The e-commerce companies on Internet take advantage of the algorithms to become more and more customer friendly. These algorithms predict and influence human behaviour. As soon as the home page of an e-commerce site is opened, we get to see the offerings of those products in which we are interested. As users of internet, we leave behind a digital trail and our data is captured. Algorithms pick up our digital footprints, and synthesize them to give us customarised choices. Based on this data, the algorithm decides the right products for us which we are likely to buy. A significantly large percentage of sales is governed by algorithms, say 35-40 per cent.
Algorithms do big data analytics in real time. On Facebook algorithms decide what news the users will get in the Newsfeed. The Newsfeed algorithm is called EdgeRank. It identifies what users enjoy most. The results are different in different geographies. Indians are interested in ABCDs or astrology, Bollywood, cricket and divinity. Basically, on Facebook you see more of what you like and less of what you hide.
Advertising is targeted based on the user data. Even the comments written on updates signal your choices. The sharing too indicate your likes and dislikes. These signals are the proxies for relevance. Search habits of the users also provide a lot of valuable information.Surfing habits also provide valuable inputs.
Algorithms make marketing specific rather than randomized. Algorithms also decide the delivery of the products through a set of couriers. That courier is chosen which best suits the customer needs Package routing is done through algorithms.
One such open source programme to process big data is Hadoop developed by a Yahoo! engineers, in 2005. There are other proprietary softwares.
LinkedIn matches job openings with candidates. Credit Card companies use algorithms to detect fraud. Since they know the user’s spending pattern, any major deviation is detected.
Algorithms do require scale to work better. They analyse big data. The more the data, the better the result. At lower levels, the accuracy is lower. At higher levels, the accuracy is to the extent of 98 per cent.
Algorithms are not very useful in all situations, e.g. new product launches, organising a flash sale, and analysing the first time visitors to a site. Here there is paucity of past data. Moreover, everything cannot be predicted. Machines process large volume of data but do not understand the natural language. Ultimately it is a combination of man and machine. Machines lack the human touch. They cannot replace the human judgement. Some companies, therefore, blend man-machine approaches. But these days, algorithms also do self-learning. A one of search or comment on social media may give a false positive. Algorithms could be cruel too, say a year’s review in Facebook may incorporote the death of your child.
Companies keep on buying the algorithms. LinkedIn bought Pulse in 2013, and Bizo in 2014. Google took over Deep Mind, an AI start up, in 2014. Snapdeal acquired Wishpicker. Flipkart invested in Giviter, a gift recommendation platform. Microsoft proposes to roll out Cortana, a digital assistant.