PopX — Experience driven personalized food menu
Swiggy is a Bangalore based food delivery startup that started in 2014. It gained its unicorn status in a mere 4 years of being founded. From there it has traveled a long distance in the last 2 years to become India’s most popular food delivery startups taking over 36% of the market share.
An estimated 400 food delivery startups were open between 2013 and 2016 raising a total of whopping 120 Million dollars.
But Swiggy became the market leader due to its aggressive expansion strategies, high cash burn strategy on marketing and most important the product they offered to its customers.
Personally, Swiggy’s customer-friendly design is what makes me order from it a lot of the time. I find what I want to eat and I order it before I can change my mind.
But other times when customers don’t know what to eat or get confused due to hundreds of food items is when the customer engagement drops.
How to increase customer retention in food delivery apps due to the loss of customer engagement?
The major reason for the loss of customer engagement is:
Overchoice — is a cognitive process in which people have a difficult time making a decision when faced with many options.
- Presence of too many restaurants: Not that having more options is a bad thing for customers but the way you present the information is very important.
For example: If you show a customer top 10 restaurants to order from, on a single page, his time to make the best decision will increase. (s)He’ll start making trade-offs between places, then the food which is offered there, then the pricing and this is something which will cause a delayed response.
Studies show that over choice leads people to delay or completely opt-out of decision-making, report lower choice satisfaction, and make poorer decisions.
Suppose you want to buy something to eat and you get this menu in return. You start filtering places in your head like do I want a milkshake, or let’s have Indian food, what about a burger and so on.
2. Presence of too many similar food items inside each restaurant:
Giving customers more than 10 types of chicken burgers could confuse him over his decision to eat a burger at all.
Some form of UX design needs to be introduced which helps them understand the difference in flavour which makes it easier for them to know that Pesto Fried Chicken Burger tastes better than Crumb Chicken Burger.
If this job was done in haste, it can result in a generic food description, for example, a food description for a Chicken Steak Burger can look like it comes with Chicken Steak, buns, onions, and tomatoes which surely customer won’t be excited to read or order in this case.
One of the solutions which I’ll be discussing in this blog is something I call PopX. It provides personalized tailored food recommendations from different restaurants.
How big of an impact can personalized recommendations have?
When personalized recommendations gained momentum, around 35 percent of what consumers purchase on Amazon and 75 percent of what they watch on Netflix came from product recommendations according to a report by Mckinsey.
Now, we know that recommendations can have a major impact on increasing revenue of companies by increasing customer engagement.
How do we personalize customers’ recommendations?
In order to personalize customer-related needs, we first need to understand customer needs and empathize with the customer’s behaviour when he decides not to buy any food. For this we need data. Lots of data to understand what is happening, why is it happening and how can we prevent it?
- Active data collection:
Active data collection refers to getting data about what customer likes, dislikes by directly asking him. It can be in the form of a simple page asking him if he likes Indian Food or does he like spicy food more than sweet food, burger over pizza and so on.
2. Passive data collection:
For a particular customer, passive data can be collected from:
A. With his previous food orders, we can look into the repeated orders, to see what he likes to eat usually.
Non- repeated orders and cross-reference them with their rating which (s)he gave for it. We can try to understand if (s)he dislikes a particular category of food and gives it a bad rating, is it the food or whether (s)he dissatisfied with the restaurant.
Using this information we can show him the most similar food items which he can eat.
B. Food cart abandonment, looking at the data we can come up with hypothesis likes whether the prices of food were too steep or he was putting all the food items he wanted to eat at one place and then narrow down his choice after viewing the food cart.
In the end, this hypothesis has to be validated by data.
How to make it more accurate for a wide range of populations?
Collaborative Filtering: If a group of customers eat the same food items and rate them 5 stars daily, we can assume that they have the same taste. Here our decisions are made based on many customers(collaborating).
Example: Homer Simpson(The Simpson) has eating habits similar to Peter Griffin(Family Guy).
So, the next time when Peter Griffin opens Swiggy PopX, we can recommend him a bag of chips that Homer absolutely loves.
- Active data collection wireframe design is inspired by the Apple Music Artist selection system.