Data-Driven Dark Kitchens
Location intelligence to expand your online food delivery reach
This is an analysis for location strategy of dark kitchens, an emerging trend in online food delivery in which restaurants expand online delivery reach without traditional brick-and-mortar stores.
For readers new to China’s food delivery scene, I’ve given a quick introduction below to help contextualize the analysis. If you’re an expert/insider, feel free to skip the following section.
Food delivery trends in China
Food delivery is the new normal in China. $34bn was spent on online food delivery in 2018 and is projected to grow 10%+ in the near term.
Driving this trend is rapid urbanization and the increasingly busy life of city people seeking convenient options. This is consistent with developed cities around the world. But China leads the world in mobile e-commerce, and by extension, China has arguably the world’s most mature on-demand delivery network and mobile-adopted consumers. This is why food delivery has taken off in China.
An offshoot of this rising trend is dark kitchens. Restaurants (but also coffee shops, bakeries, and whatever food people might order online) can skip traditional brick-and-mortar stores for “dark” kitchens, business operations that are setup mainly to serve online orders.
Dark kitchens have lower overhead costs. Since these “stores” only engage customers digitally, there is no need for prime commercial real estate. And operators enjoy a streamlined operations with no tables to wait nor seats to fill.
In this new paradigm, food service businesses must adapt to these new online consumer trends and create new sustainable business models to win. Some brands are already expanding delivery service reach by leveraging the major delivery platforms from Eleme and Meituan — and they do so without the traditional retail footprint and its associated overhead.
food service businesses must adapt to these new online consumer trends and create new sustainable business models to win
Our approach
So how should food service brands formulate a dark kitchen growth strategy? With dark kitchens, traditional considerations for brick-and-mortar stores need to be reconsidered.
To tackle this challenge, we took a cross-disciplinary approach combining food-retail factors, urban factors, and human factors. We analyzed those factors at three scales:
- City: to identify delivery hot spots by assessing relative market demand
- Neighborhood: to determine location typologies to target neighborhood demographics
- Street: to select a specific dark kitchen location site.
City: Which areas of the city perform better in delivery?
For this study, we created a hypothetical brand: Mala Bloom is an upstart food chain serving up delicious bowls of malatang. (Haven’t heard of malatang!? It’s all the craze in China. It’s essentially a scaled down, more convenient version of hot pot, which is easily the most popular category in China: $112 billion was spent on hot pot in 2016.)
Mala Bloom wants to capitalize on the food delivery trend and expand aggressively through dark kitchen strategy. So how should we prioritize the locations we open new dark kitchens?
The best location for a dark kitchen is the one with the greatest potential market demand. The starting point for our analysis is therefore to determine relative market demand for malatang food delivery. We do this by comparing different areas of the city based on existing malatang competitor order volumes.
Here we mapped existing competitor malatang restaurants and their monthly order volumes based on eleme’s platform data. Since eleme is exclusively a food delivery platform, the monthly order volumes represented here is exclusive to online orders.
While these monthly order data is not a good indicator of total market demand, it is a valuable metrix to gain an understanding of relative demand.
We use a heatmap analysis to determine the intensity of market demand for each area. The darker the red suggests the greater the concentration of market demand. Although the Lujiazui business district in Pudong doesn’t have any one restaurant surpassing monthly order volumes of 7,500, the aggregate of restaurants and store order volumes in the area has the greatest concentration of market demand.
The city-scaled analysis resulted in a cohort of 15 area “hotspots” which we used as a base for our subsequent analyses. The 15 hotspots represent the highest relative demand for malatang food delivery across the entire city.
Neighborhoods: How can I target a specific demographic?
At the neighborhood scale of our analysis, the objective is to uncover the composition and characteristics of each area (hence “neighborhood”), which helps us to further refine and target a specific demographic and inform our location strategy.
First we need to define each area’s neighborhood boundaries. In the case of food delivery, an intuitive neighborhood perimeter is what is within the delivery service coverage area. From the central point of each area hotspot, we calculated a 3km delivery service coverage using Baidu’s shortest route API, specified for scooters (which is the standard vehicle for delivery workers). This approach considers roads and navigation routes when calculating the 3km distance travelled, as opposed to a simple circular 3km radius. (Eleme does not disclose how it calculates its limit for delivery service range, but in practice is says it does not deliver beyond 3km. Also, restaurants have the discretion to further limit the delivery service range, but for the purposes of our analysis we have ignored this.)
Within each neighborhood, we plotted residential compounds and office buildings. Gold dots represent residential compounds, while purple dots characterize office buildings. The size of each dot represents size of population per datapointbuilding, calculated as 2.5 persons per unit for residential compounds, and 1 person per 15 square meters floor area for office buildings.
The data is parsed, cleaned and merged by our teams from various sources, including Lianjia (链家), 51Banban (办办网), Fantainxia (房天下), Anjuke(安居客). By cross-referencing datasets, we avoid any specific platform biases and ensure robustness.
With this we can approximate the ratio of office workers-to-residents. Again, this is not a precise measure of actual total population, but certainly a useful one to determine relative measures of population compositions. The results highlight well-known business districts including Pudong Lujiazui, Jing’an, Wujiaochang, and Xujiahui, but also highlight a few less obvious ones.
We also approximated food delivery saturation by calculating the approximate number of consumers per delivery store (all store listings, not just malatang, on eleme’s platform). The results here are interesting: the most unsaturated delivery market is the largest total addressable market by population, Pudong Lujiazui; whereas the most saturated market is Former French Concession, where population density is, surprisingly, considerably lower. Despite being considered the true center of Shanghai, the Former French Concession neighborhood actually includes many old lane house communities protected from new (i.e. taller) building developments.
Malatang is typically eaten alone, unlike the traditional format of hotpot which is usually eaten in groups. Part of its appeal is actually the convenience of being able to indulge in hot pot without all the fuss of getting together, making it a staple among 996 office workers looking to enjoy a hearty meal .(“996” is a term originally tokened by Alibaba to characterize technology teams working from 9 am to 9 pm every day and 6 days per week. The term is now used to refer to people who work a lot.)
As such, the launch strategy for Mala Bloom is to target neighborhoods with a high population of office workers.
Pudong Lujiazui has the highest concentration of office workers while being the least saturated food delivery market of the cohort. This makes for a strong candidate for a site location. Even though the neighborhood has a high office worker-to-resident ratio, it still ranks 3rd in concentration of residential compounds, which is useful for an all-day service offering. For the purposes of this illustration, we selected Pudong Lujiazui for our subsequent analyses.
Streets: Where exactly do I select?
One of the key tenets of a dark kitchen strategy is to take advantage of lower rental prices by selecting off-main-street locations. The objective of the street scale analysis is to determine a specific location while considering various factors including price and delivery time convenience.
We mapped the site locations of commercial leases within Pudong Lujiazui, filtering for commercial spaces between 20 and 30 square meters that are permitted to operate a commercial kitchen. The dots are color coded according to their leasing rental price; the darker the tone of red, the lower the price. We manually selected six site locations of the lowest price group (of course in practice you can choose more or less to consider).
Here, our team parsed and merged from various sources including Baidu and 58tongcheng (58同城).
Next, we assessed the quality of site locations from the vantage point of delivery time convenience, which we define as driving distance and number of intersections. These two factors influence the actual delivery time, but is largely dependent on traffic conditions at the time of delivery.
To calculate both driving distance and number of intersections, we used Baidu’s shortest route API to determine the routes taken (an assumption) in order to determine the distance driven and intersections crossed. The routes from this calculation are visualized in the gif above.
We calculated averages for driving distance and number of intersections for each set of routes from site locations (n=6) to the office buildings (n=354) in the neighborhood. The results in the chart suggests a comparative advantage in delivery time convenience for those in the lower left versus upper right.
While the per trip differences might seem marginal, on aggregate the time convenience savings can be significant for the bottom line. For instance, the savings of 0.5 km/trip, if we assume average monthly orders of 3,000, equals 1,500 km/month or 18,000km/year.
In China, these costs are usually borne by the delivery platforms (i.e. Eleme and Meituan). But for brands seeking to run and operate their own deliveries (more likely the case in rest of world, maybe not China except for the largest chain brands), the aggregate savings can be substantial.
Of course there are still other critical site location factors that might be difficult to quantify and map. Ultimately, a site inspection is indispensable to make a decision, after having used a data-driven approach to narrow down the options. These other factors include delivery vehicle accessibility, operating hour constraints (a concern for shopping mall sites), FDA operating licenses, among others.
Data empowers us to see from a new vantage point. Our analysis from the city, to the neighborhood and down to the streets, hopefully has demonstrated how unique insights could emerge from this approach.
At the end, strategic decisions involve multitudes of variables, both objective and subjective ones. The role of location intelligence is not meant to substitute but to augment decision-making by management teams with data-informed insights.
This case study was produced by CVI, a location intelligence company with a mission to help retail businesses make better decisions. At CVI, we take a human-centered and data-driven approach to build technology tools and to formulate strategies that empower city-shapers. If you are a food service operator, a retailer, a city planner or an urban design enthusiast and are interested in harnessing the power of location intelligence, give us a holler info@cvi-tech.com.
