Mar 10, 2014
Published By : Business Standard
While the consumption of analytics is increasing, Indian companies are way behind in terms of using analytics to drive business decisions, Srikanth Velamakanni tells Ankita Rai
There are more than 200 analytics companies in India and most of them are servicing clients in the developed markets such as the US and Europe with very few Indian clients. Is India an attractive market?
It is true that most of the analytics companies in India prefer clients in the US and Europe. It is not because Indian data is not clean. In fact, the transaction data generated in the country is clean and usable. The challenge is Indian companies don't have the people and the processes to operationalise analytics. Indian companies are yet to catch the analytics wave. While the consumption of analytics is increasing, we are way behind in terms of using analytics to drive business decisions.
When we started Fractal Analytics in 2000, we had many Indian clients. But we realised addressing local market was not a good business idea as Indian clients didn't have the required system in place. Even today, the revenues and pricing are not aligned with the responsibility an analytics company has to take in terms of providing end to end solutions to the client. It is more of a learning experience for Indian companies rather than a strategic partnership. Now, we hardly have any Indian headquarted, India-based client.
In India, companies in the space of financial services risk management, marketing and e-commerce are doing good in analytics. We are currently working in the consumer packaged goods industry, financial, insurance, technology and telecom space. The next phase of growth will come from the life sciences and healthcare space.
What is the difference between conventional analytics and big data. Is big data simply old wine packaged in a new bottle?
Big data is an umbrella catch-all phrase in the analytics industry. While analytics has been around for a long time, the use of analytics has become challenging because of the large data size and the variety of data being generated. For instance, apart from transaction data, we now have audio feed, video feed, social data, and sensor data collected from smart devices and so on. The ways of analysing and storing data has changed. Now the data can be captured and processed in a very short time. So basically big data has changed the earlier way of doing analytics.
Big data attempts to quantify human behaviour. Is it possible to predict human behaviour, with quantitative data?
Consumer behaviour is a function of various parameters, such as, which website she frequents, which restaurant she visits regularly, what are her hobbies and interests. A marketer needs to understand human beings without putting them into boxes or segmenting them on the basis of gender, age group, income bracket etc. If we can understand various factors that influence her, we can communicate with her in the manner she likes. For example, if today I am launching an Android device, I will not talk to somebody who uses an iPhone.
So if you are able to understand the pattern behind the behaviour, you can predict it. We are living in a digital world and generating thousands of data points every day. For example, Google can tell you how much you walked this month, how many places you have travelled. Banks, telecom companies and credit card companies are a big source of data. Data can give you a better understanding of human behaviour, more than what we know about ourselves. It has privacy implications but it is a very powerful tool to understand human behaviour.
With the advent of big data processing technologies, does it make sense to work with samples? How much data is enough to understand a consumer or predict her behaviour?
Consider the analogy of a photograph. How many pixels are enough? What resolution of a photograph is enough? The higher resolution, the better the picture. This is where data analytics comes in. You need a large amount of data to answer specific questions. Earlier, to sample or not to sample was not a question. Computer resources were inadequate to process large data and business intelligence professionals accepted sampling as a kind of pragmatic, albeit inadequate, necessity.
However, if you want to get granular in terms of data you need analytics. Sampling is good for primary market research. Even today the cost of market research is very high. So the sample can be defined according to the level at which a company wants to research. But in the case of data that is coming from the non-research world such as social data coming from Twitter, Facebook, or transaction data coming from banks and telcos, you need big data analytics. You can't analyse it using sampling. Computation and storage of data has become cheap. You don't have to sample if you have computing power.
What are the big challenges facing the analytics industry?
Hiring the right talent in analytics is difficult. We are in a business where talent makes all the difference in terms of what kind of solutions we can provide to clients. The average attrition rate in the analytics industry would be round 25 to 30 per cent. At Fractal, the rate is lower, around 14 to 15 per cent. Our secret sauce is to hire the best talent whom we can trust and who are passionate about analytics. Our hiring process is slow and tough. A mid-level person has to go through 8 to 10 level of interviews. We don't measure input focusing on output. We measure the results and client satisfaction. We want to become an aspirational company with which people want to do business.
Traditionally companies prefer analytics when the market is good. In the times of slowdown, the demand for analytics also declines. When is analytics useful?
Analytics is useful when a company is growing and also when it is stagnant. In general, when a company is expanding and entering into new markets, revenue side analytics becomes important. The focus then is to get the pricing right. However, during a slowdown, companies need analytics to optimise costs. In general, during a slowdown, analytics consumption also goes down. This is because it is discretionary in nature just like marketing spends.