Delivering Contextual Search Results at Flipkart

With 150 million products across 80+ categories, shopping journeys vary drastically. Various factors inform the user's search experience: How urgent is the purchase? How much spending power does the user have? Where is the buying decision made? How important is imagery in this category? Helping diverse users find the right products for their unique needs can seem daunting.

I created an extensive framework for product cards that prioritises category × cohort × intent based design goals that achieved healthy upticks in basket additions.

Introduction

In e-commerce, there are three ways products can be surfaced to a user - Search, Merchandising and Recommendations. Of these, search is by far (70%) the best driver for basket additions. The key advantage of search, of course, is explicit expression of user intent.

To best serve the user, and make as much money doing it as possible, it is key to understand this expressed intent, and cater the experience with empathy. After all, the way a user shops for t-shirts is very different to the way they might shop for an air conditioner!

Search Tokenization

A user's search query is analysed by breaking it down into units, called tokens. They fall into a few well-understood types: Brand, Category, Product, Facet (basically any adjective) to name a few. By ranking popular searches, we can understand the spread of queries that come in for any filtered set of search journeys, and define design goals for that set. This "filtered set" can be a category of products, a user cohort, or any other grouping of queries.

Factors Considered

Based on data provided by the UX Research team, the PMs, Researchers, and I identified the following factors that influenced my designs:

1. Category

80% of mobile searches are model-specific

Purchase decisions are largely driven by offline social recommendations and youtube tech reviews. As such, the customer's buying decision happens off-platform. This can be seen empirically, from the fact that the majority of search queries are made of line or product tokens, for eg: "Apple iphone 14 256gb".

What that means for the design, is that:

1. The user generally only cares about the first result (on an ItemID level)

2. We have to showcase variants of the same device (color, storage, ram options)

3. In case the product is Out of Stock, we show that we understand the user's need and provide the next best fallback.

Grocery purchases don't need overthinking

In Grocery and BGMH (Beauty, General Merchandise & Home Needs), users are already familiar with the products, and often don't require a product page visit to make a buying decision. How often do you squint at the information on say, a packet of flour, before adding it to your basket? However, key details like manufacture date, expiry, quality markers, etc need to be available at a glance, as well as clear visibility on SKU options.

1. Add to basket from search page

2. Easy variant (SKU) selection

3. Mfg / Exp information

4. Quality indicators (star rating, trusted brand, etc)

Fashion categories encourage browsing

Search queries are usually generic, for eg: "red printed tshirt". The biggest decision maker in this case is the product image, and a user expects to scroll through a variety of options before choosing their favourite.

1. Imagery should be large and clear

2. Brand name is important in higher priced segments

3. Users should have visibility on variants (size, color, pattern) and similar products

Appliances are spec & jargon heavy

TVs and Air conditioners are planned purchases, and users often hunt for deals before making a purchase. Fridges and washing machines are more urgent purchases, prioritising quick delivery, most often to replace a broken appliance. Appliances usually have key specifications that vary subcategory to subcategory, and these drive the buying decisions.

1. User should have clear visibility of key specifications

2. Flipkart should explain jargon better to facilitate a smooth buying experience

3. Sorting and filtering are extremely important to help a user narrow down their intent

2. Intent

Some users need help narrowing intent

The user is shopping for an occasion or genre, without a specific need in mind. In order to serve their needs, we implemented a GPT layer to:

1. Identify subcategories associated with a user's query

2. Run parallel searches for related categories

3. Showcase flipkart's offerings, across categories and domains to the user

A shopkeeper can help inform the buyer's purchase

For even more broad or problem solving queries, it helps to provide users with step by step guides on how to make a buying decision. We experimented with a GPT layer that would:

1. Help a user choose from the range of buying options for their intent

2. Provide context and additional information to help them make a decision

3. Progressively narrow the user's intent over several clicks.

3. Cohort

Some users are easily overwhelmed

Low TPC (transactions per customer) users are defined as those having less that 3 purchases in a calendar year. These users are:

1. Price sensitive

2. Skeptical of inflated deals

3. Low loyalty

4. Low trust in platform

5. Easily frustrated / dissuaded

Learning and Challenges

Designing for Flipkart’s vast platform required balancing diverse user needs with business goals. Each category had unique search behaviors, making a one-size-fits-all approach ineffective. Collaborating with UX researchers and PMs was crucial to extracting insights and translating them into intuitive designs. The UX research team provided detailed reports on user behavior, pain points, and conversion patterns, which informed design iterations. Their assistance with user testing and discovery sessions helped validate design decisions and refine the final experience.

Reflection

This project reinforced that great design is empathetic and data-driven. Understanding user intent helped create a seamless and trustworthy experience. Collaborating cross-functionally was crucial in turning data into impactful design solutions. The measurable uptick in basket additions validated the importance of aligning design with user behavior, reinforcing my growth in designing for intent and trust.