AI/ML Case Study on Swiggy

This blog has been written as a part of the Capstone Project to analyze Swiggy — one of the largest food aggregator platforms. This has been done to explore and understand the feasibility along with the benefits of using AI/ML to improve the product. This blog intends to summarize the in-depth analysis of the product, the insights that have been gathered, the feasible recommendations along with a plan of execution in a concise manner.

SWIGGY — UNDERSTANDING THE PRODUCT

Background

Swiggy is Indian online food ordering and delivery platform headquartered in the city of Bengaluru, India and operates in 500 Indian cities, as of September 2021. Swiggy was founded by Sriharsha Majesty, Nandan Reddy, Rahul jamini in 2014. Swiggy was inspired by the concept of providing a complete food ordering and delivery solution. The platform intends to bring the popular neighborhood restaurants closer to the urban foodie. The platform has a single window to order food from a wide range of restaurants. They have their own delivery partners and hence a ‘no minimum order policy’ on any order or restaurant. They accept online payments on all orders. Trusted delivery partners ensure reliable and speedy deliveries to treat their customers with food that is hot and fresh every time they order.

Business Objectives

Mission: Swiggy’s Mission is to change the way India eats.
Vision: Swiggy’s vision is to elevate the quality of life of urban consumers by providing unparalleled convenience.
Tagline: Swiggy karo, phir jo chahe Karo !

The objective of the company is to disrupt the kitchen at home and make their delivery service a viable option for an everyday food need. Rahul Jaimini, one of the three co-founders sums up Swiggy’s objective as — ‘make kitchens obsolete’.

Guiding Metrics

The major metrics are DAU/MAU, total subscriptions, average number of orders per user, and average value of orders per user per week. These are good indicators of user retention and engagement.

Assumptions

The company is performing well and is planning to expand its footprints across India. In the era of modern new emerging technologies, the company wants to invest in AI/ML to explore better customer solutions. The company wishes to offer hyper-personalized experience to its users to improve its user engagement further.

UNDERSTANDING THE USERS

Insights from the Surveys and Interviews

1. People who frequently consume outside food, they want to eat healthy and hygienic food. Since they order food frequently from food delivery apps, they would like assistance that could help them search for better meals (better in terms of quantity, quality, calorie content, cooking style).
2. People who daily consume outside food, want to eat different items on different days to follow a weekly/ monthly routine. Since they repeat the order cycle every week/month, they would like assistance to set up a meal plan that ensures food is ordered/delivered to them accordingly.
3. People who frequently order from outside, want to limit their budget/expense. Since they order food frequently from food delivery apps, they wish to receive assistance that could help them search for more discounts on food deals than restaurant ratings.

Pain points / Frustrations

1. Difficult to ascertain and manage both quality and quantity of food.
2. Inconvenient to match tasty as well as healthy food.
3. Difficult to search for home cooked food.
4. No option to set up a fixed meal plan to ensure food is delivered on time.
5. Difficult to actively search for relevant discounts on food deals.

Aha moments

1. One-stop destination for favorite food and restaurant search.
2. The ability to pay using such a huge variety of options. Very convenient.

PROBLEM STATEMENT

The app experience is not as personalized as most users want it to be. Considering the sedentary nature of our lifestyles in today’s time especially post the onset of the pandemic, users who are ordering frequently wish to maintain a diet by adopting a meal plan that would ensure to maintain a diversity of food in their orders while not degrading their health even more. Also, they look for better discount options so that it doesn’t end up burning a hole in their pockets. Due to the lack of appropriate features to address these pain points, Swiggy is at risk of failing to attract new customers and also in retain current users who want to create a positive impact in their daily lives and want to make smart and healthy choices. AI/ML can be leveraged to address these user pain points and find a feasible and appropriate solution to the problems in hand.

NDERSTANDING OF THE CURRENT USER BASE

The current user base represents majorly the unmarried young professionals who are tech-savvy. These users are willing to experiment and try new delicacies regularly. These users do have a strong opinion about restaurants and their food service. The users want to enjoy a good, tasty meal while not compromising the healthy quotient of it. These are busy professionals who often do not have the skill or the time to cook. Hence, they demand fast deliveries and good service.

Assumptions

1. The users are particular about the Quantity and Quality of Food.
2. The users prefer to spend time at home rather than going outside. They want a platform to order their preferred food from their preferred restaurant.
3. The users are health conscious.
4. The users follow diet plan or routine. They want to pre-plan their meals which allow them to place orders days/weeks in advance.
5. The users want a fixed meal budget. They want to have a budget plan per week/month so that they spend ordering food within the budget.

PROPOSED SOLUTION

Reimagined Product

Using Machine Learning model to better collect and store data collected from the user and optimizing the classification and recommendation algorithm which will consider a lot more data points to improve the recommendations and personalize the app according to the user and their needs.
This will be implemented in phases and the MVP (Minimum Viable Product) will have the following features as the must have features :

1. Provide a “food graph” which breaks down a food dish by recipe, cooking style, ingredients used, calorie value, and variations of the dish
2. Smart notifications based on schedule meal plan of the day
3. Hyper personalization of restaurant search based on meal recipes Also, the app will later be incorporated with nice to have features:
4. Related Suggestions
5. System powered meal and budget summary providing the budget estimation guide for the entire month
6. Customized feed by blending the prices of food with the users’ past food preferences

MOCKUPS & PROTOTYPE

Home Page

As a part of the three main features to be included in the MVP, the following features are included on the home page of the app:
A questionnaire to personalize the app content and recommendations for the newly signed up user or the logged in user.
Reason- This will help in personalizing the content of the app and recommending better options according to the user’s preferences.

1. Smart notifications based on the temporal data, order history, calorie content and cooking style of previously ordered dishes. Reason- This will help in reducing the search time of the user.
2. Weekly curated recommendation list based on user order history and preferences filled up in the questionnaire. Reason- The recommendation list will help in getting to know user choices better to improve the recommendations further.

Search Page

As a part of the main features to be included in the MVP, the following features are included on the search page of the app:
Multiple filters to help users in finding the desired food item in less time.
Reason-Will reduce the search time of the user.

1. Most frequently ordered dish recommendations. Reason-Users willing to try new food items can have their pick using this. This will give us information about user choices also.

Restaurant Menu page

As a part of the main features to be included in the MVP, the following features are included on the restaurant menu page of the app:
Showcasing the food graph ( ingredients used, cooking style, calorie value, and variations of the dish) for all the items.
Reason-This will help the user in making smarter and healthier food choices as it is providing all the information that they are looking for.

Link for the prototype : https://marvelapp.com/prototype/7cf53dj
METRICS

The following metrics will be considered to map the performance of the features:

1.Restaurant Menu Click Through Rate (CTR) — The no. of clicks made on the menu of the restaurant. More no. of clicks will indicate more user engagement and vice-versa.
2.Offers/discounts click through rate — The no. of clicks made on the different offers or discounts. More no. of clicks will indicate more user engagement and vice-versa.
3.Time Spent on the app before placing the order– Time spent by the customer on the app before they end up placing an order indicates the duration of search time on the platform.
4.Recommendations leading to a conversion — Count of the no. of recommendations which got clicked and ended up in a successful conversion (to a placed order) can help us in mapping the precision and recall of the recommendation algorithm.
5.Frequency of login — The no. of times (consecutive times in a day/consecutive days in a week) a person logs into the app indicates the degree of user retention.

Short Term Goals

To reduce the search time of the user (time spent in searching for the preferred dish).
2.Improve the efficiency of recommendations.

1. To make the app more personalized for each user.
2. To improve customer engagement.

Long Term Goals

1.To attract a much more diverse set of customers (the ones who don’t form a part of the existing customer base)
2.To introduce the concept of meal plans and subscriptions (like a tiffin service/community based food service)

1. To incorporate the facility of providing not just restaurant based food but also home cooked food to the customers according to their preferences.
2. To increase customer retention

SUMMARY

The pain points in the existing app were identified and have been attempted to be solved using AI/ML in the solution proposed in this article. Implementing the above mentioned features will not only help in improving the customer engagement but will also help us in collecting more information about the user choices which will help in training the ML model consistently and will thus improve the quality of recommendations given to each user. This will make the app more personalized for each user. Thus, it will help us in achieving our ultimate goal of increasing user retention in turn driving revenue up the scale.