3 Ways to Create Better Data Stories!
Fairly new but highly important to the world of Data Science, Storytelling takes its roots in data science methodologies that ensure that any analysis/project undertaken by a data science team, sees the light of day, not just in terms of having a complicated structure/code, but simply, an action that it brings.
And what better way of compelling an action to be made than a story?
Data Storytelling is yet another way to transcribe data-driven insights by addressing root causes and relating them to BUSINESS VALUE driving growth.
Note: Although the steps mentioned in the article seem fairly simple, the idea for every data science team is to churn out highly accurate and precise numbers as early as possible for stakeholders to consume and act upon.
Based on my experience, the simple steps mentioned in this article are enough to make any data story better.
Data Storytelling takes its place in the data lifecycle owing to the fact that companies are data-driven today and every part of the company has to connect with another and grow as a whole.
Taking the process and citing business impact made by changes/issues being faced, is a very easy way to communicate with the larger audience on what should be done more or less.
What are the key elements of a data story?
On a high level, every data story has the following 4 elements;
- A narrative or a plot
- Loads of Visualisations
- Context
- An outcome/result of the project
Read through this article if you want to understand how to write a data story,
and let’s jump into how you can write a better data story (in 3 simple ways).
Think of it like this,
You are a student who wants to choose a college to apply to, there may be 100 different colleges that you may choose one from to go for, the catch being that… you will do your research. A college might seem better because it has better sports facilities, or another might be better due to a bigger library. What makes the difference is… understanding what you are looking for.
The same thing makes a difference in the data story as well, What are the stakeholders looking for? Or if it is an open-ended analysis or hypothesis you are pursuing, what might be something the stakeholders in the company should know about?
Which brings me to the first thing I’d want to highlight —
Understand the Objective of your Analysis
This will help you keep looking in the right direction of your data, and dig deeper to find more meaningful insights in the analysis. Be open to accepting any surprising insights you find out, and do report them, but don’t dive deeper into other sides of the numbers that may be irrelevant to your analysis, to save time and resources,.
How can you get this done?
- Talk to the stakeholders in your analysis directly to get their perspective on the objective
- Ask questions about caveats to your analysis that you need to avoid or consider
- Study any previous analyses in the same horizon and any competitor’s work in similar areas to have an outside perspective if possible
- Understand more about the underlying business value and areas of impact from your findings.
More about this here:
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Let’s assume you got that right.
You know what you are looking for and are confident about your work. The scope is defined, the numbers look alright, and you have given your audience, a black-box ML model that works on a very strong GPU and your predictions come out as 99% accurate in terms of the test data. You present to your audience, the ROC curve along with the MCC coefficient that you so beautifully have carved out.
Wow. You have failed to understand the intuition behind storytelling.
Keep it Simple!
The simpler your analysis is to understand, the easier it is to remember and take action upon. There are a few basic things that can go wrong with a data story that is over-complicated. Overfitting your data is easy when you over-complicate your model, and the lower your explainability, the higher the uncertainty. A simple analysis that sticks to the point, sends a clear message to the audience, and can be acted upon with haste.
How can you be simple?
- Don’t quote too many numbers, keep it simple and talk about the impact
- Avoid any graphs that are overloaded with information — simple bar charts, line charts should be good. Quick to read, easier to understand.
- Avoid any unnecessary jargon or give precise descriptions if you use any.
- Cut out any unnecessary parts of the story — stick as close to the plot as possible unless you want to explain any external effects to the numbers.
P.S. I found this beautiful Wikipedia page that talks about this idea — Analysis Paralysis, find it here:
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Great, you have both ideas now. You made sure your analysis was simple enough to understand and you stuck to your context. The stakeholders commend your work and clearly, you have enough to get you the next hike :) There’s only one question left for you and the analysis to answer then…
Always ask…. What next?
This is one of the most important questions that every analysis needs to answer. What next?
- If it is a problem you’re addressing — what are the next steps to fix it?
- If it is a new opportunity to take on — what are the next steps to confirm if it is a viable investment to make?
- If it is a recent change to your product or service — what are the next steps to improve it?
- If it is a descriptive analysis — what are the next steps to take to track the KPIs constantly?
- If it is an ML model’s efficiency in the question — what are the next steps to take to improve the model?
There’s a simple motive behind asking “What next?” — Improvement. The more you can improve your analysis, the better the insights you will prepare to leverage in decision-making.
At the end of the day, the analysis you present as a story must be rememberable and should carry a solid message that is understood by your audience. The aforementioned can help you achieve the same, only if you don’t get carried away with your analysis paralysis!
Leave a comment if you think I missed out on any other pointers that are relevant to the article! (Thanks!)
Do leave a clap and check out my previous articles here, if they interest you!
For all my articles:
Connect with me on LinkedIn: https://www.linkedin.com/in/yash-gupta-dss/
~ P.S. All the views mentioned in the article are my sole opinions. I enjoy sharing my perspectives on Data Science. Do contact me on LinkedIn at — Yash Gupta — if you want to discuss all things related to data further!
