Toast App

My thesis in graduate school was a project called Last Eats. It was a web platform of restaurants recommendations based on radical selectivity. It was positioned as the opposite of Yelp - trusted, selective recommendations for the best restaurants based on the knowledge of people you know.

What we learned by making Last Eats was that there was a community that craved an alternative to Yelp and the other anonymous recommendation engines. They wanted connection and they sought experiences that they couldn't find elsewhere. After taking into account all of our user feedback and the sustained appetite for the more mood, experience and location based discovery, we decided to break from our earlier platform and create something new.

Enter Toast.

Toast is a food recommendation app that leverages your trusted network to make the decision of where to eat the fastest, easiest and most satisfying experience possible. We’re doing this through simple, trusted, actionable recommendations from friends, driven by mood and location.

Like it's predecessor, Toast is a word of mouth recommendation platform, but with some distinct differences.

  1. It's a native iOS app
  2. Mood and location are the foundation of the platform
  3. You can add as many Toasts or recommendations as you want. There is no limit, but you can designate a current favorite or "Top Toast" 
  4. Restaurant reservations are enabled
  5. We're launching in NYC only (with an emphasis on everything south of Union Square as well as Williamsburg)

From Last Eats to Toast

When we started Toast, our model of radical selectivity evolved. We softened the be-all, end-all singular destination limitation and extended our premise to include extraordinary food experiences that are tied to mood and emotion. We realized with all of our research into Yelp/Google and all other impersonal digital food discovery tools that mood based discovery was highly unreliable on other platforms.

If I'm looking for great Thai food, I can source any generic digital restaurant ranking/evaluation tool to find an approximation of this end desire. If on the other hand I want restaurant experiences that are "fun", "comforting" or "adventurous", these criteria are distinctly personal and much harder to reliably source from strangers - or people who don't know the nuance of my own personal taste (i.e. my definition of "adventurous" is likely very different than that of a random person on the internet).

As you can see below, when a user types, "Adventurous Restaurant" into Yelp, the only results that come up are extremely fancy, expensive eateries with no logic or justification as to why each is recommended or what differentiates it as adventurous. This is a prototypical example of the failure of high profile recommendation engines to reliably source trusted, mood based experiences.

Competitive Analysis and Market Positioning

We began our analysis of the market, and the entrenched competitors within, by systematically breaking down the food discovery leaders into relevant categories. These included: tastemakers, established tech leaders, first movers, photo based discovery, restaurant delivery/reservation apps, socially optimized food discovery and industry insiders.

We then partitioned the companies into quadrants that express their relative positioning to Toast respective of relevance and trust.

If Toast could deliver the most trusted, relevant recommendations that were tied to delivery, take out or reservations, we could differentiate our offering immediately.

Investigating Successful UX Patterns

As we began to think about how we wanted to build the fastest, easiest, mood driven food discovery tool, we did an audit of popular apps that were driven by stellar user experience. These did not have to be food discovery apps, as we wanted to identify UX patterns that were easy, intuitive and fun to use while helping their intended users make quick, empowered selections.

Although we ideated on dozens of different ways to select a mood and location, we ended up employing a vertical carousel pattern that was made famous by Urbanspoon, one of the App Store's first wildly popular food discovery apps. We wanted this process to be quick, elegant and fun and we feel the carousel accomplished this goal.

With regards to the content discovery process, we used a quality over quantity paradigm and wanted to make sure we gave each restaurant its proper consideration. As a result, we never show a list of options, but instead each option is presented one at a time. Thankfully the explosive popularity of Tinder normalized this and made the lateral swipe progression powerful, intuitive and easy. Unlike with Tinder, we wanted our users to have the power to browse forward or backwards. There are plenty of apps that do this effectively so we had lots of confirmation that this style of content discovery could be applied to our platform.

Since the value of our platform was directly linked to diverse, quality recommendations, we had to make the process of making a toast  as easy as possible. We leveraged a common content creation pattern of top right additions - a convention made popular by Twitter, FB Messenger and many other products.

Testing Out Different Approaches

Before we committed to our first build of Toast, I designed a few alternative approaches that we tested via flinto, InVision and other prototyping tools. Some of these took similar approaches and some were drastic departures.

Among our considerations were the following. We tested a home page where we showcased a user's friends who are the most active on the platform as "Toastmasters." Furthermore we gave them a place of prominence on the bottom third of the screen enabling a user to quickly see a their most recent Toasts specific to a location and mood. This ended up being a feature only potential power users would benefit from most. It also ended up complicating the first page of the app unnecessarily.

Additionally, we tested a swipe to discover experience that had a more formal, Zagat-like brand and feel, with multiple friends photos shown on the same line as if to indicate their participation in a group conversation. At one point we also hoped to integrate and partner with a few select delivery services, so we explored exposing that capability within the bottom fixed navigation bar.

Our full on Tinder for restaurants model tested the worst by far....

First Point of Contact in App

When a user logs into Toast, we immediately have them select the mood that they are feeling and or seeking. Then we default to their current location, although they can easily choose a different one if they prefer.

The moods we focused on for our launch included the following: Adventurous, Boozy, Casual, Celebratory, Cheap, Comforting, Fun, Healthy, Hungover, Indulgent, Light, Romantic, Simple and Swanky. This list was culled from extensive user research sessions was past users of Last Eats and new users from our target demographic (18-45 year old restaurant goers).

Once they select a mood and a location we immediately showcase a series of restaurant that fit the collective criteria, sourced from friends and defaulted to the most recommendations and or closest location. When a user finds a restaurant that looks appealing, they can easily read toasts (reviews) from friends, determine the type of cuisine, price point and walking distance (although the typical walking distance for a user was under 30 mins, the screenshots below were taken away of NYC, thus generating wild distance estimates).

A user can then click on a restaurant to see photos of the food, a map of the location, menu, directions, website and more. The Foursquare API provides much of this data for us. Our task was to take all of this data make it as consumable and immediately actionable as possible.

Our ultimate goal with Toast was to get our users to an incredibly delicious, unique dining faster and more reliably than any other food discovery platform. First, we made every restaurant on Toast tied to one click Pick Up or Reservations. This was a critical point of emphasis for us as means to enable our users to take action on a compelling option faster and easier than any other food discovery tool.

Not Just Friends, But Friends of Friends

One of the strategic differences between Toast and Last Eats was finding a way to meaningfully expand the circle of trusted reviews. As a result we followed the model that was leveraged by the popular dating app, Hinge. We prioritize your immediate friends' recommendations but also those of your friends' friends. This third tier allowed us to grow our repository of quality, mood based recommendations while maintaining trust as one of the central tenets of our value proposition.

I am great friends with Ian McCormick. Based on my proximity to and trust in Ian, I can reasonably expect his friends to have a relatively similar set of values. As a result, my trust in Ian enables my extended trust to Tal, Ian's friend. The intimacy within this dynamic is  inherently preferable to that of recommendations from strangers.

How We Determine What Restaurants We Show

Our algorithm takes the following variables into account :

  • A user's proximity to the restaurants that fit their mood
  • The number of times a restaurant is given is Toasted.
  • Whether or not the restaurant was given a Top Toast designation - which was the equivalent of the singular, be all end all Last Eats declaration. We brought back a level of radical selectivity from Last Eats with the arrival of Top Toasts. These are singular designations that are tied to just one restaurant, put on a pedestal. Algorithmically these toasts carry more weight than a regular toast and are thus positioned earlier in the discovery feed. As with Last Eats, these designations can change at any time.
  • The level of social engagement a Toast has garnered - i.e. the number of likes a review for a place has received.
  • Eventually, our goal was for users to self-select their best friends on Toast so we could appropriately bias their reviews as part of our ranking algorithm.

User Research Leads to New Features

In addition to the standard form of consuming content on Toast through the swipe to discover platform, there were a number of early, committed users that demanded another way to discover content. They wanted to see all of the Toasts specific to a mood or a friend relative to their position on a map. Additionally, we added the ability for users to just see Top Toasts declarations or to just see recently added Toasts.

We also added even more social functionality through an Activity Feed where users can see bleeding edge toasts made and endorsed by people within the community.

User generated hashtags added another layer of specificity and personalization that can be created and followed as another means of discovering distinct content. Although we provided some stock hashtags that we easy for users to select and add to their Toasts, we quickly noticed users were adopting their own tags for the experiences that mattered most to them.

    Lessons Learned and Future Opportunity

    Toast validated an approach that caters to experience...

    Burden of creation, peak food photography around time of launch, yet another thing to do now

    Network Effect

    Focused on a smaller demo, like college students?

    While we hit some major milestones and grew quickly, we had a hard time sustaining our growth and

    Wanted trusted recommendations to move into other domains where a friends recommendation for an experience, and the cost of getting it wrong is big - travel, hotels, bachelor and bachelorette parties.