Here's How Zomato Improved OCRs

In partnership with

Hey 🖐

Ever scroll through Zomato for like, an hour, drooling over pictures 🤤 and then... NOPE. Out of the app you go, with nothing but a rumbling tummy and a vague sense of indecision? A lot of people do that. But guess what?

There is a goldmine of data behind that struggle, and Zomato's been digging in to turn those window shoppers into loyal customers 🤑

Speaking of which, ever wondered how Zomato seems to know exactly what you are craving before you even do? Nope, it's not magic 🌟 it's data!

Grab a snack (or order one from Zomato!), and let's dig into the delicious details of how they pulled this off. Trust me, it's a game-changer.

Deep Dive into Zomato

Zomato is India's leading online food delivery platform, but its reach extends beyond deliveries. It has become a one-stop shop for everything food-related, boasting over 100 million downloads and serving a massive customer base. So, what does Zomato offer?

• Restaurant discovery: Zomato allows users to browse restaurants across various cuisines and locations. They can filter by cuisine type, delivery time, average order value (AOV), user ratings and reviews, and even study curated collections for special occasions or dietary needs.

• Food ordering & delivery: Users can place orders directly from their favorite restaurants or find new ones through Zomato's platform. Zomato partners with a vast network of delivery partners to ensure swift deliveries.

• Table reservations: Going out for a dine-in experience? Zomato allows users to book tables at partner restaurants, eliminating the hassle of waiting or missing out on a coveted spot.

• Online reviews & ratings: Zomato allows users to review and rate restaurants. This valuable user-generated content helps others make informed decisions about where to eat.

• Zomato Pro: Under this membership program, people get exclusive benefits like free delivery on orders above a certain value, restaurant discounts, and priority service. This program helps Zomato boost customer loyalty and generate recurring revenue.

Now, you might be thinking about how your favorite restaurants come to Zomato and how they process your order. Zomato's reach extends beyond individual consumers. It offers solutions for restaurants as well:

• Zomato for business: This suite of tools allows restaurants to manage their online presence on Zomato, including menus, photos, and customer reviews. It also helps them with analytics tools to understand customer preferences and optimize their offerings.

• Hyperpure by Zomato: This is a quick commerce venture by Zomato, delivering groceries and other essentials within minutes. This addition allows Zomato to tap into a large market and address users' immediate needs.

After all this, Zomato has revolutionized the food industry in India by:

• Increased restaurant visibility: Especially for smaller eateries, Zomato provides a platform to reach a wider audience and compete with established brands.

• Convenience for consumers: Food discovery, ordering, and delivery are simplified into a single platform, offering massive convenience to users.

• Growth of food delivery: Zomato is a major driver of the Indian online food delivery market, creating new opportunities for restaurants and delivery partners.

The Challenge

While Zomato boasts a massive user base and offers a convenient food delivery experience, the struggle the company often faces is - how can personalized restaurant suggestions increase user engagement and order frequency.

The primary objective was to improve the user experience by providing personalized suggestions that target individual preferences, thereby driving higher engagement and more frequent orders.

Specific Challenges Faced

1. Understanding user preferences

• Zomato needed to analyze a vast amount of user data to understand individual preferences accurately. This involved parsing user interaction data, such as clicks, search queries, previous orders, and demographic and behavioral data.

2. Data integration

• Integrating data from various sources was complex. The challenge was to combine user interaction data, demographic data, and restaurant information into a cohesive dataset that could be effectively analyzed.

3. Developing an effective recommendation algorithm

• The team needed to fine-tune algorithms that could provide accurate and relevant recommendations. This required choosing the right combination of collaborative filtering (leveraging user similarities) and content-based filtering (considering item similarities).

4. Ensuring real-time processing

• Personalized recommendations needed to be generated in real-time to improve the user experience. It needed a robust backend infrastructure to handle large-scale data processing and deliver quick responses.

5. A/B testing and validation

• Zomato had to design and implement A/B testing to validate the effectiveness of personalized recommendations. This involved dividing users into control and test groups, ensuring the tests were statistically significant, and accurately measuring the impact on key metrics.

6. User acceptance and trust

• Introducing personalized recommendations required careful review of user acceptance. The challenge was to ensure that users perceived these recommendations as helpful and trustworthy rather than intrusive or irrelevant.

Data-Driven Question

The core question driving this initiative was: "Can leveraging user data to provide personalized restaurant recommendations improve user engagement and boost key metrics such as the number of orders per user and user retention rates?" To address this question, the team focused on the following sub-questions:

• What patterns and preferences can be identified from user interaction data?

• How do different demographic segments respond to personalized recommendations?

• What is the optimal way to present these recommendations within the app to maximize engagement?

• How do personalized recommendations impact short-term metrics (e.g., click-through rates, order frequency) and long-term metrics (e.g., user retention, lifetime value)?

Action Taken

To tackle these challenges, Zomato's team adopted a structured approach:

1. Data collection and cleaning

• They aggregated data from multiple sources into a centralized data warehouse.

• Cleaned and preprocessed the data to ensure accuracy and consistency.

2. Segmentation and analysis

• The product team segmented users based on various criteria, such as demographics, behavior, and past interactions.

• They conducted in-depth analysis to uncover patterns and preferences.

3. Algorithm development

• They developed recommendation models using a combination of collaborative filtering and content-based filtering techniques.

• They have then iteratively tested and refined these models to improve accuracy and relevance.

4. Implementation and testing

• The team implemented the recommendation system within the app and conducted A/B testing to compare its performance against generic recommendations.

• They analyzed the results of A/B tests to validate the effectiveness of personalized recommendations.

5. Continuous improvement

• They set a feedback loop to gather user feedback and data regularly.

• And they made ongoing adjustments to the recommendation system based on new insights and performance metrics.

By addressing these challenges systematically, Zomato aimed to leverage the power of data analytics to provide a more personalized and engaging user experience, ultimately driving higher user satisfaction and business growth.

Data and Methodology

Zomato utilizes a powerful combination of data sources and analytical tools to tackle the challenge of low order conversion rates. Here's a breakdown of their data & methodology:

Data Sources

• User behavior data: They track how users interact with the app. Every tap, swipe, and search tells a story. Zomato monitors:

  • Searches performed and filters applied

  • Time spent on restaurant pages

  • Abandoned carts (orders not completed)

• Restaurant data: Information about Zomato's partner restaurants is crucial. They analyze:

  • Cuisine type and dietary options offered

  • Average order value (AOV) - how much a typical order costs

  • Delivery times

  • User ratings and reviews

• Historical order data: Past orders are a treasure trove of insights. Zomato examines:

  • User demographics and location

  • Time of day orders are placed

  • Past order preferences (what users typically order)

Analytical Tools

• User segmentation tools: Zomato groups users based on shared characteristics like:

  • Age and location

  • Past order history

  • Browsing behavior (what cuisines they tend to search for)

• Recommendation engine analysis: They evaluate how well the current recommendation system performs. It includes understanding:

  • Whether suggestions are relevant to user preferences

  • If the algorithm prioritizes restaurants with high conversion rates

• A/B testing: This scientific approach allows Zomato to test different features and see what works best. They might:

  • Compare different recommendation algorithms

  • Test variations of the user interface (UI) for search and filters

By combining this data and leveraging the power of analytics, Zomato gains deep insights into user behavior and restaurant performance. It empowers them to make data-driven decisions that improve the user experience and boost conversions.

Results

The personalized restaurant recommendations by Zomato yielded significant improvements in key metrics, showing the effectiveness of data-driven product improvements.

1. Time spent on the app: Users who received personalized recommendations spent an average of 35% more time on the app than those shown generic recommendations.

2. Click-Through Rates (CTR): The click-through rates for personalized recommendations were 28% higher than those for generic recommendations. This suggests that users found personalized suggestions more appealing.

3. Number of orders per user: There was a 22% increase in the number of orders per user per month among those who received personalized suggestions. This indicates that tailored suggestions encouraged users to place more frequent orders.

4. Order frequency: Users who received personalized recommendations placed orders 1.5 times more frequently than those who did not. This significant increase highlights the impact of personalization.

5. Retention rate: The retention rate for users receiving personalized suggestions improved by 18%. This means users were more likely to return to the app and continue using it.

6. Average Order Value (AOV): The average order value for users with personalized suggestions was 12% higher than those who received generic recommendations. This increase in AOV contributed to overall revenue growth.

7. Monthly revenue: Monthly revenue saw a 15% hike due to increased order frequency and higher average order values. This shows the notable financial impact of implementing personalized recommendations.

Conclusion

Zomato's personalized recommendations aren't just about suggesting what to eat—they are about creating a more engaging, satisfying, and loyal user base. It's a win-win for both users and the business, proving that when it comes to data and personalization, everyone can benefit.