Space to Watch: Food & Beverage Automation

Written by Bogdan Cristei of Shack15 Ventures and Tim Yang of Tyche Partners; they are deep tech investors based out of the San Francisco…

Bogdan Cristei

24 May 2023 — 12 min read

Generated with DALL-E

Written by Bogdan Cristei of Shack15 Ventures and Tim Yang of Tyche Partners; they are deep tech investors based out of the San Francisco Bay Area.

Summary

During the pandemic, restaurants have heavily digitized the payment and ordering process, both on-line and in-store. However, we do not see any material changes in the industry related to the food preparation process itself — partly because the chef, cooks and staff are very difficult to replace or augment through automation. Because more and more customers are shifting to takeout orders, we think automating the food preparation process makes sense. The most suitable categories for automation include fast food, quick service restaurants, pizza and the like.

There are four main types of approaches:

Enhancing the existing meal preparation process (eg Miso Robotics and Hyphen)
A new way of cooking certain foods (eg Picnic)
Central kitchen (eg Cloud Kitchen, Kitchen United)
Better vending machines — (eg Xook)

The top four key pain points are:

Quality — automation has to provide the same quality of food
Customer reach — automated onsite food preparation could increase customer reach
Deployment difficulty — cleaning, restocking, down time, and meal preparation time
Scalability — the automated solution needs to be easy to scale

Topics covered include taking the ‘robot foodtech’ idea to the limit, effectively arriving at ‘better vending machines’. The top four takeaways are:

Moat — distribution, physical footprint, and access and leverage with the retailers dominates over economics and tech.
Distribution — most likely a better vending machine as a distribution channel is unlikely to work.
Venture approach — have there been venturable businesses here? The answer is no, not yet.
Conclusion — we cannot think of examples that are particularly great businesses to bet on as VCs.

Table of Contents

Introduction
The Current State of Automation
Examples
Key Pain Points
Back of the Envelope Market Sizing Exercises
How Much do Things Cost?
How Food and Labor Costs are Calculated In The Restaurant Industry
What Are the Ranges?
Understanding “Prime” Costs
Typical Restaurant Revenue Breakdown
Overview of Various Kitchen Business Models
Category Takeaways
Better Vending Machines ?
Examples of Companies in the Space
How About Ghost Kitchens?
Other Examples of Companies
Further Reading
Generative Al for Food
AMRs for Serving Food / Robot Servers
References

1. Introduction

Consumer behavior is changing, and there are labor shortages created by the Covid pandemic. According to WSJ, more people are now doing takeout orders at both fast food and full service restaurants: “Americans Are Gobbling Up Takeout Food. Restaurants Bet That Won’t Change” “America’s biggest chains test digital-only restaurants and more drive-throughs, gambling that heightened consumer demand for food to go will last well beyond the pandemic”

This trend could result in three main implications for venture capital investments:

Format — restaurants to take on new formats in order to fulfill takeout orders
Throughput — more digitization and automation to increase throughput
Recipes — new types of food and recipes that are more suitable for takeout orders

As it relates to this write-up, we will primarily focus on the enabling technologies such as robotics, automation, machine learning and AI that could enable the first two trend changes, namely throughput and format. The current restaurant-tech landscape can be divided into four quadrants, as presented by Zenput below: back of house, front of house, operations execution and point of sale.

During the pandemic, restaurants have heavily digitized the payment and ordering process, both on-line and in-store. For example, Toast — a restaurant SaaS leader — saw exponential growth and adoption in 2020 and 2021. Number of locations increased from 1,000 locations in 2015 to 48,000 locations in 2020. Toast, Inc is a “cloud-based restaurant management software” company based in Boston, Massachusetts. The company provides an all-in-one point of sale system built on the Android operating system.

However, we do not see any material changes in the industry related to the food preparation process itself — partly because the chef, cooks and staff are very difficult to replace or augment through automation.

On the other hand, most industry leaders say baristas and drive-through operators are replaceable, since here we have a combination of controlled environment, repetitive task, usually no need to be creative and less degrees of freedom compared to food operations. See “Restaurant Hosts, Baristas, and Drive-Through Operators at Greatest Risk of Robotic Automation” [4]

2. The Current State of Automation

Automating the food preparation process makes sense, since more and more customers are shifting to takeout orders. Automation may get increased traction, especially as higher throughout and labor savings get more interesting. The most suitable categories for automation include fast food, quick service restaurants, pizza and the like — all these can be cooked based on simple recipes. There are four main types of approaches:

1) Enhancing the existing meal preparation process — keep everything as is and replace human labor via automation; for example, automated dispensing machines at Chipotle, robotic arms to cook fries, etc… (also see Miso Robotics and Hyphen)

2) A new way of cooking certain foods — a totally new way of making food, such as a complete automated pizza cooking system (from dough to pizza, see Picnic as example)

3) Central kitchen — A centralized place to cook a variety of foods; highly efficient production units without a storefront that are optimized for delivery (examples include Cloud Kitchen, Kitchen United)

4) Better vending machines — machines that cook pre-prepared food (see Xook as an example)

Examples

Key Pain Points

We spoke with a handful of food automation companies; below we are including a summary of key pain points as they roll out their solutions to end customers.

Quality — automation has to provide the same quality of food in a stable manner at faster speeds

Customer reach — automated onsite food preparation could increase customer reach. For example, automated pizza cooking in an amusement park.

Deployment difficulty — cleaning, restocking, down time, and meal preparation time are key factors in deployment

Scalability — the automated solution needs to be easy to scale (i.e. food agnostic, easy to operate and maintain, reasonable size that fits into the space, etc…)

For now, central kitchens and enhancing the existing meal preparation process would best fit the venture capital financing model. We believe these areas will see more VC investment as restaurants roll out more automated process to increase throughput and efficiency as the “to go” trends mentioned above increase in popularity. A quick overview of scoring included below with the following as grades: 1 → Good; 0 → Neutral; -1 → Not Great

3. Back of the Envelope Market Sizing Exercises

Just to get a feel for the opportunity, we run some numbers below for the United States — note that these are far from exact numbers.

Fast Food and QSR Employees [US]

Assuming there are ~ 5M workers and annual salary per worker is $25k
~ Total wages = 5,000,000 workers * $25,000 / worker = $125B
Assuming 20% are involved in cooking we get $125B *0.2 = $25B
Assuming 1% automation penetration we get $25B * 0.01 = $250M

Fast Food and QSR Locations [US]

Assuming there are around ~200k fast food locations
Assuming 5% of locations have one robotic solution and each arm has a value of $30,000
Then the opportunity size is 200k locations * 0.05 * $30,000 = $300M

Pizza Market [US]

Total available opportunity = number of outlets * revenue
There are about ~500,000 outlets selling pizza in the US
US is about ~1/3 of the global market
deploying a robot at 2500 stations equals roughly 0.5% market penetration

Coffee [US]

There are about ~40k coffee shops in the US
They get about $180k revenue for the first year and $300k revenue in year three
See example of robot coffee baristas here including Briggo, Cafe X, Rozum Café, Monty Cafe, Aabak [1]
Some chains have shut down in the past in order to train baristas once quality of product decreased — see example of Starbucks Shutting Down 7000 Stores [2]

Robotic Servers [US]

Assuming there are ~2 million food servers at $25k annual salary
~ Opportunity = 2M * $25,000 = $50B
Assuming 1% penetration we get $50B * 0.01 = $500M

4. How Much do Things Cost?

See — “What Percentage Should a Restaurant Spend on Payroll?”

How Food and Labor Costs are Calculated In The Restaurant Industry

Food and labor costs are calculated as a percentage of the total volume of sales
If a restaurant does $20,000 per week and the total cost of food and beverages is $7,000 for that week, then the food cost is 35%
If, at the same restaurant, labor (including payroll taxes and benefits) equals $5,000 for the week, then the labor cost is 25 percent
Total prime costs are 60% in this example

What Are the Ranges?

Certain fast food restaurants can achieve labor cost as low as 25%
Table service restaurants will see labor in the 30% to 40% range
Food costs (including beverages) for the restaurant industry run typically from 28% to 35%

Understanding “Prime” Costs

In order to make money in the restaurant business, prime costs should generally be in the 60% to 65% range
How that breaks down between food and labor is less important than achieving a prime cost maximum that produces a satisfactory profit.
If one of the prime costs is in the higher range, the other prime cost must be in the lower range to achieve profitability.

Typical Restaurant Revenue Breakdown

Source: “What Percentage Should a Restaurant Spend on Payroll?”

5. Overview of Various Kitchen Business Models

Source: Ghost Kitchen TBnds and Emerging Technology; Intel.com; [5]

Examples of Global Centralized Food Preparation Companies

Investments in The Virtual Kitchen Space by Company

Source: Virtual Kitchen Investments in U.S. Near $600 Million; Food On Demand; Feb 13, 2020

6. Category Takeaways

Better Vending Machines ?

If we take the ‘robot foodtech’ idea to the limit, we arrive at “Better Vending Machines” as a the main concept, so the question we need to ask here is “what would the business would look like if the tech were as dumb and commoditized as vending machines?”

Success hinges on preferred location access (static rent vs % of revenue) and extremely efficient ops over a tight geographic area.
While gross margins on product are high (>50%), costs are dominated by how many machines can your crew serve per day/week/time-unit.
Crews often talk about how they preferred one location over another because of how pallets and their trucks were loaded, enabling them to more quickly when servicing a given vending machine (i.e., collect cash, replenish product, remove expired) and go to the next location.
Small operators in high-value locations were kicked out by bigger players with more efficient operations

What if the vending machine products were wholly owned by the manufacturer, and there was no other way to obtain the product or substitute ?

Seems like distribution, physical footprint, and access and leverage with the retailers dominates over economics and tech. We cannot think of examples that are particularly great businesses to bet on as VCs.

Have there been venturable businesses here — and if so how?

It really feels like the answer is no, not yet.

Pain Points

If food prep takes longer than 1–2 minutes, the solution is most likely impractical as convenience is key.
It is important that these machines are never down, they are easy to use, and food options are abundant.
The system needs to be reliable; important factors here include developing your own in-house robotic systems with algorithms written by an internal team, so that execution can be fast. It is important that the algorithms be under the team’s control.
These systems will most likely need to be refilled once per day at most locations — this presents an important logistics problem
The short term moat could be: deals and ability to quickly scale across number of highly profitable locations
The long term moat could be: “food court in a box” — where you own the kiosk as a service brand, perhaps a brand like Keurig that is part of the “office experience”
All the companies in the space are going after locations with pre-existing demand and labor shortages, so the competition is fierce for a limited number of locations
Not easy to make something like this go viral

Most likely a better vending machine as a distribution channel is unlikely to work; vending machines work best with products that are suitable to be distributed via vending machine — see Pop Mart example below.

Examples of Companies in the Space

How About Ghost Kitchens?

“Some experts predict that the ghost kitchen business model will dominate the industry. It’s not hard to see why. The lower overhead, rich customer data, and ability to test menus and brands rapidly make ghost kitchens an innovative incubator for the industry.” [5]
“The social dimension of breaking bread in a sit-down restaurant remains priceless. It seems more likely that brick-and-mortar restaurants and ghost kitchens will develop a symbiotic business model that helps them both succeed.” [5]
This could also be a growing bubble and the investments might be pitching forward over those proverbial skis. John Gordon, the principal restaurant analyst and consultant at Pacific Management Consulting Group, said there are a lot of $1 million kitchens going up without the orders to support them or eyeballs to see them. [11]
Interesting Bloomberg article: Ghost Kitchens Are Dying and Nobody Noticed

7. Other Examples of Companies

Note: web links have been generated with ChatGPT, some may be incorrect

Picnic: https://picnic.app/
Hyper: https://www.hyperpizza.it/
Yokai Express: https://www.yokaiexpress.com/
Hyphen: https://www.hyphenpizza.com/
Briggo: https://www.briggo.com/
CafeX: https://www.cafex.com/
Rozum Cafe: https://rozumcafe.com/
Monty Cafe: https://www.montycafe.com/
Aabak: https://aabak.com/
SJW Robotics: https://www.sjwrobotics.com/
One Stop Kitchen (OSK): https://onestopkitchen.net/
Kitopi: https://www.kitopi.com/
Local Kitchens: https://www.localkitchens.com/
All Day Kitchens: https://alldaykitchens.com/
Kitchen United: https://kitchenunited.com/
Afresh: https://www.afreshtechnologies.com/
Rotender: https://www.rotender.com/
Cecilia.ai: https://cecilia.ai/
The Mini Bakery: https://www.theminibakery.com/
Miso Robotics: https://misorobotics.com/
Kitchen Robotics: https://www.kitchenrobotics.co/
True Bird: https://www.truebird.com/
Smile Robotics: https://smile.co/
Solato: https://www.solato.co.uk/
Milkit: https://milkit.io/
Peanut Robotics: https://www.peanutrobotics.com/
Knitscope: https://www.knitscope.com/
SoftBank Robotics: https://www.softbankrobotics.com/
Makr Shakr: https://www.makrshakr.com/
Bear Robotics: https://bearrobotics.com/
Milky: https://www.milky.ai/

8) Further Reading

Generative AI for Food

“Generative Al has the potential to revolutionize the way restaurants are run. Restaurants can use generative Al to create new and innovative recipes, analyze food trends, and optimize their menu offerings based on consumer preferences and dietary restrictions. It can analyze sales data and help restaurants optimize their inventory management processes by predicting future demand and ordering the right amount of ingredients at the right time, which can reduce waste and save money.” [13] PitchBook Q4 2022 Foodtech Report March 7, 2023

AMRs for Serving Food / Robot Servers

Each robot can serve between 15–30 tables; makes the experience cooler but not really moving the needle on savings so tough to see this being “venturable” See example below — Haidilao Hotpot (Bear Robotics — 1 AMR)