During the summer after my freshman year, I worked as a Business Data Analyst Intern for Dun & Bradstreet (D&B) in Center Valley, Pennsylvania. I spent the bulk of my internship using R Programming and Excel to analyze large amounts of data from D&B’s SQL servers and increase automation as much as possible. I also worked to create a new approval workflow using SharePoint and Microsoft Flow in order to automate approval changes impacting the division.  Finally, I worked on many ad hoc assignments as well, such as web-scraping 54 potential new data sources and creating summary presentations from data.

I chose to do this internship largely to gain exposure working in a corporate environment and continuing to build on my analytical skills. I felt that Dun & Bradstreet was an innovative company where i could learn much more from the other data analyst professionals in my group. I would also be able to apply knowledge from my previous statistics/mathematics courses to the real world.

Throughout my internship, I came up with several innovative solutions impacting the company. For one project, I used R to analyze different variables in D&B’s database to find high correlations with linkage. After my analysis, I recommended 7,000 to be encoded as “linked” in the system, which can easily be repeated with future data. This improved automated process will save D&B employees time from manually checking linkage. Another project involved creating a workflow. Previously, when an employee wanted to make a change to existing systems, they would have to email several people individually for approval, who would then email others for approval as well, often taking weeks before final approval. I  simplified the process by using SharePoint and Microsoft Flow to create an automated workflow. Now, an employee simply has to fill out an online form with the details of their change. The approval form will automatically be sent to their managers and others involved in the process, saving time for employees. Upon testing the workflow, the average time until final approval was 2 days, a significant reduction.

While most projects went well, my group and I had difficulties finishing our final intern project, which involved automating D&B’s Match system. We initially developed python code to search D&B’s databases, but had to input a series of search queries. Since we had hundreds to go through, the question became to what extent we should do it. One person only wanted to find the bare minimum of queries (matching 75% of the data) so that she could finish other projects, while another person wanted to match 99% of the data, which would take too much time. It was also becoming a toxic environment as the person who wanted 75% match was removing our queries and arguing with the rest of the team. After staying calm and having several meetings to discuss, we were able to refocus on the project in a unified direction. Unfortunately, we lost too much time and were only able to match ~82% of the data, which was above our 75% minimum but short of our 95% goal. It was an important learning experience for me as I worked with people who had opposite viewpoints. Being able to unify the team was an important leadership lesson for when I work on other teams in the future. I also understood the importance of not wasting any time in these large but time-sensitive projects.

The most important takeaway from my internship was learning to balance many different projects and work on numerous teams. As I had many projects to work on everyday, I had to prioritize my time efficiently and focus on the most important tasks on hand. I also worked with a wide variety of interns, employees, and managers, and adapted to their work-styles. Being in a corporate environment was also a great experience. Finally, I was able to continue working on my Excel/R analytic skills and build PowerPoint presentations.

Here’s an artifact showing my conclusions when analyzing the relationship between Duns Number and Linkage:

I used R to analyze the data and converted my output into Excel. If the proportion of Linkage within the Duns Number was above 80%, I would recommend all candidates with that Duns Number to be built as linked, as shown. After receiving this output, I encoded these new linkage rules back into the system and explored the relationships between other variables and linkage. This would save employees time as they will not have to manually check for linkage in any of these candidates.