Interview: Flipkart-Manager Ram Papatla über Mobile und Künstliche Intelligenz.

von Florian Treiß am 26.September 2017 in Interviews, News

Ram Papatla (Flipkart)

Mit künstlicher Intelligenz den Handel stärken: Der indische Amazon-Rivale Flipkart, der im Frühjahr 1,4 Mrd Dollar von eBay, Microsoft und Tencent einsammelte sowie in diesem Zuge eBays Indien-Geschäft übernahm, wappnet sich mit KI-basierten Innovationen dafür, in der Zukunft 1 Mrd Kunden zu bedienen. Ram Papatla, Vice President Product Management bei Flipkart, wird dazu in einer Keynote bei der Mobile in Retail Conference der GS1 Germany am 18./19. Oktober in Berlin nähere Einblicke liefern. Bereits im Vorfeld der Konferenz hatten wir Gelegenheit dazu, mit Ram Papatla sowohl über die Bedeutung von Künstlicher Intelligenz für den Handel zu sprechen als auch, wie wichtig Mobile für die Flipkart-Kunden bereits heute ist.

Location Insider: What is your role at Flipkart?

Ram Papatla: My role at Flipkart is to lead the Growth and Consumer Experiences Products. We have several products and solutions for Consumers to enable shopping more accessible and affordable. Some examples are Personalization, Discovery, Community Reviews, Catalog, etc. As part of the Growth function, we ensure our Advertisers/Brands can reach the right audience to drive consumer awareness and consideration. In addition, I partner very closely with our Data Science group to drive AI-first Research within Flipkart.

Location Insider: Flipkart is India’s leading e-commerce company. How does the Indian market differ from European countries like Germany, France or UK?

Ram Papatla: We are experiencing unprecedented growth on back of several ecosystem changes such as smartphone penetration, Regulatory intervention to reduce friction in payments and transactions and our own ability over the last 10 years to invent and organize several new capabilities. The diversity and the volume of our market is daunting, we serve 100 million customers and that count is rapidly growing, we are pushing deeper in Tier 3 towns of India where access to affordable quality products is a challenge. There are several industry-first concepts that we are inventing to build trust for customers that are traversing to an online experience –  for example: best in class Mobile Website that can work in a 2G environment, offering true value for phones for an exchange with a new phone by grading the customer’s phone real time, providing access to credit based on customer’s history on Flipkart, etc.

Another interesting area is efficient generation of traffic as we compete for our customer’s time – there are several experiments we’ve run to develop engagement drivers such as games, bidding, etc. We’ve also built personalized notifications to provide relevant information to our customers on specific products, offers attached and affordability constructs.

Location Insider: What can the European market (or companies) learn from the Indian market, and maybe vice versa?

Ram Papatla: I think the diversity of India is very unique and we’ve placed a bet that we are going to transform shopping for Indians through technology. Technology is a way for us to offer the best Product Quality and enable Accessibility of commerce for All Indians. There are specific customer expectations such as having a store salesman to accompany and guide them in shopping unlike west where customers prefer exploring the aisle on their own. We are building experiences that are transformative as customers move online. For example: Our Home Page, the content we offer in our Home Page is highly personalized, we can offer specific content and merchandising that matches a life stage of a customer. Another example, is a mechanism for in-app notifications and visual reminders – if a customer forgets to complete a transaction or buy a high value item such as a laptop without a sleeve, we can gently nudge them through an in-app intervention or a widget on our Home Page to add these complementary products.

Location Insider: At the Mobile in Retail Conference, you will talk about „Scaling for the next Billion: Mobile Digital Natives, Driving Standards through AI“. Could you give us a brief outlook on your keynote?

Ram Papatla: Flipkart’s 10-year transformation of 100 Million customers (and counting) using Mobile to access shopping has created an unprecedented data pool of Indian consumer shopping behavior!  This data includes a decade of consumer shopping awareness, consideration and purchase and the most comprehensive Knowledge Graph of products in the country. We’ve built several innovations based on this valuable data from offering Recommendations to Credit Lines. Based on the foundational data, we have started an AI-led innovation which will help us serve our customers better than anyone else in the country. Come hear about the standards that have been put in place (Flipkart’s Vertical Definition Framework) and how we are using AI-first approach to drive transformation of shopping experiences for our customers. The talk will also provide a validation of standards from customers through community feedback tools.

Location Insider: In what other areas is AI helpful for Flipkart’s business and customers?

Ram Papatla: We’ve been focused on several areas within ML and AI for several years. There is core Data Science group that partners very closely with the rest of the organization to provide capabilities to raise the bar on our Customer and Merchant experience. To provide some additional context – Using Deep Learning, we are able to almost fully automate the processes of Product Creation and Quality Control, at a negligible cost, with unprecedented accuracy and practically infinite scalability. We are able to classify images into the correct taxonomical slots, extract important features, identify duplicates and similars, clean up bad images and even determine higher-level aesthetic features that optimize the conversion funnel.

We also have deep learning based image similarity estimator powering our visual recommendation and market research. The image similarity detector can match across diverse backgrounds, presence/absence of human models/mannequins, widely different perspectives. This allows us to recommend lifestyle products that are visually similar to something the user is browsing.

Another area is NLP – The ability to let our customers express themselves naturally via speech or text, and for us understand exactly what they are looking for or asking about, is now possible with the recent AI-powered advances in Automatic Speech Recognition and Natural Language Processing. These technologies give us the ability to understand the intent of a customer, parse out the key pieces of what they are looking for, and either search our catalog with a superior level of precision, or connect them with the appropriate resource for issues like customer service. Taking the technology one step further into a commerce chat-bot or assistant, we are able to drive a multi-turn conversation with the customer, helping them refine their query or find an answer in an expedited manner. Applying Natural Language Processing (NLP) to our rich corpus of User Generated Data. We get several thousand reviews a day. Today, we already use NLP to Moderate the reviews automatically into Accepted, Rejected and Discouraged. The AI technique we use are very robust to spelling and grammatical errors, as well as transliterated words from Hindi and regional languages. We detect profane reviews as well as spam automatically. Algorithm accuracy is 98%+. We have develop deep native understanding to derive sentiment, build an automatic product scorecard to extract key reasons to buy weighted by user reviews.

Location Insider: How much traffic do you have on your website? How much of this traffic comes from mobile devices? How many transactions come from mobile devices?

Ram Papatla: India is a market that is backed by Smartphones and now penetration of affordable data plans. Our growth is reflective of the macro situation in India, we are the most frequent visited commerce Website, with 70% from Mobile Apps and other 20% from Mobile Site. As more Indians started getting access to data, we’ve ensured channels such as our Mobile Site are on parity with our Mobile App.

Location Insider: Should retailers integrate the mobile channel into their client’s customer journey and how?

Ram Papatla: I can’t think of a good reason to pass up on a strong Mobile footprint for any Retailers customers. Customers are on Mobile and are expecting smooth, friction free shopping experiences. I think the answer to how is pretty simple – ensure that the primary experience for customers to engage is built for Mobile and not a compromise. All of our workflows will have to be rethought starting with how customers discover new products, should you integrate a new finding method other than Search, expectation of payments, creating a strong community to shopping reviews which is reflective of Mobile customers experiences on other sites. I think there is also a tremendous opportunity for Retailers that play in offline and online to enable to new shopping experiences such as touch and feel of an online exclusive product in an offline store.

Location Insider: Which role do you see for brand manufacturers in online retailing, especially for the grocery brands?

Ram Papatla: The way we think about Brands is they are offering a customer experience with us, so it’s pretty important. Brands are starved for data to clearly understand how customers are engaging with them in a commerce world. We have a fantastic product called Brand Index which I will feature in my talk that helps them understand customers awareness, consideration and purchase on our platform. With these powerful insights, Brands can plan for the right interventions.

Location Insider: Thanks for the interview – see you at Mobile in Retail Conference!

Newsletter abonnieren

Abonnieren Sie den kostenlosen Newsletter von Location Insider. Wir liefern darin täglich gegen 11 Uhr business-relevante Hintergründe zur Digitalisierung des Handels.

Hiermit akzeptiere ich die Datenschutzbestimmungen.

Artikel teilen