COVID-19 and Data Science
No, this is not a regular post of having incomprehensible graphs or “Flattening the curve” showing the impact of Covid-19 on countries or humanity at large. This is more about what we as a “Community of Data Scientists” should learn from such a big event of our lifetime.
Just 4 months back all our forecasting models were running smoothly with nearly impossible 99% accuracy and were making us proud. Now, how many of these models are running with same accuracy? I guess, very few.
While downward trend is observed in most industries be it Aerospace, Transport or Retail, at the same time Pharma and digital space is thriving the most.
I believe by now everyone knows ‘Zoom’ for all the right or wrong reasons, so to set things in perspective: Zoom has added 2.2 M customers in first 55 days of 2020 while it added only 1.9 M in entire 2019. Could any statistical model have predicted it with even 50% accuracy? No
It must be now clear where we are heading here. In next few lines, I will try to show why external factors are important in any statistical model or analysis for solving any kind of business problem using data and how should we go about it?
What are external factors and why they are important for you as a data scientist?
The economy, politics, competitors, customers, and even the weather are all uncontrollable factors that can influence the models or decisions.
With so many uncontrollable factors in play all we can say is: Change is inevitable and having the flexibility to deal with unforeseen market situations can mean the difference between survival and extinction for an organization.
Do we have a framework to contemplate all such external factors? Yes and No
We do have frameworks like PESTLE or SWOT which can help us in organizing our thoughts when we research about external factors but without any doubt it is next to impossible even for an expert to perfect it. In the end it’s all about reading signals and getting an edge in the chaos.
How to incorporate such factors in our models?
Discussing technicality, feasibility or accessibility of external data is topic of discussion for another day but in today’s world where data is generating at an exponential pace especially from social media platforms, reading such signals is not as difficult as we could have assumed.
To summarize it: Noise is as important as Trend or seasonality.
Stay safe & healthy!
Shiva is a keen follower of scientific trends and is an Asimov fan. Believes solid execution is key to the success of any strategy and is focused on building a world class data science team at Prescience. Has a B.Tech from IIT Delhi and MBA from IIM Lucknow with 20+ years of experience in the technology space.