A hypothetical restaurant registers a fifteen item menu for distribution via Kitopi. There’s value on both sides of the transaction. The restaurant provides the following: brand, menu, and cooking intelligence. Brand experiences the least augmentation. A customer orders from that restaurant on the delivery app just as if they were ordering from the physical storefront. The restaurant’s logo is stamped onto every item leaving the kitchen and is intrinsically linked to the production of the final product. A restaurant on Kitopi tailors the menu for core delivery items; compressing a forty item menu to fifteen. Products that aren’t differentiated, are distracting, or act as fillers are jettisoned. A restaurant with ten burgers may choose five or even one.

Taking this to a greater extent provides clarity on a broader shift in the digital restaurant space, particularly in the case of virtual food halls. At these halls, customers can order from any of sixty or so brands in a single order. The sixty number isn’t pulled out of a hat, that’s how many brands operate out of a single 2,000 square foot Kitopi - the Dubai based cloud kitchen company - hub. You can get your burgers from one place and your fries from another. This breaks down the friction that previously protected menus. Much of restaurant menus are filler to cater optionality. and subjective preference. Basically, a customer might order guacamole because they prefer guac even if the Ceviche is “better”. Consumers aren’t always ordering specialties so to speak. Even at a great restaurant only some fraction, a small fraction, is objectively superior.

In the case of a virtual food hall, however, consumers don’t have to settle for subpar starters. Instead, the best starters from each restaurant are delivered in one order from the food hall. A customer in the mood for both sushi and brusell sprouts is able to get the rolls from one spot and the sprout appetizer from another. Reduced switching costs act as a forcing function to hone in on core offerings. There are drawbacks to this optionality, however. Selecting items from different restaurants is uncertain.

Are consumers going to have enough effort to track a Brussels sprout brand? Will a starter from one spot match the flavor profile of a different restaurant’s entrée? Envision a complete disaggregation - items from sixty brands are listed in one menu. Starters are lumped. Mains and desserts too. Bondi Sushi’s 'Shrimp Avocado Rolls' is sitting alongside a 'Kick My Boss - Spicy Burger' from Rock House Sliders. This is basically a micro-branded Cheesecake Factory - home of what seems like America’s and, most certainly the World’s, longest menu running 250 items. The menu has a small number of branded items of the type you’d see in this virtual food hall such as the Oreo Dream Extreme Cheesecake. So too, it spans cuisines from French to Korean to Mexican and carries breakfast, lunch, and dinner options. Flipping its pages inspires the thought that its ten different restaurants lumped together.

Assume, however, that the average delivery consumer is overwhelmed by this optionality leading to suboptimal ordering. One solution is bundled menus. Delivery apps or Kitopi itself could coordinate their offerings into a number of select menus. Some of Kitopi’s kitchens produce for one hundred distinct restaurants. Spotting each restaurant at twenty items opens up over one thousand unique options for groupings of ten items into boutique menus that pick and choose the best combinations. These menus could be A/B tested across audiences to optimize satisfaction. Taking this further, the quality of 'best' can be analyzed from two angles for middle and low priced restaurants. One is what’s most popular, the other is what any individual’s subjective opinion is. For the latter, Kitopi would be able to deploy recommendation algorithm style strategies to consumer tailored menus. Customer preference for starter combinations and the like would be taken into account. Kitopi would merely be doing what every other content-based platform does - Netflix, YouTube, Amazon - by tailoring content to preference. 

There’s potential for the creator economy angle as well. The “Travis Scott Meal” at McDonald's gave fans of the music artist the ability to order as he does. The meal led to the strongest revenue month in a decade for the Golden Arches and caused a shortage in chopped lettuce. Scott’s offering didn’t consist of unique items, but rather just branded an order: Quarter Pounder with Cheese, Fries, Sprite. If Kitopi can create their own menus, creators could just as easily do the same. Any branded individual could create their 'branded' menu in the same vein as Scott by selecting from the two thousand possible items. 

From the restaurant side, food itemization is akin to licensing out their content library individually rather than as a whole. Kitopi may only want to serve Samosas and Butter Chicken from an Indian restaurant because those are deemed the highest traction items. Other spots do Dosas better. Maybe it won’t be this narrow and restaurants will do ten dishes with enough uniqueness and taste value to get picked up. Restaurants doing one thing well is an indicator of the quality of everything else, but how strong an indicator is the question? Will that Samosa restaurant start focusing on only Samosas to make them their key offering? After all, they could license out these Samosas to hundreds of kitchens to serve a broad customer base. Restaurants get high leverage on selling one item well out of the distributed, scaled kitchen network. Without delivery, there aren’t enough local customers to support a Samosa only shop. In the global network, however, there are millions in the Samosa market. If that restaurant gets a ten percent kickback on each item sold, they have the potential to stand up a streamlined, scaled business on the back of their 'best' offering. The outcome will likely be analogous to grocery store aisles carrying a few different types of Samosas. 

Once a restaurant comes up with a promising item, develops the recipe, and deploys it to the Kitopi base, there are a few options. One is to sit back and collect licensing fees riding that Samosa money until a better one comes onto the market. Inevitably, however, one will come onto the market. This requires restaurants to operate like software companies, the Samosa needs to be versioned, iterated upon, and customized to stay ahead of the curve. This is part of the reason why restaurants will become narrow. Massive distribution means one offering is enough of a business. Those who hone in will make that item at a level far greater than competitors. That’s all to say there may very well be an exclusively brussel sprout brand in Kitopi’s kitchens soon.

A similar inflection event is happening in food. Major restaurant brands are outsourcing their menus to manufacturing operations as Apple outsourced iPhone designs to Foxconn (the famous electronics mega-factory). For the likes of Papa Johns and Nathan’s Famous Hotdogs, delivery bags wear the tag not of “Made in China” but “Made in Kitopi”. They could just as easily read “Made in a Ghost Kitchen”, for Kitopi is one of many restaurant companies using cloud kitchens to service orders for app pickup. In five years since its founding, Kitopi has become the leading brand in the Middle Eastern market by providing rapid global distribution with serviceable quality.  

The differences between Kitopi and Foxconn are their operational time horizons and modes of contracting. Unlike Foxconn and other Chinese factories which have the latency of shipping products back to distribution centers,  Kitopi liscenses out a menu and services delivery directly to drivers picking up from its facilities. Production is based on demand and can be rapidly shut off so the risk for restaurants and Kitopi is exceedingly low. Kitpoi receives an order from delivery apps for Papa Johns, cooks the pizza, and hands off the order to a driver. Ghost kitchens maximize just-in-time production. Papa Johns receives a ten-percent kickback for providing Kitopi the menu and cooking intelligence to sling their pies. Apple and Papa Johns both get irreplicablly efficient assembly of their product, albeit with disparate incentive models.

Kitopi stays ahead of the curve to launch new brands quickly. When a restaurant approaches Kitopi to explore a partnership, the “Supply Chain Growth” team has already produced an outline of how to port over the restaurants’ menu. By day five of conversations the menu is approved. How does Kitopi do it? The answer lies in streamlining a restaurant’s offerings for 80% coverage of existing SKUs (read: ingredients) from the company’s supply chain. This not only allows a drastic reduction in order complexity but also superior pricing. Just by getting onto Kitopi’s supply chain, a restaurant’s costs are reduced by 10% on ingredients. The more hamburger buns Kitopi orders the cheaper. “Secret Sauces” and the like that are irreplaceable constitute the marginal 20% that are special ordered. Kitchen workers receive less than a week of training before a brand is available to consumers. That’s a two week turnaround from first conversation to being able to order “Bondi Sushi” (the Miami based Sushi bar) on Suadi Arabian delivery apps. 

It’s reasonable to expect that the sushi served in Dubai will not taste exactly like its counterpart along the coast of South Florida. Say the Sushi is 90% of the taste, or even 75%. At the end of the day, however, consumers still taste the core essence of a restaurant sitting eight thousand miles away. That’s the power, and is standard for global brands. Lulu Lemons’ first batch of product runs used higher quality fabrics than those since moving to China. Reducing quality standardizes the product for tens of millions of consumers who want to experience the “essence of the brand”. Just as leggings can be good enough, so can food. Especially, when that food is being produced at an error rate of 0.2% for delivery compared to the restaurant standard of 5%. Kitopi presents reliability and “good-enough” quality for a lower cost.

Streamlining to unlock speed for restaurants extends past the startup-phase to cooking and assembly. Take the example of a Salmon dish requiring six steps, three kitchen appliances, and twelve minutes to prepare at an American restaurant. Kitopi reduces down to three steps, a single smart oven, and seven minutes. That’s a forty percent reduction in prep-time with fewer appliance choke points. As a rule, food is cooked and ready for pick up in under nine minutes. Kitopi’s reached the heights of software-enabled operational efficiency. The company has a team of over a hundred engineers in Krakow, Poland whose job is to analyze and optimize the process. Everything is tracked and managed - down to each employee washing their hands for twenty seconds. The Kitchen Operating System (KOS) consolidates orders from delivery apps, splits them on an item-by-item basis and deploys them to stations throughout a kitchen. Cooks then work on these items in parallel at different workflow stations. Items are given barcodes and placed on a central conveyor inching them to their assembly location. Drivers are integrated into this system by prioritizing - or delaying - orders based on arrival time. Parallel processing combined with drivers-in-loop ensures hot-and-ready food for pick up. What’s compromised on taste authenticity is made up for by food that’s sanitary, temperate, and expedient; no cold soggy fries that’ve been sitting while the steak in the steak frites is cooked.

Scale is the linchpin to the ghost kitchen model and manufacturing generally. It doesn’t matter how cheaply a Chinese factory can produce Nike products if they can’t scale up to service millions of customers losslessly. Machinery and processes get ironed out as production ramps up. A single 2000 square foot Kitopi facility churns out 3,000 daily orders over sixteen operational hours. That’s almost 200 per hour. The kitchen is live to deliver them food whenever a customer is awake and hungry. Restaurants that handle their own delivery keep regular kitchen hours and negotiate online orders with in-person diners to their mutual detriment. Outsourcing delivery gives restaurants “always-on” availability and capacity to service their consumer bases’ needs. Restaurants on Kitopi’s network experience a 150% increase in delivery revenue. A brand isn’t just able to scale via one facility through Kitopi, but rather can deploy their menu to every kitchen in the distributed network. Kitopi has a presence in the UAE, Saudia Arabia, Bahrain, Kuwait, Qatar - the bulk of the Middle East. This means that within two weeks, the entire urban Middle Eastern population can get South Beach’s finest sushi to their door in half-an-hour. 

The transition to cloud kitchens, exemplified by Kitopi, marks a pivotal shift in the food industry, mirroring outsourcing trends observed in manufacturing. Kitopi’s model demonstrates how technological innovation and operational optimization can redefine traditional business models, making quality food more accessible across vast distances with minimal compromise on taste. While the nuances of taste and local authenticity may vary, the ability to deliver the essence of a brand’s culinary experience to a global audience underscores the power of this shift. As cloud kitchens continue to grow, they pave the way for a future where the geographical boundaries of culinary experiences are further blurred, making global cuisines more accessible than ever before. This not only signifies a leap for the food industry but also poses intriguing questions about the future of dining and culinary traditions in an increasingly interconnected world.

Though evolving more rapidly, the space is following a similar trajectory to Amazon’s dropshipping phase. Dropshipping took off on the back of a mature fulfillment offering from Amazon and the growth of consumer trust in online shopping. A standard process emerged: select a niche product, find an overseas supplier on Aliexpress (or similar), slap a brand on the whitelabeled product, and get fulfilled by Amazon’s warehousing and delivery chain. Over five thousand results spit back on a search for “iPhone charger” - that's the outcome of dropshipping. Most of these results come from the same factory in maybe a different color with a different brand. Like Doordash, or more accurately the other way around, Amazon’s cracked down on the practice. Addition of new products is limited by strict regulations that emphasize uniqueness and tangibility. 

Bloat isn’t limited to phone chargers. Thousands of burger brands serve effectively the same sandwich under a different name. Brand is the strategy. To mirror the dropshipping chain: Doordash is Amazon - the search platform; ghost kitchens are Chinese factories that produce the generic products for consumer branding; and dropshippers are the virtual brands posting menus under their logo. NextBite serves as a case and point. The virtual brand partner of iHop and Noah Schnapp creates simple, generic food fertile for a logo sticker to be slapped on. Five of the same chicken sandwiches with five different storefronts. Each of which are pushed out on Doordash for distribution. This is more or less dropshipping. 

Missing from the comparison is an AliExpress analogue. What does the AliExpress of food look like then? The site functions as a marketplace connecting consumers directly to wholesalers and manufacturers with global distribution. Cutting out the middleman, AliExpress enables customers to create their own “brands” on top of an existing product base. In the food market, however, brands are vertically integrated - controlling production and marketing. Even for ghost kitchens, NextBite runs both the kitchen and the brands. In the AliExpress model they would do one or the other. Not both. 

Functionally, this platform would connect customers with an existing food base ready for production. A food base of finished items; that means burgers rather than buns and chicken tenders rather than raw chicken breast. Unlike AliExpress, however, products aren’t fabricated at once and shipped to an Amazon fulfillment warehouse. Rather, this model would work more like a contract.

Take an example of someone who wants to start a restaurant brand with no existing cooking knowledge. Their first step is to find the type of food for their brand. They scroll TikTok, examine trends, and decide on some niche items. Next, they need to find a producer since they don’t have any knowledge of recipe creation or how to create any scale around a food item. They go to the food version of AliExpress. This site lists tens of thousands of food items, ready for “whitelabeling” and distribution to customers. Each item is a generic version of some distinct food item offered in house for production by ghost kitchens. 

Another piece is that each Ghost Kitchen has a unique delivery network, due to the locations of the physical kitchens and delivery radiuses to get food to consumers’ doors within thirty minutes. The brand “founder” has the decision of which geographies to distribute to and to what capacity. This allows fine grained control over the market. For example, maybe there’s a food trend in five specific zip codes. In each of these zip codes, then, the founder could select the available ghost kitchens and none of the others in the network. 

Since orders are serviced from a single kitchen, the founder will have to find a match for their products all from one kitchen. Kitchens are reaching a capacity of thousands of unique items, so this is more feasible than it might sound. Perhaps networks will specialize around certain types of cuisines - for example a Chinese food network - or workflows such as “bowl style” restaurants. Realistically, such specializaiton is only immediately possible in urban centers where there’s enough order flow to support different cuisines. In suburban geographies it’s less feasible, but perhaps a ghost kitchen network from China for example will have locations in demographics with  higher Chinese cuisine preference. 

Continuing along, this founder hones in on five or so items they’d like to produce. For each, they tweak a few elements - subbing sauces in for those they think fit better and the like. Next they select the production volume to detmermine pricing and order management for ingredients. For example, they might buy five hundred burgers per week. They will have to pay for all these burgers (or at least a fraction of them) upfront regardless of if they are actually produced. The kitchen needs the cash to order ingredients as well as to account for the employee time spent. It’s the same with AliExpress - you have to pay for all five hundred lava lamps even if you don’t sell them.

One notable step is that the founder will have to find packaging in line with their brand that’s sent to the fulfillment kitchen. Once purchased, the founder puts their restaurant up on Doordash and its in service. Customer’s orders will be forwarded to the fulfillment kitchen, prepared, and handed off to the driver. A full service restaurant is created and it's possible these burgers are being delivered in Ohio while the founder is in Beijing or even the other way around. 

In this system it's important to recognize the value provided at each step. The ghost kitchens supply the product. They develop the recipes, train the staff on production, and operate the cooking facilities. AliExpress, or in this case a food analogue, provides access to a global marketplace of consumers looking to purchase. It connects them with customers who will do the marketing. Kitchens don’t have to worry about running brands or interacting with end customers - they just do the food fulfillment. NextBite picks whichever element they do best - kitchen production or brand operation.

The restaurant industry's digital transformation, marked by phases of discoverability and orderability facilitated by the internet and delivery platforms like DoorDash and UberEats, has now entered a new era with ghost kitchens. This "scalability act" redefines the essence of being a restaurant in the digital age by separating the creation of recipes from their replication. Similar to Content Delivery Networks (CDN) for web content, ghost kitchens allow restaurants to upload their "recipes" to a network, where orders are cooked and delivered by the nearest kitchen. This model offers unprecedented scalability and significantly lowers the barrier to entry for launching new restaurant brands.

Digital restaurant brands on platforms like UberEats have seen rapid growth indicating a burgeoning market. Initially dominated by corporate entities, the evolving landscape now invites a wider array of participants, echoing the early internet's "dot-com" boom. However, this ease of entry also presents challenges, such as market saturation and potential quality dilution. Platforms and ghost kitchen networks may need to enforce stringent criteria for new entries to maintain quality and diversity.

In effect, ghost kitchens are reshaping the culinary world by enabling rapid scalability and innovation, much like the internet did for information. As the industry navigates this transformation, it will need to balance the proliferation of new brands with the maintenance of culinary standards to enrich rather than dilute dining culture.

Uniquely, interacting with the internet is universally accessible. Anyone with some coding skills, or even none at all (via website builders like Wix), can launch a site from the couch. Barriers to entry are slim to pushing information out into the world. Though not quite as simple as deploying code to a server, launching a digital restaurant is magnitudes simpler than standing up a physical store; Physical infrastructure is abstracted away.

Infrastructure abstraction lowers barriers to entry leading to a surge of restaurant brand origination.  Launching a “restaurant” is reduced down to a three step process: develop a menu, deploy the menu to a Ghost Kitchen Network, and post the “restaurant” on delivery apps. Restaurant “founders” don’t have to ever cook the food or even try the food they are pushing out to customers. Written instructions are all that’s required. Just as “dot-coms” flooded the early internet, a stampede of digital brands onto delivery app platforms is the inevitable outcome.  

The deluge’s already begun. UberEats’ digital brand volume quadrupled in the two years between 2021 and 2023 charging from ten to over forty thousand. For context, the delivery app hosts a total of nine-hundred-thousand restaurants. That means digital brands compose only about four percent of the total brand base and are the fastest growing segment on the platform. There’s a lot of room here.

Digital brands popping up so far on the app have thus far been primarily corporate offshoots. That’s to say that they’re owned and operated by “Digital Brand Partners” such as Virtual Dining Concepts. VDC operates the popular Mr.Beast Burger - a burger restaurant promoted by the YouTube megastar of the same name. Digital Brand Partners often run a portfolio of dozens of restaurants either built in-house or in collaboration with a public face or existing brand.

Today, distributing via a Ghost Kitchen Network requires some level of corporate or brand presence. Over the next few years, however, as overhead to launch continues to fall, so will the size of the entity required to do so. As of now, Kitopi, the Dubai based digital restaurant company mentioned a priori, operates up to seventy restaurants out of a single two thousand square foot kitchen. Each of these brands is an established global presence liscenced out for distribution to Kitopi. 

Uber founder Travis Kalanick’s entry into the space Cloud Kitchens, on the other hand, offers a distributed kitchen as a service model providing restaurant brands with space for rent in commercial kitchens. Restaurants provide the staff, Cloud Kitchens the space. This is a different model to Kitopi and peers who staff and run their own kitchens. Kalanick’s approach is more of a real estate play, but the restaurants in his space are still “ghost kitchens”. 

The “scalability act” opens the opportunity for merging the two models to create a true ghost kitchen network that's globally integrated, easily deployed too, and equipped to serve manifold cuisines. This new company would operate a network of kitchens that each have a capacity of hundreds of brands. Restaurant “founders” would be able to purchase space like they do server space. These “founders” could be anyone as they’d be able to start a restaurant by uploading recipes online, purchasing a “spot” in the network, and gaining instant distribution for their brand. If the brand takes off, they simply buy a larger “spot” and distribution out of more kitchens. Growth of digital brands will be rapid in light of such reduced friction. 

When this future arrives, problems of the same origin UberEats faced in Spring of 2023 will become manifold. During this time, eight thousand digital brands got the boot for being effective spam. The majority of these deletions were thinly veiled clones of each other - re-listing the same menu under a different brand. When ease of deployment rises, volume rockets and median quality deteriorates. Filters are already being built to protect from spam. Doordash, for example, requires restaurants to submit five unique items to be considered a “restaurant”. Spam aside there are going to be a huge number of “food white-lables” that, just like the internet, are a part of a mass of really bad content (read: restaurants) saturating the market. 

During the “discoverability act” in the late nineties, the internet served only as a remote directory. When visiting a new city, a traveler could Google “restaurants in Kyoto” and have a directory webpage pop up with nearby options. The “orderability act” arrived in the following decade. Brought on by companies such as Doordash, UberEats, and the like, this second phase serviced online discovery, ordering, and delivery to consumer’s doors. Delivery platforms moved down the chain from restaurant discovery to individual food delivery. Though orderability marked progress, delivery still represented an external interaction with a restaurant - the workings of the kitchen remained the same. 

Restaurants have serviced takeout for over a century. All that’s changed is where these take-out orders are coming from and who’s picking them up. Rather than calling in, diners click check-out on a delivery app. Rather than getting picked up by the ultimate consumer, take-out orders are picked up by a driver. Products such as OrderMark aggregate tickets from different delivery apps into one system anyways, so the internal workings of the kitchen hardly change at all. A ticket comes in, the order is cooked, and it's sent out to the customer whether they are in a booth or sitting at home on their couch. 

Restaurant digitization’s third act, however, is creating a structural change in the restaurant industry that will do for food much of what the internet did for information. The “scalability act” redefines what it means to be a “restaurant”. With the establishment of ghost kitchen networks, a restaurant is no longer tied to the replication of its food, but instead only its creation. 

Creation is the act of producing the recipe for an item - the act of coming up with something new. It’s the process of developing food. Replication is the procedure of cooking that item over and over to serve customer orders. The goal is the opposite of creation: It’s not to create something new, but rather to create something as close as possible to the menu’s listed item. A good creation is most different. A good replication is most same.

Ghost kitchens take care of the replication as an analogue to how Content Delivery Networks (CDN) handle web distribution. When a company deploys a website to the internet, the “recipe” for serving the site is uploaded to the CDN’s network of servers. Whenever a user types in the website’s address - in effect “ordering” its “delivery” - the network server closest to that user replicates the page based on the downloaded recipe to deliver said website to the user. 

Made up of hundreds of kitchens spread out around the world, Ghost Kitchen Networks (GKN) operate equivalently. A restaurant who wants to serve on the network uploads their “recipe” - their creation. When a user orders on a delivery-app, the ghost kitchen nearest the customer cooks up the order from the instruction set and hands it off for delivery. CDN’s send internet packets. GKN’s send DoorDash packets. 

Ghost kitchens remove the means of production from the restaurant. They detach the kitchen replication process. So too, restaurant’s are provided with an effective global scale. Ghost kitchen brands like Kitopi, the five year old juggernaut out of Dubai, operates hundreds of kitchens spanning a dozen cities. Each of these kitchens services three thousand daily orders, providing any given restaurant with a hypothetical capacity of a million orders a day - instantly. To deploy a restaurant into the kitchen network requires two weeks. It often takes longer to launch a website.

To operate Mise En Place at scale, Kitopi leverages a system of central production units that feed into smaller, localized kitchens. The French phrase for “everything in its place” empowers precise kitchen actions via a meticulous standard for ingredient preparation and organization. Handling the bulk of labor- and time-intensive operations, Kitopi’s “hub kitchens” are large warehouses equipped for this prep: marinating meats, mixing sauces, chopping vegetables, and partially cooking items. Each hub delivers to fifteen or so “satellite kitchens” for finishing and assembly. Centralized prep minimizes slow-down. Satellite kitchens don’t set-up or clean-up, don’t manage ingredients, and don’t make large quantities. Workers are, instead, freed up to maximize order throughput without regard to the time-intensive steps bogging down integrated kitchens.

In-kitchen workflow is divisible into three segments: order management, assembly, and kitchen layout. An order might look like: Steak Frites, Soppressata Pizza, and Kid’s Quesadillas. Upon receipt, this order is disaggregated and sent to separate stations on the line. Satellites are organized by workflow not by brand, so the pizza is sent to the pizza station rather than the “Sicily’s Finest” station. In action, this means that one oven might be used for five restaurants - who collectively benefit from the streamlining. Despite their disparate requirements, all items are handled individually, timed for instantaneous completion, and processed in parallel. The quesadilla finishes at the same time as the steak - and both finish in under eight minutes.

First, what are virtual food halls? Operated by brands such as Kitopi in the Middle East, Rebel Foods in India, and Byte Kitchens in Northern California, virtual food halls are ghost kitchens that produce for a number of different restaurants. That’s to say between five and seventy completely unique brands are operated and distributed out of a single two thousand square foot kitchen. One brand might be Shake Shack while another other is a one-off sushi spot. Critically, however, both the ShackBurger and the Spicy Tuna Roll can be delivered in one DoorDash order. On-app, the items are presented as though they are coming from the same restaurant because, in effect, they are - or at least the same kitchen. The result is thousands of food items from dozens of different cuisines made available for delivery to a consumer’s house in under thirty minutes.

Now, what is the “iPad Kid”? To model the phenomenon requires the creation of a hypothetical family of four: parents, an older daughter, and a younger son. It’s Friday night and this family’s in the car headed to dinner. The son lights the match. “Mom, I want to play on my iPad.” Mom is faced with two options: let the young man enjoy his Flappy Bird or the option she chooses. “I’ve said no five times today. N-O spells NO.” This young man isn’t too pleased, and he fights back. Mom and son go back-and-forth, back-and-forth in a dance well-practiced but never perfected. A quick capitulation. Junior makes Level 5 in Flappy Bird before they reach the restaurant.

This encounter embodies a classic intra-family dynamic: differing preferences that lead to the path of temporary least resistance. If the son doesn’t get his iPad - four parties are discontent. Dad is mad because he was distracted while driving. Mom is irritated because her son tells her “she stinks”. The older sister is upset because she now has to deal with her pouting brother, and her pouting brother is upset for self-evident reasons. But he does get the iPad under the settling stipulation that it's only for the car ride.

No strangers to compromise, this family went through a similar back-and-forth when deciding where they would go eat on that Friday night. Dad wanted a burger and a beer after a week of work, the older sister wanted to try a trendy new Asian fusion place, and so on. The case of the “iPad Kid” took hold, however, and they settled on the same place they go every Friday. Each is pacified but not satisfied.

A virtual food hall, had it been available during this pre-car ride discussion, would have been the optimal outcome. Providing expansive optionality, the food hall enables each member of the family to get exactly what they want. Friction is non-existent. They each just pass the phone to select the restaurant of their preference. There’s no argument and there’s no settling. Mom doesn’t have to order a black-bean burger that she doesn’t really like because it's the only semi-healthy menu item. Junior doesn’t have to eat a quesadilla when he really wants nuggets. Those weren’t bad options but they weren’t great either. Everyone settled so they would all be ok.

One potential drawback is the element of “cohesion”. It’s easy to imagine a parent saying “We’re all eating from one place because we’re eating as a family”. The concept of “one place” needs to be re-evaluated here. If all the food items come in a bag with one logo, are they really from different places? They’re from the same kitchen after all.

The event of “going out” as a family will fade to an occasion, for whenever it starts to feel like a chore, someone will take out their phone and ask “virtual food hall?”. “Virtual food hall!!” everyone will exclaim in ascension and DoorDash’s CEO will see the order counter on his desk click. Plus one.

Optimized to be zero-waste, Rebel’s kitchen innovation is best illustrated by looking into three machines: the Fryer, the SWAT, and the Rotimatic. You could work a Rebel fryer with one of those claw grabber toys that are basically trash pickers for kids. That’s to say it's a pick-up, drop. Pick-up, drop kind of situation. Once food is dropped into the bubbling oil, whether it be french fries or chicken wings, the fryer gets to work. A camera attached to the top scans the item, uses AI computer-vision to detect its shape, and then triggers a specific temperature. Oil is adjusted to this temp and a timer is set for the ideal cooking. Ding. Pick-up. Repeat. 

Cleverly named, the Size-Weight-Apperance-Temperature (SWAT) machine is the barrier sitting between that fried chicken, as well as every other item produced in a Rebel kitchen, and the customer’s mouth. Each item is placed into the SWAT for data extraction across each of the four metrics by an array of sensors and cameras. If the crispy chicken isn’t crispy enough an alarm sounds. Kitchen employees dispose of the order and their error is logged into the data layer. Everything is logged into the data layer. Rebel operates what it calls the AI Data Brain - a database consisting of every data point logged from their kitchens. Every machine that can be is plugged into this system to be tracked, measured, and optimized for the bottom line. 

The SWAT and the Fryer are just two examples. Rebel’s deployed over forty unique internet-of-things enabled appliances into their digital kitchens. Not included in that list, however, is the Rotimatic. That’s because it's found in over seventy-thousand households instead. Sold since 2018, the Rotimatic takes in water, oil, flour and outputs the beloved Indian flatbread. What’s unique, however, is the first few times that bread won’t taste very good. Users rate their initial Roti on a number of metrics that are plugged into a machine learning algorithm to adjust for a better and better product. Now Rebel. Let’s say a customer likes their pasta extra al-dente. They should be able to rate their noodles in-app so next time the pasta pot will tweak cook time using the same tech as the Rotimatic. Rebel can and will do this, showing its power as a tech innovator first and foremost.

Rebel Foods, on the other hand, takes the “ownership” approach. Rebel operates hundreds of its own ghost kitchens to fulfill its in-house brands such as Faasos and Mandarin Oak. Their Innovation Center conceptualizes, tests, and tweaks new brands for the market. “Internet restaurants” are then integrated into Rebel’s pan-Indian supply chain and pushed out for delivery via the distributed kitchen network. Model two: controlled and scalable. Recently, however, Rebel’s begun to leak into a third model via their partnership with Wendy’s: “the liscensing model”. Two hundred and fifty virtual Wendy’s will be popping up across India over the next few years - all fulfilled by Rebel. Like a traditional franchise, Rebel produces and distributes all the food and Wendy’s gets a royalty percentage on every Baconator sold for liscencing out its menu. 

Champion of the “liscensing model”, however, isn’t headquartered in India or the US. Kitopi is a Dubai based cloud kitchen company taking charge of the fastest growing region for ghost kitchens: the Middle East and North Africa. Outpacing the US in growth by a projected 4% annually, MENA is becoming the center of attention for digital restaurant brands globally. Kitopi’s become the giant in the region through unmatched supply chain and operational efficiency. Whilst Rebel produces seven brands per kitchen, Kitopi produces fifty. That’s fifty unique brands flowing out of a single kitchen. How is that possible? It’s possible because instead of only operating a distributed kitchen network, Kitopi operates distributed kitchen components. In every urban market, Kitopi runs at least one “central kitchen” responsible for sanitizing vegetables, butchering, prepping, mass sauce production, portioning, etc. The products of the central kitchen are then distributed across the city to smaller “finishing kitchens” that assemble the food so it's out the door in eight minutes or less. Yes, fifty brands and eight minutes. Order from any of these brands, and the entire order should be at your door in less than half an hour - all pumped out of the same location. Such efficiency makes Kitopi the gold standard of optimization and is the reason why hundreds of brands are liscensing out their menus to the company for distribution.

During Google’s early days, the product team faced a narrow decision: what shade of blue should the links be? To identify the optimal color the team took a hyper-methodical approach by putting users into buckets representing 2% or so of the total and displaying one of about forty unique shades to them. #1A0DAB produced the highest click yield and so #1A0DAB it was. Over two hundred million dollars of revenue was produced from this exercise in what’s become common practice across the tech industry: A/B testing. That is, deploying numerous iterations of a product in test markets to optimize the approach. 

Rebel Foods’ culinary innovation center might as well have a Google sign across the entryway. Their process of originating and testing new restaurant concepts mirrors what the Silicon Valley behemoth would likely come up with. First step is data. The innovation center leverages machine learning algorithms to detect trends, missed opportunities, and new markets in the food sphere. Once an opportunity is identified, Rebel comes up with a provisional menu optimized to take advantage. The menu along with cooking instructions are then handed off to five or so Chief Delight Officers (general managers) to deploy in their kitchens - five out of hundreds. 

That’s where the process starts. What follows is a period of testing and measuring. Testing and measuring. Rebel will continue to iterate on a new restaurant until one of two outcomes is reached: the menu reaches product market fit or it’s discarded as a null concept. Stipulations for PMF are codified: a delivery app rating above 4.2 and an net promoter score of above 40. Rebel’s hit rate on these metrics is impressive: seven out of twenty restaurants tested during 2018 surpassed them and were deployed to the entire kitchen network. Regional tailoring is leveraged in this model as well, Rebel can code their menus to certain taste profiles depending on where the production kitchen is located. They can A/B test their audience.

As Tesla deploys updates to a car, Rebel deploys software updates to their restaurants. When holidays come around, Rebel’s data scientists identify the special items that will play best. Leveraging their massive presence in the pan-Indian supply chain, they can deploy these concepts within six weeks. It tracks, then, that Rebel’s job board has listings for data engineers, UX designers, and AI researchers. Rebel Foods is a software company first.

Rebel Foods, the decade-old Indian ghost kitchen operator, has developed the largest network of digital restaurants in the world. Rebel’s CEO Jaydeep Barman, however, has a larger goal in mind. He wants to open up ten thousand fully functional restaurants by the end of the year. Over the next set of articles I’m going to dive into the three reasons why Barman’s estimate is not only realistic, but is potentially conservative. In short, why Rebel Foods may become the largest food brand in the world. And why you’ve never heard of them.

First, a brief overview of how Rebel operates. Rebel’s business is first and foremost one of distributed ghost kitchens. What Barman calls a restaurant is really a cloud kitchen with distribution capability for a particular digital brand. So if both my friend and I order Wendy’s to our houses, but the food is cooked in two separate kitchens, that counts as two distinct “restaurants” in Barman’s book. 

Hundreds of such ghost kitchens are peppered across India. These kitchens aren’t assigned to a single restaurant. Instead, they fulfill a number of different brands. So that spot making my Pretzel Baconator might also be making Spaghetti Pomodoro sold under a different logo. We’ll call this the “virtual dining hall” model. Like a dining hall, Rebel’s kitchens are outfitted with all the equipment to make cuisine ranging from wok based stir frys to pan cooked pizzas, for that’s exactly what they do. Rebel runs six or so separate restaurants from a single kitchen. That means launching one thousand “restaurants” requires less than two hundred distinct fulfillment locations. This “one-kitchen-multiple-brands” approach enables exponential revenue growth without increasing fixed costs. Rebel can spin up and fulfill an entirely original brand at the expense of only ingredients and creative effort. 

Critically, Rebel owns and operates all of these kitchens through general managers referred to as “Chief Delight Officers”. For comparison, NextBite, an American corollary of Rebel’s, doesn’t own their fulfillment. Instead, Nextbite depends on kitchens provided by existing restaurants with excess capacity. They don’t own the vertical production pipeline. Upsides to doing so, as Rebel does, are innumerous: Rebel’s friction to launching a new product (restaurant) is minimized to effectively zero. The company can immediately deploy new ideas to their CDOs for testing and feedback. Rapid innovation and iteration of this sort defines Rebel’s rapid ascent of the digital restaurant mountain. 

Two restaurants originated by NextBite, the virtual brand partner, create an ideal case study for this issue. Crack’t and Hatch House are both digital breakfast sandwich shops. This means that their not so radical takes on the McMuffin are cooked and handed off to delivery drivers out of fulfillment - also known as “ghost” - kitchens. They’re not just both breakfast sandwich shops though, they’re the same breakfast sandwich shop. Both use the same bread, the same cheese, the same eggs. Both serve up five options, have one side - Farm Rich Whole Grain French Toast Sticks (which are precooked and frozen for the curious), and bolster incredibly similar cook assembly diagrams. Nextbite isn’t pulling wool over anyone’s eyes here: the company’s two plays are using an identical base sandwich. 

The canvas of their offering is the same, so what, then, are the differences? What clears NextBite from getting the boot from UberEats? One word: Peppers. Hatch House is designed to appeal to a more upscale audience. An audience with a more mature palette than Crack’t. As a result of data mining and trend analysis, NextBite must have found that peppers are the best way to make a sandwich for kids into one for adults. Result being Hatch House’s addition of Cherry Peppers, Pepper Jack Cheese, and Chili Crisp Aoili to their offerings. 

“Make everyday something special” says Hatch House’s front page. “Take your breakfast to the max” responds Crack’t’s. A clean farmhouse logo vs bright colors and bold text, Hatch House comes off as a mellow option to the parents of Crack’t customers; Customers who are likely ordering their “Hangover Helper” because it has a good name and they want hash browns in their sandwich after a late night. 

Two ways of looking at this play: NextBite is simply getting the most out of their fundamental product or they’re deceiving customers by passing off the same food as separate restaurants. Either way, it's a clever move to attract the broadest audience whilst sidestepping the new delivery app rules around having unique menus. So too, this sandwich scenario demonstrates that a fine line separates brand and food. We eat with our eyes, and we pick restaurants with them too. 

First: repeatability. NextBite’s model depends on what are called fulfillment kitchens to cook the food for delivery pick up. These kitchens usually use excess capacity in existing restaurants. They have the tools, they have the employees and can make a few extra thousand by cranking out calorie bombs. Crack’t needs to replicable across kitchens from Georgia to California. Ingredients must be mainstream and processes simple enough for the least skilled worker in the chain. In execution, this means that Crack’t offerings are made up of only thirty SKUs (ingredients) that are orderable from two major American purveyors - US Foods and Sysco. Fulfillment kitchens tap order on this shopping cart and then head over to the Shopify to purchase the logo stickers to slap onto the product. 

Assembly is the same story. NextBite provides a graphic that may as well have come from one of that iPhone game CookingFever, where kids cook and assemble virtual fast food. This “Build Chart” elicits one obvious thought - couldn’t a robot just do this? Looking closer at the sandwich assembly procedures and that question turns to a comment. A robot could definitely do this. The procedures are less than seven steps and consist of “cook eggs”, “slice bagel”, “spread sauce” or some version of that. 

Urban Dictionary here: the word “Cracked” is an internet gaming word for insanely good. This evidences the second layer of NextBite’s strategy: marketing. Dollars and thought do seem to have been applied to the Crack’t website. The pictures are high quality, the logo is appealing,  and the site is generally visually enticing. In the name too, NextBite is making it clear that brand matters. Brand that appeals to young people likely to order “The Hangover Helper”. Product isn’t the emphasis, but marketing is. A major miss is the lack of a strong social presence. Crack’t doesn’t have Instagram or TikTok - seems like an oversight from a company trying to get into Gen-Z’s wallets. 

Looking at the Crack’t page on Doordash, one thing immediately stands out. The numbers are weird. Sandwiches cost $10.80 or $13.20, french toast sticks cost $7.70, and delivery closes at 5:40pm. $XX.99 is no more. NextBite is trying to optimize every part of their operation. They concluded that these prices and operating hours are by some metric most likely to produce the highest margins, and so that’s how they do them. Is it new? Yes. Is it confusing? Yes. Will it last? We’ll see.

Major players in the casual dining space across America are operating a parallel strategy: Launch new digital brands operated via ghost kitchens that appropriate their existing ingredients into indulgent hangover food items. These menus focus on only a few such items, for instance Outback Steakhouse’s “Tender Shack” has effectively two choices: “Dang Good” fried tenders and “Nashville Hot AF Tenders” that can be ordered as a combo, on their own, or in sandwich form. Further, the items that are offered bolster names that sound as though some middle manager was tasked with producin titles that are “trendy” and “young”. They’re loud, they’re aggressive, and yes they’re a bit out of touch, but these restaurants are all clearly trying here. Trying to make their brands accessible and enticing to a new audience. 

Restaurants like IHOP and TGI Friday’s feel worn out and old, which is ok if you’re forty and it's a wednesday morning but not if you’re twenty three and you have a hundred different alternatives. Disconnected from the younger generation, these brands are looking for an in by any means. Look no further for evidence of this than the partnership between IHOP and Noah Schnapp - the Netflix Stranger Things teenage actor. Facilitated by Nextbite, the delivery-only food brand launching service, IHOP fulfills orders for Schnapp’s new “restaurant” TenderFix. If you remove the logo, TenderFix and Tender Shack could be the same restaurant. TenderFix serves two fried chicken sandwiches, with side options as follows: waffle fries, chocolate chip cookie, brownie. That’s it. That’s the entire menu. They didn’t even try. Why aren’t they called “Walloping Waffle Fries” or something like that?

This is about as clear of a branding play as you could imagine. Schnapp slaps his name on some chicken, IHOP makes that chicken, and Schnapp’s fans eat that chicken. That’s the gist of it. Now Nextbite deserves an article of its own, but there are two things important to note here. The first is the strategy description on its website: Data mining, trends, psychographic analysis. Basically, Nextbite is doing the same thing TikTok does to get you to keep watching videos. The company is finding the simplest dopamine rush possible to get people to buy food - in this case it's glutinous chicken sandwiches. The second is that Nextbite also runs IHOP’s “Super Megga Dilla”. Yeah, the Pizza Quesadilla place. 

As their existing audience ages out, the evolution of these brands depends on catering to Gen-Z. Will this be entirely through virtual brands? And will these virtual brands wholly dependent on branding be able to hold weight in the next decade? 

Item one: quality. If the food doesn’t taste good, a restaurant won’t be successful. Bottom line. However, in 2024, taste is a commodity. Wheat, sugar, even milk: those are traditional commodities. The quality of these raw ingredients plays a big part in differentiating outcomes. Bread using better wheat and better sugar is going to taste preferable to an alternative with inferior composition. What happens, then, when two products use identical high-quality ingredients? This is where "food intelligence" comes into play, encompassing everything from a chef's choice of recipes to their expertise in determining when the bread is baked, so to speak.

Traditionally, culinary expertise, or “intelligence alpha,” was highly valued. This expertise, however, is no longer exclusive to experienced chefs. For instance, many prefer a Sweetgreen bowl to a meal prepared by an experienced chef at the same price point. That Sweetgreen bowl may very well have been assembled by an automated robotic arm. Non-experiential dining has reached a scale where the art of producing sufficiently “tasty” food is largely demystified. Marginal gain in taste for a consumer from experimenting with lunch options is narrowing to zero. Cooking intelligence, once a prized attribute, is accessible. This shift indicates that the nuances of taste and food preparation are no longer as significant as they used to be.

Speed, like quality, has transitioned from being a distinctive advantage to a baseline expectation. Initially, the allure of these “bowl restaurants” was their ability to deliver high-quality meals swiftly. The promise was a good meal served fast. However, this speed advantage is diminishing in value. The growing prevalence of fast-casual dining options, coupled with the ubiquity of delivery services, has reduced the significance of time as a demarcator. Ordering Chipotle via DoorDash may save about ten minutes compared to other options, but the delivery for most restaurants takes between thirty to forty minutes anyway. Consequently, the aspect of immediate access has become a standard in the industry.

This leaves the brand. In a market saturated with options that offer tasty, fast, and reasonably priced meals, the decision-making process for daily dining has shifted towards perception and marketing. Bowl restaurants cultivate trust with their audience, encouraging repeat visits for familiar favorites. To stand out in this crowded market, restaurants must engage in effective marketing. This includes appealing design, strong visuals of their dishes, and an active social media presence. Building a strong brand identity is now crucial. With numerous establishments offering the same chicken and grain bowls, the challenge for each is to answer: Why should consumers choose yours?

Ghost kitchens, operated ideally, should serve the same function. A restaurant deploys its menu to a network of these remote kitchens. Patrons in cities containing distributed kitchens, then, are able to order from a nearby fulfillment center. This is the idea of the Food Delivery Network, the "FDN". Delivery not in the direct language of DoorDash or UberEats, but in the language of delivering to locations where the food isn’t originated. Uploading to these networks is a great analogue to deploying to a server - you have to upload recipe and method instructions in the same way packages are installed and startup instructions provided. The user has to teach the server instance how to distribute their content. How to cook up the special sauce.

The premise that you can’t get exactly what you want to eat when you want is friction. If your favorite spaghetti dish is served in Boston and you live in San Francisco, you’re out of luck. This doesn’t have to be the case. Optimally, that Boston restaurant uploads their menu to the FDN, and a kitchen near you produces and delivers that delicious spaghetti to you. Friction removed. The restaurant profits from this transaction, and both parties get a net positive value add. You downloaded food into your stomach because the restaurant uploaded that food to the network.

There’s nothing intrinsically new about this idea. After all, franchises are effectively the same concept. A restaurant like McDonald’s makes their menu globally available by opening more and more franchises. Installing more and more servers effectively. What’s different about Ghost Kitchens is that they’re a fraction of the cost to start and operate, making them viable for a greater breadth of restaurants whose models aren’t already set up for mass distribution. "Simplicity" is the name of the game (acknowledging that presently this isn’t necessarily the case).

Travis Kalanick’s CloudKitchens is aptly named. For that’s a key dimension of what Ghost Kitchens are: a cloud for restaurants to deploy their menus globally. Operationally, these kitchens aren’t at a point of efficiency - restaurants aren’t either, for that matter - where this pipe works as cleanly as outlined above. When they do, however, access to food is going to open up in a way unseen up to this point.

Taking this to a greater extent provides clarity on a broader shift in the digital restaurant space, particularly in the case of virtual food halls. At these halls, customers can order from any of sixty or so brands in a single order. The sixty number isn’t pulled out of a hat, that’s how many brands operate out of a single 2,000 square foot Kitopi - the Dubai based cloud kitchen company - hub. You can get your burgers from one place and your fries from another. This breaks down the friction that previously protected menus. When you walk into a restaurant, you accept that much of the menu is filler to cater to broad tastes or support optionality. Many of the starters are put there because people might prefer that option based on their preference for that category rather than its alpha over the alternative. Basically, you might order guacamole because you just prefer guac even if the Ceviche is 'better'. Consumers aren’t choosing the restaurant’s specialties per se. Even at a great restaurant only some fraction of the menu is going to be outstanding.

In the case of a virtual food hall, however, consumers don’t have to settle for subpar starters. Instead, those starters can be ordered from a different restaurant in the food hall. If you’re in the mood for sushi you’re able to get the rolls from one spot and the Brussels sprouts from another. Such a reality acts as a forcing function for restaurants to focus only on their core offerings. There are drawbacks to this optionality, however. Selecting items from different restaurants requires higher friction as well as uncertainty.

Are consumers going to have enough effort to track a Brussels sprout brand? Will a starter from one spot match the flavor profile of a different restaurant’s entrée? To answer that question, envision a complete disaggregation - so the items from all sixty brands are listed in one menu. Starters are lumped. Mains and desserts too. Bondi Sushi’s 'Shrimp Avocado Rolls' are sitting alongside a 'Kick My Boss - Spicy Burger' from Rock House Sliders. One note in support of this model is that it’s basically Cheesecake Factory - home of what seems like America’s and, who are we kidding, the World’s longest menu running 250 items. The menu has a small number of branded items of the type you’d see in this virtual food hall such as the Oreo Dream Extreme Cheesecake. So too, it spans cuisines from French to Korean to Mexican and carries breakfast, lunch, and dinner options. A customer flipping its pages could really imagine it having at one time been ten different restaurants lumped together.

So this model could work. Let's assume, though, that this optionality overwhelms the consumer leading to suboptimal ordering at the least. One possible solution is bundled menus. Delivery apps or Kitopi itself could look at each spot and coordinate their offerings into a number of menus. Some of Kitopi’s kitchens produce for one hundred distinct restaurants. If we spot each of these restaurants at twenty items - that’s two thousand unique items and potential for 1026 unique options for groupings of ten foods. This would provide the possibility to pick and choose the best from each brand to coordinate integrated menus. These menus could be A/B tested across the audience to find the optimal combinations. Taking this further, the quality of 'best' can be analyzed from two angles for middle and low priced restaurants. One is what’s most popular, the other is what any individual’s subjective opinion is.

For the latter, Kitopi would be able to deploy recommendation algorithm style strategies to create menus based entirely on consumer preferences. The platform would be able to learn what types of starters you prefer in combination with which mains and the like. Do you order dessert? What do you order when you do? Is the way you order Japanese food different from when you order Mexican food? By coordinating offerings catered to these tastes, Kitopi would merely be doing what every other content-based platform does - Netflix, YouTube, Amazon. Bringing recommendations down the chain to menu construction. This is feasible for group orders as well. Kitopi could create a menu based on its restaurant base that merges the preferences of five, or however many, people.

There’s potential for the creator economy angle here as well. The 'Travis Scott Meal' at McDonald's gave fans of the music artist the ability to order just as he does. Demand is clearly there - the meal led to the strongest revenue month in a decade for the Golden Arches and caused a shortage in chopped lettuce. The meal didn’t consist of any unique items, but rather just Scott’s effective order: Quarter Pounder with Cheese, Fries, Sprite. Simple as that. If Kitopi can create their own menus, creators could just as easily do the same. Any content creator could create their 'branded' menu or combo selecting from the 2000 possible items. They would get some rev-share from orders based on their menus and Kitopi wins because it will receive order spikes in the same vein as McDonald's did.

From the restaurant side, food itemization is effectively akin to licensing out their content library individually rather than as a whole. Kitopi may only want to serve samosas and butter chicken from an Indian restaurant because those are deemed the 'best', highest traction items. Some other spot does Dosas better. Maybe it won’t be this narrow and restaurants will do ten dishes with enough uniqueness and taste value to get picked up. This is almost certainly true, because restaurants doing one thing well is a strong indicator of the quality of everything else, but how strong an indicator is the question? Will that samosa restaurant start focusing on only samosas and make that their key offering? After all, they could license out these samosas to hundreds of kitchens to serve a broad customer base. Because of the distributed kitchen network, restaurants get high leverage on selling one item well. Without delivery, there aren’t enough local customers to support a samosa only shop. But there are millions of app customers who are in the samosa market. If that restaurant gets a ten percent kickback on each item sold, they have the potential to stand up a streamlined, scaled business on the back of their 'best' offering.

Or not. Once a restaurant comes up with a promising item, develops the recipe, and deploys it to the Kitopi base, there are a few options. One is to just sit back and collect the licensing fees, ride that Samosa money until a better one comes onto the market. Inevitably, however, one will come onto the market. This requires restaurants to operate like software companies, the Samosa needs to be versioned, iterated upon, and customized to stay ahead of the curve. This is part of the reason why restaurants may become incredibly narrow. Massive distribution means focusing on one item is enough of a business. Those who do will make that item at a level far greater than competitors. That’s all to say there may very well be an exclusively brussel sprout brand in Kitopi’s kitchens soon.