How much Analytics is too much Analytics?

Being one of the biggest ecommerce firms in India, we have to constantly innovate and improvise so that we do not disappoint our 15Mn+ customer base.

Having world class Product and Engineering teams definitely helps, but there are times when you have to look outside for inspiration and ideas. We keep reading articles on who is doing what, across the globe. We also keep meeting companies that offer to enhance our capabilities by some product or algorithm they have.

Most of these products that we come across have been cooked with 1 portion of technology, 2 portions of need and some 20 portions of unnecessary data collection and modeling. There are companies that claim to tell you the intention of a customer's trip on the basis of direction of his/her car movement; some claim to tell you the choice of content of customers by tapping into the digital signals floating around them; then there are some who claim to improvise your recommendation of clothes basis the songs stored on the phone; the list is long, but you get the drift.

Little off track - there was a Bollywood movie, not so much in past. The central theme of the movie were two friends who have a sort of understanding. One of the characters would have a dream, and then would narrate it to the other one. The latter one then would analyse, put in some correlation models and come up with a number. The two would then bet all their money on the number in lottery - and bam! The money is doubled.

This was the movie I would think of, sitting in these meetings.

While I have no doubt in my mind, that any and all of these products would have required a genius mind and a dedicated set of resources - I am not Impressed.

These data collation processes, algorithms, modeling, correlations and regressions are all theoretically brilliant - but they miss the point that a human is a mix of choices. 

Humans are not machines, which in most cases run on the same processor, OS, RAM. Humans are Humans, where the processor changes with mood, OS changes with company and RAM changes with the brand of alcohol he/she is sipping on ;)

The recommendations in the artistic areas (esp Music and Videos) do not seem to acknowledge that a person who watches House of Cards, can be a fan of Justin Bieber and also have Mad Max as the favorite movie.

May be some day we shall see an algorithm that is able to assess the emotion of the other person "while" having a conversation, and quickly recommend next words to impress. May be there is one already.