I am Ruchika Menon, and I started my career in an AIML startup as a Business Analyst. So, I had only theoretical exposure to AI and ML. In my current job, when I was staffed on a project that dealt heavily with data science, I decided to enroll with Great Learning to develop my technical knowledge.
I was a Marketing consultant on a project that focused on enhancing the marketing scope of a client. One of the new ideas that we brought in was the angle of Machine learning in Marketing. The use cases we touched upon were very specific to the client itself to augment and optimize their marketing efforts.
As mentioned above, we wanted to optimize the marketing avenues of our client. Our project dealt heavily with personalization, so we had a comprehensive idea of the needs of our customers. We were able to tie back his/her behavior with their demands and identify the next best course of action. Generally, we would send the customer multiple communications to address all their needs, but we wanted to try introducing an ML model to prioritize the customer only in one campaign pertaining to his most vital need.
Through extensive data availability and personalization, we could identify the varied needs of a customer. But we wanted to be able to identify a singular most important need so that we can optimize the communication to make every contact count.
We would send multiple campaigns to the customer, informing him of a variety of products that he was eligible for and has shown some affinity towards. However, not necessarily all would convert. By using past and latest data, we were able to incorporate seasonal and latest buying trends to identify which campaign he would most likely convert in.
It was resulting in over communication with the customers and wastage of resources as well. We were well versed in python and SQL; those were primarily the tools used to solve the problem. We used these languages to do the data munging and model building of the regression model.
We saw that the month of the year and the latest purchases (both volume and price) played a significant role in the conversion of a campaign. Most importantly, we were able to identify a few customers that, regardless of the campaign sent to them, would not buy in that month. These customers were pushed to the bottom of the campaign to not waste resources on them. They were only contacted if the top 80% of the customers were already contacted.
We decided to use a regression model using past and latest behavioral data to identify the pattern the customer exhibits before he converts in a campaign. We created multiple regression models in python and our campaign management tool to find the propensity of the customer to convert in the campaign. We accordingly slotted the customer in the campaign where he had the highest propensity and revenue potential.
We were able to get an uplift of almost 40% in reachability and conversion from before due to the implementation of the model.
As a marketeer, this helped me optimize my communication and ensure that the resources at my disposal were utilized efficiently.
Great Learning overall gave me a very clear picture of the E2E process required to build a model. My concepts when building the model were solid, ensuring that the model had a high confidence level and, in a way, also giving me the option of building multiple models with different variables. The knowledge that I received truly helped me enhance the marketing program I was on.
