Frictionless Food Recommendation App
Description of Idea: There are dozens of food / restaurant recommendation mobile apps and websites, the most popular obviously being Yelp, but none are frictionless or work in the background. This app has higher quality recommendations based on more data while being much easier to use. The first step in setting up this app is syncing it with a variety of optional accounts that the user already has. The user would have the option to sync the app with their checking account and credit cards (like I do with Mint.com already) so that all restaurant transactions could be logged and analyzed. Additionally, the user could choose to also sync the app with their Yelp, Facebook, and Foursquare accounts to pull in existing check-in and restaurant rating data.
The problem with even the best food recommendation apps is they face the chicken and egg problem. They can’t recommend places or dishes to eat until they have enough data from enough different users. But they can’t attract users without having enough data. Worse is that many highly touted apps like Alfred begin by requiring you to ‘teach’ it your preferences by telling it where and what you like to eat.
That process is redundant. I started reviewing restaurants on Yelp in 2007 and I’ve eaten out several times a week for the last 10 years, paying with a debit card 90% of the time. There is a shit ton of data already there. Why do I need to start over by entering it in an app? Last.fm took this approach with music. The first step with Last.fm is syncing or ‘scrobbling’ your account with iTunes. My iTunes account already housed years of data about the music and artists I listen to. Last.fm used that existing data to recommend new artists and songs to me. They didn’t ask me to start from scratch and tell them who I like listening to.
Why can’t a food discovery app do the same?
The Problem Being Solved: This method is a more efficient way to discover restaurants that you have a high probability of liking. There is more data to analyze and requires less of the user. Making this a mobile app adds the benefit of being able to use the phone’s GPS to recommend nearby restaurants.
Another problem this app could solve is not knowing an optimal place for multiple people, a problem no other app seems to want to tackle. This app could recommend a restaurant to a couple or group of people. By associating another account with yours, the app could find common tastes between the account holders, factor in location and time of day and then make a recommendation. Again, this app is based on using as much existing data, primarily purchase history, as its main data source. Every time someone eats at a restaurant the app takes note. Understanding more about the user’s preferences at that point isn’t nearly as difficult as starting from zero. With Yelp, users assign a single rating to a restaurant as a whole. A 4 star rating doesn’t tell readers whether that 4 stars accounts for the food, a specific dish, the service, the ambiance, the prices or a combination of them all. The user might explain in their novel of a review, but that requires reading through each one. Often people will rate a restaurant 1 star because of the way they were treated by an employee that could be fired a week later, regardless of the quality of the food. They might specify that in their review, but their low rating is still reflected in the restaurant’s overall rating.
This app doesn’t start with asking people to rate things. It tracks the places they buy food from. If the app detects that a person is eating at the same restaurant frequently, it could prompt the user to explain by rating elements like service, food, parking, price, ambiance, etc. to understand more. If the user goes to a restaurant once in a year, why even bother asking them for their opinion? Their behavior indicates that there is something they do not like.
Lastly, utilizing the user’s purchase history as the main data set bypasses the typical chicken and egg problem. The app becomes more valuable as more people sync their accounts, but it is useful out of the gate to the very first user since it’s a personal log and recommendation engine based on what they’ve already demonstrated they like. Recommending new restaurants is as easy as finding commonalities between users in close proximity to one another.
Use the Data Collected
While asking a user to sync sensitive information like their checking account is a hurdle, using and gamifying the data collected in interesting ways could reduce that barrier. Imagine a cross between Last.fm and OkCupid’s Blog:
- “You ate Mexican food 18 times in May and 74 times in the last 6 months.”
- “You love In N Out after 11pm on Friday nights.”
- “Your friend John Doe eats at Soul Burger twice as often as you.”
- “You typically tip $3.58, a 6% decrease from last year.”
Key Features: The premise of this app relies on the ability to sync the app first and foremost with a checking account and/or credit card accounts to monitor food-related transactions. Other accounts that could be synced in addition to a checking account or as an alternative are Facebook and Foursquare for check-in data.
A strong differentiating feature would be the ability to ask for recommendations for multiple people. The app calculates the suggestions by comparing similar frequented restaurants between the parties involved.
The option to allow the user to fine tune preferences by providing ratings on aspects and even dishes would be nice as well as giving the user the option to add in their own review should they choose to do so.
Competitors: There are dozens of food discovery apps. None do anything close to this. Leading apps other than Yelp are Foodspotting, Nosh, and Alfred. Foodspotting and Nosh are picture based and Alfred tries to recommend restaurants by suggesting similar restaurants as ones you’ve told it you like.
In my opinion, Yelp is in the best position to implement this method of gathering data. They have a well-known name, tons of existing food related data and a wide user base but I believe their dominance has led to some complacency. Yelp’s challengers have struggled to overcome the chicken and egg dilemma and are picking differentiating angles that aren’t always that useful.
Furthermore, I believe that the majority of people aren’t actually looking for a new place to eat all that often. The most common uses cases seem to be when people are traveling, in a rush, a new relationship or are out with friends. Factoring in everyone’s tastes is difficult, yet the data is there.
I would love if someone would build this app! And who knew 42 year old athletic women were so easy to please?
Author: Emil Gallardo