Dr. Zachary Garfield
Research Director
Background
Restaurants are increasingly focused on delivery. Even before the COVID-19 pandemic, restaurants understood that food delivery was no longer just for pizza joints, but a way to scale up revenue without diminishing margins. A restaurant may only be able to handle 50 to 100 in-person covers per night, but delivery orders are only limited by the kitchen’s capacity.
COVID-19 made delivery into a necessity. It was, in many markets, the only way a restaurant could remain open. But delivery customers are distinct from dine-in customers, with their own unique preferences, order combinations, and motivations. They also have unique ways of signaling their satisfaction, or lack thereof, back to the restaurant. As part of its F&B marketing work, Fourdozen has conducted order analyses to help restaurant owners better understand what delivery customers were telling them, and make appropriate decisions about delivery menus, pricing, and specials. This article provides some background about menu analysis work we conducted in Hanoi, Vietnam during 2019 and 2020.
Problem
Restaurants and their staff need to better understand the demands, preferences, and behaviors of delivery customers. In many cases restaurant proprietors have access to rich data which can provide insights into their menus and their customers in ways that can inform strategic marketing decisions and increase the efficiency and profitability of operations. Many restaurant owners however lack the required statistical expertise, or are too busy to conduct the kind of data collation and analysis that can meaningfully inform evidence-based decisions. Moreover, point-of-sale systems are often not harmonized with delivery apps, leading to fragmented datasets that are essentially unusable.
Our Big Idea
Fourdozen solved this problem by making everything easier: easier to collect and aggregate data into analyzable sets; easier to draw conclusions from those data; and easier to take action.
Fourdozen conducted novel analyses of delivery order data from a popular local restaurant in Hanoi, Vietnam, spanning 14 months, comprising 1,500 orders from over 750 unique customers. These orders were from before, during, and after the first two waves of COVID-19 that hit the city, including during a total lockdown period when restaurants were unable to serve in-person.
Analyses were designed to provide actionable insights into how menu items actually perform on a delivery menu. It examined their interrelationships, and identified the items that drove customers to order again, and place larger orders.
Fourdozen used hierarchical cluster analyses to identify which items were most commonly ordered together to identify customer “profiles.” We relied on elastic net penalized regression analyses to reveal which items most strongly associated with both one-time customers (i.e., missed opportunities) and customers who ordered the most over time (i.e., big spenders). This revealed actual menu item performance in a way that order volume simply cannot, uncovering items that may have made a good impression among a list of items, but disappointed when they arrived at a customer’s home..
Finally, we used time series analysis to better understand how new and return customers varied across waves of the COVID-19 pandemic and how our marketing services were driving orders from new vs. loyal customers.
Results
Menu analyses provide a wealth of information because there are so many variables to look at. With this project, we were able to make a variety of actionable conclusions that contributed to the restaurant’s bottom line.
Identifying and Removing Items that Disappoint
One of our most surprising findings was that a top selling menu item could actually be sabotaging restaurant success. Any restaurant owner when looking over their books would have assumed that this item – an appetizer with a low cost, and a low margin, was a hit. It consistently sold good numbers as an add-on to orders for both lunch and dinner. But our analysis revealed that new customers who ordered this item were much less likely to ever order again. The menu appeal which drew them to order the appetizer was not matched with performance, and meant that Great Wall could be missing out on returning customers.
Targeting Offices with Lunch Specials
The majority of deliveries were sent to houses and apartments rather than hotels and offices, and yet we discovered that orders to offices and hotels are more likely to have a higher grand total than to houses or apartments. It followed that increasing orders to offices could help bolster revenue, and we therefore recommended:
- Creating a special office lunch set combo for groups of 2, 4 and 6 people;
- Offering delivery deals at lunch time for those who have multiple, large orders, to encourage group orders from the office; and
- Conducting old school paper marketing efforts at hotels and offices, such as providing paper menus.
Segmenting Lunchtime Orders
The analysis was able to reveal which orders were specifically “lunch only orders.” We discovered a number of different orders and order combinations that were only ordered during lunchtime and therefore recommended the client double down on these items by creating a lunch box set that included them.
Keeping Vegetarians Coming Back
We also discovered that at this restaurant, vegetarians tend to order once and never again, which led us to identify a number of potential problems with key vegetarian dishes. We suggested making the menu more vegetarian friendly and to revise and improve current vegetarian or vegan options in order to be more competitive and distinctive from competitors.
Detailed Customer Profiles
Lastly and perhaps most interestingly, the order analysis helped us to identify a number of different customer profiles. That is to say, we were able to see the most popular dishes at various times of the day and days of the week, and which items were most often ordered in combination.
Given these insights, we were able to create logical order pairings that we gave internal names such as “the dinner spread,” “the tried and true,” or the “drunken munchies.” Based on this, we were able to make suggestions such as reorganizing the menu to encourage pairings, creating combo meals, running promotions, and improving the overall efficiency of the kitchen.
