Amod Malviya | Technology is not just an automation tool
4 min read 31 Jul 2013, 10:06 PM ISTFlipkart’s Amod Malviya on why the company chose to develop its own business software
Premium(Aniruddha Chowdhury/Mint)
Bangalore: Amod Malviya, senior vice-president and head of engineering at Flipkart.com, is the man behind the online retailer’s innovations aimed at helping the company differentiate itself from rivals. In an interview, the Indian Institute of Technology, Kharagpur, graduate explained why Flipkart chose to develop its own business software instead of buying what was available in the market. Edited excerpts:
There are so many off-the-shelf software packages available, but Flipkart still runs its own enterprise resource planning (ERP) software.
When we were evaluating a whole bunch of ERP solutions we were looking at, if these solutions would scale if we were to scale business 50 or 100 times—will the same solution hold? This was during early 2010. Second aspect was interoperability—how well it could integrate with our existing systems, and finally we looked at whether those solutions were pure automation products or they offered any level of intelligence. On all of these parameters, not a single product available satisfied us. Many failed just on scale.
Do you always look for home-grown solutions?
Lot of companies think about technology as an automation mechanism. We have a fundamentally different approach to that—we don’t look at technology just as an automation tool because then we would be missing out on a lot of other advantages in scaling intelligent systems. We don’t like to build solutions to address a specific solution, we like to build systems that think, decide and are decision-making systems.
Can you elaborate?
For instance, if you look at the problem of pricing, you are dealing with products that have lesser margins, but attract lots of customers and another bunch of products that give us good margins. That’s basically a play between margins and topline. Most solutions would like you to decide the logic, the route. But we like to think about pricing systems in a way where after giving it the problem statement, it arrives at a decision itself, suggests different solutions. We tell our pricing systems, this is our topline, this is the bottomline I want and let the system suggest possible paths. Popularity of a product, supply chain costs associated with that product line are among data points that are fed into the system for it to arrive at a solution.
Similarly for supply chain systems, we realized that the solutions available in the market focused just on automating it. While enough benefits can come from that, we believe that at some point in time it could even become a constraint for the business. Our payment system also evolved from a home-grown platform. For now, we are not patenting these solutions, but at some point in time we will have to look at ways for protecting our IPs (intellectual properties).
Is there a big data challenge you face?
In terms of the volume of data we generate, it’s massive—700-800 gigabytes per day. That was about few months ago actually. Now, it would easily be one terabyte per day. Big data is one area where we actually work with some proprietary solution providers. In fact, a lot of decisions at Flipkart are now automated and driven by data.
We are planning to build more platforms for information discovery and management on top of existing big data products. The available products are only focused on how do you crunch the massive chunks of data, we are focusing more on creating a single data view on a single platform. Going forward, we will look at giving the power of big data to everybody in the organization, of course depending on appropriate access rights, etc. We will democratize the data.
There is a lot of buzz and myth around big data. The key is to invest in the right data scientists and have people who can make sense of insights beyond just crunching the data. I am hesitating in talking about big data because it’s such a heavily marketed term that a lot of people end up making mistakes by attempting to buy available products. What we are calling it is data-driven decision-making, which goes beyond just crunching structured and unstructured data.
