Fall is the most common time to start school. When I was still in college, my main intention was always just to pass the course without having an intention to make use of the time, people and resources available. Having a guided project to work on is a good start.
However, I wish I had imagined how any particular project would have fitted into my portfolio, so that I could save time by just putting ready-to-go products together. In this article, I want to share a structure that I have stuck with for a long time and wished someone have told me whenever I approach any projects.
In this guide, there are a lot of things I have experimented with, so they are all available on my Github repos. If it is still unclear, the way I usually do the initial research of how people do things is through reading Medium articles. Let's dive right in!
Understand what I want to showcase
A big drive for any of my projects back then was the technical analysis I wanted to perform, due to the lack of my own working experience. I realized that even though I knew how to write codes, I was not exposed to different types of business requirements and techniques to solve them. In order to find out the missing link, I read a lot of job descriptions and followed working professionals. You can refer back to this blog post to find out how I gathered information.
Back then, I was interested in user's behaviors, which covered a lot of online businesses. Therefore, I collected a list of possible analysis I could do for users’ segmentation, for example: funnel analysis, RFM analysis, cohort analysis, survival analysis, etc. I also looked up some KPIs, metrics such as Customer lifetime value (CLV), average purchase value (APV/AOV), retention rate to incorporate into my findings language.
Most importantly is the one question: So .... what?
In order to figure out this important question, there are other questions to answer. For example, I would imagine if I worked for a company, what type of questions would they ask for me to choose this analysis technique over the others. I would also have a basic idea of what data points required for me to proceed, so that I could hunt down my perfect dataset.
Find or create the datasets to practice the analysis
There are many good datasets available online. Kaggle, Amazon, Google, Wikipedia are some really popular sources with variety of topics and even some inspirations of what other people have done with them. When choosing a dataset, besides looking for the one that fits the analysis, I also checked for the quality of the dataset, in order to determine the data preprocessing requirements, based on what I want to showcase (Examples: dealing with big data, cleaning data, combining different data sources, etc.).
In some cases, if I know exactly how the data would look like and I want to hurriedly start the project, I can even create the dataset using this website, called Mockaroo.
Another way of getting the most updated data is to make use of APIs or to create a web scrapping program in order to attain the desired data. This approach is suitable for bigger projects, in which retrieving the data is an important component that illustrates our additional assets with Python. This can be as simple or as complex as one wishes, but it is always a good idea to check for APIs first, in order to avoid reinventing the wheel.
You can check for a project using Coingecko API here. Another example of working with a Python Web Scraping program can be found here.
Structure the project:
There are many ways to tackle the same problems. However, it is important to follow a specific structure with labeled steps, and explain the findings clearly with actionable plans. I find having an executive summary at the beginning and the end ensures a clear expectation for users to follow.
In order to communicate with end users about the most important insights, having proper visualizations to illustrate is the key. This is the place to showcase the inner designer in me, so I spend time to tweak small details that create a more pleasing and interesting piece for the audience. This can include matching colors with the brands, or adding animations to make things more lively, or simply tidying up the axis and the scales.
This tremendous effort, unfortunately and realistically, is not often viewed by the recruiters. Nevertheless, it trains the brain to follow the patterns, so that it is easier to follow and review when needed.
I have been asked to explain the most complicated project, a project that transform raw data into insights, and many take-home assignment that I could quickly start without being panicking because I have had a solid structure to follow.
I have also adopted the top down communication concept, which is similar to the inverted pyramid structure in Journalism. We always want to put the most important piece of finding at the top, in order to save users’ time and effort, but also to make sure that they are completely delivered. There are many good online resources to learn about this, but this video is the one I found very helpful to refer to: TALK LIKE A CONSULTANT - Top down communication explained (management consulting skills)
In many school projects, I was asked to include either a report of a presentation to describe the process and the findings. It is a great place to for putting all the pieces together and emphasizing the motivations, the challenges, and the future considerations to further improve the projects. This can, and should, be done on the Readme document on Github but in a more concise manner, which can be seen in this example here.
Projects are one of the most crucial piece that actually filled up my non-experienced resume and helped me hit multiple keywords in the job descriptions. I hope this guide is useful by adding small polishes to your brilliant projects. Have fun projecting!
Comments