project as all of our diverse expertise and perspectives helped us discover interesting aspects of data and ethics that ultimately allowed us to deliver on our task.
Conclusion
One aspect we hope you will remember about our project is that it was a matter of trial and error. If you are to learn about something practical such as developing a mastery of Power BI, get your hands dirty and start experimenting with data. Because this process can be very demanding on the individual, MBA programs could decide to incorporate this learning in a collaborative project. Especially during the early stages of learning, team members are dependent on each other to fill in the gaps of their knowledge, which in our case, provided a sense of ‘we are in this together’. Another aspect to our project is that ethical considerations are inevitably intertwined with data analytics. We tend to separate ‘hard factual data’ from the people who produced it, or whose data is being collected. The impact of this project could be that it allows students (as future business analysts) to think about data while also considering the implications of their use for the outside world and its impact on society.
We believe that, in order for data analytics to have the best benefits for society, certain norms and standards are required. Ethical predicaments can have grave consequences for individuals if rules of privacy are violated or data is wrongly interpreted due to a lack of context. Therefore we highly recommend Power BI users to make use of modules such as Intro to data classification and protection, Fundamentals of Responsible Generative AI and Train a model and debug it with Responsible AI dashboard. While the use of big data and its analysis is an ongoing journey where new ethical pitfalls are constantly being discovered, an obligation for responsible data usage and collection is an absolute necessity. These modules help to understand and build a base of knowledge concerning data protection, transparency, and fairness, which should be at the basis of every data analysis.
Additionally, our team believes that machine learning, if done in a responsible way as introduced throughout the course, has great potential for society. It helps with making better decisions in terms of, for example, customer satisfaction. We recommend the modules Introduction to machine learning and Confusion matrix and data imbalances, as they are a concrete starting point for developing a machine learning model for your own data.
On behalf of Group 3, Thank you for reading our summary and reach out for any questions or recommendations!