In the times pre-COVID (and hopefully again soon), many of us enjoy discovering new places to eat. Most often, we discover those restaurants through word of mouth or use online forums that offer categorized lists of restaurants. To find and diligence these restaurants, we would use rating platforms like Google Reviews or Yelp. But in the past, there hasn’t been a data driven platform to offer tailored recommendations of restaurants to try. The Infatuation might be the closest, starting as a newsletter and moving towards the recommendation space, but it still primarily relies on a user actively searching its map or reading its posts versus automatically generating a unique recommendations list. This is where Beli, currently in beta, enters the picture, using data analytics to differentiate itself and offer users unique recommendations for new restaurants to try.
Beli starts off new users by having them rank restaurants they’ve been to against one another to determine the user’s preferences. Next, Beli also has these users identify who their friends on the platform are so that their respective preferences can be compared. In doing so, Beli can then use the data from these respective lists to generate restaurant recommendations for the users (based off similar tastes and correlations between rankings).
This is the core concept behind Beli, but utilizing their data capabilities, there are many opportunities for the company in the future. Within the restaurant space, Beli has differentiated insight into groups of users that fit different profiles and, even more, when these users look through the recommendations, they can bookmark ones they want to try (building even more platform intelligence). As such, Beli can allow restaurants to target groups that are most likely to be customers in the future either through retargeting efforts or through potential Beli products, like enhanced listings for restaurants that can target users that fit chosen profiles.
Even more though, Beli’s information is extremely relevant for companies outside of the restaurant space. By better understanding user preferences, other advertisers can utilize Beli aggregated data to target customer sets that are more likely to use / purchase their products. For example, for customers that have tastes for higher-end restaurants and seem to try restaurants around the world, high end hotel chains, such as Four Seasons, can retarget these customers on Beli’s platform. Additionally, these companies can use this aggregated data to target these customers off Beli’s platforms, on Google, Facebook, etc., using unique IDs (Beli currently ties to user email address). Given the unique insights that Beli has into its customers’ preferences, this is extremely relevant data for advertising efforts across the online space.
Of course, there are certainly challenges that come along with Beli’s model. First, for Beli’s algorithm to generate the highest quality recommendations and for its data to be worthwhile to other businesses, Beli needs to acquire a significant amount of this data. This requires attracting a large number of users to its platform and, given that early customers will experience less of a benefit (due to less data), Beli will need to make the initial experience valuable enough to encourage early adoption. Additionally, restaurant choice is often extremely personal and can be dependent on individual experiences (e.g. one person has bad service or the chef has an off night) – as such, it is incredibly complicated to adjust for these differences and generate the most accurate recommendations as possible for given users. And if users do not like the recommendations they are initially given, they are unlikely to trust or use the platform going forward. Last, the restaurant industry is a difficult industry in which to aggregate large amounts of data – restaurants are extremely localized with low repeat rates, making it difficult to acquire large amounts of information on an expansive number of restaurants. This forces Beli to rely on extremely strong data and algorithms to create comparable restaurant sets and imply similar user preferences and the resulting recommendations.
Given these challenges, I have several recommendations for Beli and its success in using data analytics going forward. First, they need to seamlessly consolidate significant data and should find a shortcut to do so – one recommendation is to integrate with restaurant reservations systems so that they can automatically update information on where users are eating and remind them to rank the restaurants immediately afterwards. This could prove difficult as reservation systems might see Beli as a potential competitor, but if Beli can add value in getting people to eat at more restaurants, then there might be some mutually beneficial partnerships. In doing so, Beli can build additional data to help more quickly ensure recommendations are accurate and personalized. Additionally, Beli should focus on selected cities to start rather than expanding too widely so that they can consolidate localized data and perfect the growth model. The platform serves little value without the data, so rather than spread themselves too thin early on, they would be better off building enough data in one area and proving their model. In these ways, Beli can help resolve many of the difficulties they will face in building the data and analytics necessary to power their algorithm and prove a success.