How to Create a Data Analytics Project That People Want to Read
8 Actionable Tips to Turn Heads With Your Work
The most important part of building one’s career in a field is to be known by one’s work. When it comes to data analytics, having impressive projects to showcase your knowledge and expertise trumps all other methods, including certifications and courses.
So, how does one build an impressive project? Even more important, what makes an analytics project impressive?
In this article, I present to you 8 tips that have helped me become a Kaggle Notebooks Expert by building narratives across different datasets. So, without further ado let’s get started.
Tip 1: Choose the goal of your analysis over the tools you want to use
“Always remember, your focus determines your reality” — George Lucas
It is easy to get lost amidst the fancy set of data visualization packages that are constantly finding their way into our thoughts. While there is nothing wrong with learning new tools, there has to be some kind of temperance when working on a project.
The end goal of an analytics project is to not flaunt the knowledge of a new tool, but to discover useful patterns within the data provided. Hence, it would be more fruitful to concentrate on asking questions of the data than worry about which tool you must use.
Well of course, there are cases where you might prefer one library over the other due to the requirements of a project. In these cases, it is relevant to work a bit on deciding the library to use. But beware, do not let your choice of library or language guide your analysis!
Tip 2: Have a methodology in place
“If you can’t describe what you are doing as a process, you don’t know what you are doing” — W. Edwards Deming
A methodology is essentially a contextual framework to guide research. In simpler words in the context of an analytics project, it helps you stick to having a process while working on your project.
It is important to have one because having a methodology ensures that you have a clearly-defined pathway to your goal. Also, a methodology becomes very effective when you have to explain your project to another.
This is because you are in complete control of every step you have undertaken starting from the data acquisition phase to the result communication phase; and all other intermediate steps!
A lesser-realized benefit of having a methodology is that it deepens your thinking about the project. For example, I included a step of “Understanding my Bias” in my project where I analyzed US Police Racial Violence. This inclusion helped me ensure that my finds were not affected by my internal bias.
Tip 3: Brainstorm like your life depended on it
“The best way to get good ideas is to get lots of ideas and throw the bad ones away” — Linus Pauling
The first attempt at analyzing a dataset can always be very confusing. Especially if it has a lot of instances and features like the Kaggle Survey Challenge 2020 dataset.
This is where brainstorming becomes important. Simply put, brainstorming is the generation of new ideas by allowing these ideas to freely flow from the mind onto a physical(paper) or digital(computer) location.
Though most definitions of brainstorming call it a group process, there are studies that support the argument that individual brainstorming generates higher quality ideas than group sessions.
In the context of brainstorming individually for your data analytics project, there are 3 useful steps to get you started.
- Read the Dataset Description — What do you think was the main priority of those who collected the data?
- Read the description of features — Which features according to you point best in the direction of the main priority from step 1?
- Read previous work — If anybody has worked in the past with the same or similar kind of data, review them
After this, you will be able to write down all the ideas that you generate onto a piece of paper (or a digital record if you are not old school like me). And yes, brainstorm like your life depended on it. Keep writing down ideas that you can use to analyze your data till you dry out your mind.
Brainstorming is the backbone of your analysis, hence it has to be thoroughly accounted for.
Tip 4: Perform preliminary analysis to identify the most promising narrative
“For we do not think that we know a thing until we are acquainted with its primary conditions or first principles, and have carried our analysis as far as its simplest elements” — Aristotle
After your brainstorming session, it is possible that you have multiple ideas that you want to pursue as the narrative of your analysis. However, you should choose only one main idea if you want your work to be clear and impactful.
In order to make this choice, it will help to quickly fire up your system and write some code to perform preliminary analysis. This could be a part of your Exploratory Data Analysis and hence will require visualizing the data at hand before centering in on the most promising story to tell.
For example, in this year’s 2020 Kaggle ML and DS Survey challenge, I found out through preliminary analysis that the growth of Indian respondents under 21 years has been the fastest when compared to any other set of respondents in the surveys. This helped me build my case for the rest of the analysis.
Tip 5: Use a storyboard to build your narrative
“The storyboard for me is the way to visualize the entire movie in advance” — Martin Scorsese
When it comes to storytelling with data, I personally believe Cole Nussbaumer Knaflic to have one of the greatest minds in the field. While there is an extensive list of ideas she has propagated via her books and many talks, one idea that I find extremely resourceful is the storyboarding process.
As professional writers say, a good story has 5 parts — Exposition, Rising Action, Climax, Falling Action and Denouement. In the context of your data analysis project, you could follow a similar structure to make your storyboard.
Introduce the data you have, move onto the “why” of your analysis or main goal, analyze across the sub-goals you have chosen, report insights while connecting them to your main goal and finally combine all your finds, choose the most important ones and report them as decisions the concerned stakeholders could take.
Tip 6: It’s not about you, it’s about them
“It’s not about you, it’s about them” — Clint Eastwood
Moving onto a more philosophical idea, I urge the reader to understand that any analysis we perform is never for our eyes only. The impact of analytical work is only related to how useful it is for the stakeholders involved.
Therefore, don’t put a graph into your report if it is unnecessary. Just because you put hardwork into it, it does not mean you need to flaunt it. If it’s not coherent with your whole analysis, it must see the inside of the recycle bin.
It is also important to ensure that you are able to tie your results with actionable goals that the stakeholders can take.
Also, ensure that you never mislead your audience with visualizations that don’t make sense.
Tip 7: Get some initial feedback
“Feedback is the breakfast of champions” — Ken Blanchard
The first complete draft of your analysis must always be shared with someone who has no idea about the topic you are telling a story about.
This is effective for the following simple reason — If a person with no background in topic X can understand and appreciate the data-driven story you tell about topic X, then it means that your narrative is coherent, clear and engaging.
If this does not happen, then its time to go back to the drawing board and work through your narrative again. Reiterate till you attain the objective!
Remember that feedback is not meant to make you happy, but improve. So, don’t be hostile if you hear something that you didn’t like!
Tip 8: Pay attention to detail
“The difference between something good and something great is attention to detail” — Charles R. Swindoll
Even the smallest things like the themes of your plots, the typography and colors you use in your report can have a deep impact on how a reader perceives your work.
One of the best examples of analysis where the author has placed great focus on details is the Birdcall Recognition EDA by Andrada Olteanu. Even the visualizations have been made to match the color scheme of the birds!
Consistency is another very important aspect of analytical reports. You don’t need a flamboyant report. You need a minimalistic one that speaks the story of the data that has been analysed.
Other tips that have helped me
The following links include tips from some of the best data analysts and storytellers who have had great positive impacts in my journey so far
- John Miller’s “Some Best Practices for Analytics Reporting”
- Rachael Tatman’s “Six steps to more professional data science code”
- Ben Wellington’s “Making data mean more through storytelling”
- David McCandless’s “The beauty of data visualization”
Hope this was a useful read! Cheers :)
