How Myntra is using data science and ML in its analytics framework
In an exclusive interaction with Express Computer, Ved Antani, Vice President, Engineering, Myntra discusses how the company is solving fashion problems using deep personalisation with rich data platform and data science. Further, he reveals Myntra’s crucial measures for accurate data integration
Please share a brief about Myntra’s data and analytics journey.
Myntra is a differentiator in the Indian fashion e-commerce industry, and the critical distinction is how we develop highly personalised products for customers. Our application experience is tailored for fashion and that needs different way of thinking about user experience, personalisation and data science.
Myntra’s data engineering insights and data science platforms have evolved into one-of-its-kind in the fashion industry with this aim in mind.
Data is part of our DNA, at Myntra. As part of the core e-commerce funnel, data means everything for us. Decision making for business or product definition is dependent heavily on what story the data tells us. Our in-house AB testing and user segmentation systems enable us to test and validate product definition and experiment rapidly to build the best experience for our users.
The same data is also used in enriching our data science.
Please share some of the use cases of data analytics at Myntra?
We recently launched a new function called, Growth Hacking. It is an engineering function within Myntra, where we rapidly experiment with small to medium-sized product ideas on wide-ranging areas. These experiments include, for instance, providing a click-checkout experience, or tweaks in post-order experience to make the user journey more meaningful. The aim of Growth Hacking is to either improve conversion and retention rates, or some business goals.
In order to release experiments in Growth Hacking, we conduct A-B testing for new features and release it only to fixed a set of people. This is to ensure that we also measure the success of this feature. Myntra releases experiments continuously, and measures the impact. Also, these are the fast-moving pipeline of experiments that we can push out to users.
What is the analytics roadmap that you’re setting up for the organisation? Is there a specific approach that you’re looking at?
Myntra is betting high on making data science and machine learning prominent across the entire buying and push orders process. Gradually, the core analytics systems will become even closer to data science and ML platform. Myntra is investing significantly into building a ML platform and real-time data science pipeline, because we believe that data science within ML models become valuable if they run in real-time. Therefore, Myntra is focusing on building the ML platform for deployment training and data science models’ execution. We are also developing a real-time platform to gain access to real-time model feedback.
At the scale that Myntra operates, we solve complex problems related to fashion e-commerce, using extremely in-depth data and data science. Gradually, we are investing in these platforms, and we continue to add bigger capabilities to offer a delightful customer experience.
What is the kind of infrastructure built for analytics? What steps have you taken for having accurate data?
Myntra has a lot of checks and balances at every step before data ingestion into the warehouse, clean-up and processing. These steps have a clearly laid out telemetry, failure reports, and test risk coverage, in order to ensure that even a small deviation from the expected output is captured. This has been an iterated process, because we continue to improve monitoring and editing framework as we go along. For even for a minute shift or failure in our data pipeline, we have a solid alerting and testing framework to report.
If you have an interesting article / experience / case study to share, please get in touch with us at [email protected]