Email
sinha.rk.dr@gmail.com
Does AI/ML help in reducing persistent noise in consumer demand?
A common question I have been asked in multiple meetings and conferences is… If the consumer demand is inherently so noisy, should we even attempt to deploy AI/ML in Demand Sensing?
Let me go back to the basics. What is noise? It is nothing but a signal yet to be deciphered. Think of how security agencies crack secret messages wrapped in encapsulation. The stream of bytes looks random and noisy to us, but it does contain a signal which needs to be deciphered.
It’s the same with consumer demand. What looks like a lot of noise actually contains various useful signals which we haven’t understood yet. Our attempt over years has been to use more and more sophisticated time series models to understand the underlying demand pattern. However, these have been deployed with limited number of independent variables through a multivariate regression model.
The remaining noise can possibly be deciphered through techniques like Neural Networks which look for hidden patterns invisible to the conventional regression models. That’s where AI/ML is really useful.
If we use ML algorithms to just automate the conventional regression models in a more efficient manner, we are leaving a lot on the table. Think about it…