Pepperfry uses AI, ML for product discovery

Delivering modern shopping experiences with yesterday’s technology has become challenging. How are you simplify your technology footprint to innovate quickly and deliver intelligent customer experience?
The technology marvel is evolving at a great speed and dominates a large aspect of our lives today. As far as e-commerce is concerned, the core focus is on how we use technology to help us deliver on our strategy. The choices that we make on technology are not driven by explicit simplification intent. We believe technology is the tool that helps one manage business and operational complexity. In our experience many of the more mature technologies have also evolved substantially over the years. A good example is PHP, which originated as early as 1994 and is therefore one of the oldest web technologies around. However PHP has seen massive improvements over the years. These improvements come in the form of performance – PHP7 introduced in 2015 is twice as fast as PHP 5.6 – and better error handling and robust frameworks such as Laravel. It is also relatively simple to hire and train someone on PHP, making it easier to get resources to hit the production floor quickly.

We also use Java to power all our services. Java has seen a resurgence in popularity since it became the default platform for Android developers – even though Kotlin popularity has seen a sharp spike of late.

Pepperfry’s technological architecture is completely developed organically and in-house. Our discrete products enable the front-end experience, the fulfillment and logistics experience, the customer support experience as well as the omni-channel experience. We bind these discrete systems together with a Data Control Network which ensures that appropriate information is available to the relevant user, both internal and external, at the right time.

Moreover, we have been focused on enabling our mobile application for furniture shopping through a tremendously user-friendly interface with superior technological amalgamations. As new technologies emerge, it is becoming increasingly important for retailers to extend the brick-and-mortar experience to their online channels.

Consumers demand an omni-channel shopping experience, from buying a product to returning it. How are you delivering in this area?
A segment of consumers is still cautious while making purchases online as they prefer the touch and feel aspect when it comes to big-ticket products like furniture. Customers are uncertain of the end quality of a product that they have only seen online. Taking these aspects of the Indian consumer into consideration, we decided to pioneer the omni-channel approach in the industry. We have turned, what is classically considered the omnichannel experience, on its head and its worked very well for us. While historically many organisations implemented omnichannel by trying to replicate an offline experience in their online channels, it has often not been able to reap anticipated results because organisations are rarely able to avoid internal channel conflicts and online (which is the more scalable part of the business) ends up taking a backseat.

Therefore, being an online company, our approach towards omnichannel business is different. We established Pepperfry Studios wherein we aim to replicate and offer our consumers the online experience. Customers who walk in are guided through the site while simultaneously the products that are in the Studio serve as illustrations for what the customer eventually ends up buying. So customers get a ‘touch and feel’ experience but at the same time their purchase choices are not limited to the products which are available in the store. Instead, they have our entire catalogue to make a purchase decision. Not only does this provide a better selection to customers and a more fruitful experience, it also eliminates the traditional channel conflict where each channel – online and offline – is competing to reserve inventory for itself. This approach has been a complete game changer for us, and our Studios are now contributing to more than 30 per cent of our total business volume.

Additionally, we are focused on driving users from our front-end platforms (website, mobile site and mobile app) into our Studios, wherever we feel that they will benefit from the expertise and assistance of our design experts. For example, a user who has placed products into the shopping cart but then subsequently not taken any action or for some reason was unable to complete the online payment, is an individual who will benefit from an assisted buying experience. Through our technology, we pipe these leads into the nearest Studio which then gets in touch with this user and in a large number of cases is able to convert the lead into a satisfied customer.

How are you delivering AI-driven experiences to customers; what are the new areas where you investing in terms of technology-specific use cases of AI, RPA and analytics?
Pepperfry is heavily investing on new-age technologies with an aim to become the most tech-enabled business. We have been using AI and machine learning extensively to provide consumers a superior shopping experience. One of the largest problems that an e-commerce platform attempts to solve is Product Discovery for users. When a customer visits our portals for the first time, we find that they often have a broad sense of what they want, but are open to evaluating multiple options. Therefore, our task is to highlight the most relevant products in order to display something that they like and therefore move forward in the purchase journey from simply browsing to actively evaluating a product.

This is one of the areas where we make extensive use of AI – or more accurately machine learning – to determine the appropriate products for a particular customer and the sequence in which these should be displayed to them. Essentially, we decode each Browse & Search session into multiple attributes or factors; these could be product attributes, platform attributes, customer attributes, temporal attributes, etc. They are then used to determine both, the set of products to be displayed and the order in which they should be displayed.

We also use a similar ML based approach to determine what are the characteristics of products that are more successful than others and therefore what kind of products should the Supply or Category teams be sourcing. ML is also used to power recommendations that we display to users. For example, if a user is on an item specific page, we are able to display similar products to that user as well as products that are frequently bought along with that particular product, so that we are able to expand our users’ purchase basket.

Finally, an application of ML that we are enthusiastic about and are investing behind in, is Visual Search. This will help customers get a set of similar products based on a piece of furniture that they might come across offline. For example, if I went to someone’s house and liked a particular sofa that they have, Visual Search will help me identify the products on Pepperfry that are most similar to the one that I liked offline.

As far as AR and VR is concerned, we believe there is some time before these technologies mature and realise their full potential in terms of customer adoption. There is no doubt that 3D technologies help visualise products better; however its unclear how big the delta is in terms of customer experience, especially when you consider some of the constraints that users in India face. To experience AR, there is a significant cost in terms of the app size, which translates into data costs for users. If we were to build AR into our mobile apps, then a large fraction of users might have to download a significantly larger app than they need and this can quickly lead to friction. However, this is definitely an area that we are investing behind and we also have a superior concept app ready. This application will be launched when we are certain that a large segment of consumers is actually able to make use of the functionality.

Furthermore, we have enabled a 3D experience for a large number of our furniture products where a user can see a product in 3D and also interact with that product even on their mobile phones. One can change the orientation to view it from different angles, zoom in and out and also get the experience of opening cabinet doors, pulling out drawers, etc. We have noticed a significant lift in conversion rates for products which have 3D models which clearly conveys that there is high consumer affinity towards this feature.

As compared to traditional companies, what do the startups like Pepperfry look for in the tech architecture? What sort of product innovations are in the pipeline?
Startups have limited resources in comparison to traditional companies. Resources translate into both time and manpower, hence our architecture needs to be such that it can help us scale and deploy with a quick pace. At Pepperfry, we have opted for a service oriented structure which helps us build our functionality in small but impactful portions, without interfering with processes that are already ongoing and effective. We deploy a feature as soon as it has been developed and tested, thereby reaping quick benefits. The service oriented architecture also helps us achieve higher software reusability and avoid multiple pieces of code doing that same thing, therefore making the job of the QA / Testing team simpler.

One of the other key areas where we are investing is the ability to optimise last mile delivery. This is a problem where we are looking at improving asset utilization. If a single vehicle can do more deliveries in a single shift or in certain cases do more shifts in a day, we gain significant supply chain efficiencies – both in terms of time taken as well as cost. We have run a successful pilot in one of our larger markets and are looking to roll this out across the country soon.

Technologies like data analytics, AI, machine learning are slowly disrupting customer services. How is Pepperfry looking at these opportunities?
Pepperfry is India’s leading online home and furniture marketplace wherein our customers spend more than INR 10,000 per transaction on an average due to furniture being a big ticket category. Therefore it becomes essential for us to overcome any trust related barriers with our potential customers by engaging with them through multiple touchpoints and superior services.

For starters, we have built a customer support platform from scratch. This has allowed us to ensure that an agent who is interacting with a customer has all the relevant information available to her in order to be able to resolve the customer’s query in the first contact itself. By ensuring that we have high FTR (First Time Resolution) rates, we are able to reduce the number of contacts required for resolution and therefore maintain a high level of service using a relatively modest sized team.

While the customer always has an option to interact with a real agent during the course of their purchases, we also have a chat experience powered by bots, ensuring that routine questions are answered seamlessly in real-time.

What are your plans for expansion, in terms of reach and customer base?
In a span of 7.5 years, we have been able to successfully entrench our position as a leader in the home and furniture segment with an online traffic market share of more than 60 per cent.

Growing at a CAGR of 83 per cent, we aim to grow to be a billion-dollar business by March 2021. We are scaling up our offline presence across markets extensively, especially in key catchment areas to be easily accessible and enable superior consumer experiences. With the introduction of Pepperfry Bespoke, we have also evolved into the nation’s first full-stack player in the home and interior segment and hope to continue to grow in this segment that is valued at $30 billion annually.

For categories like furniture and home solutions, it is essential for us to interact with our consumers through multiple touchpoints. We aim at launching 150 Studios by March 2020, while also offering a wider product catalog to our consumers by scaling up our 10 existing house brands to serve a larger demographic and their varying needs. Through our last mile delivery network of more than 350 delivery trucks that currently cater to over 500 cities across the nation; we aim to double our footprint in the next couple of years.