On Dollar Slices, Pizza Vectors, Prosciutto Zones and Topping Hyperspace

New York has several things that might be described as “iconic” — Central Park, a subway system that runs all night, deli coffee cups featuring ancient greek amphorae — but when it comes to food, there are few meals as iconic as pizza (ok you bagel people, we hear you :). Pizza in New York City has evolved in countless ways since it first arrived in the late 19th century. Once a fairly uniform experience, eating pizza in NYC has grown to encompass many styles, restaurant ambiences and price points. Like music and coffee, pizza in NYC is both ubiquitous and highly variable. As such, it provides a fascinating lens on the culture of the city.

Published in
13 min readAug 6, 2018


While the primordial slice of New York Pizza — a composition of flour, tomato sauce and mozzarella — may have set the definitive mold, today’s pizza landscape is hugely varied. From vegan slices that dispense with cheese altogether to $2000 truffled pies literally covered in gold, hipster-neo-sicilian pies to Dominos, the multitude of pizza styles seems to be ever expanding.

From left to right: vegan pizza, $2000 gilded truffle pizza, neo-sicilian pizza from Emmy Squared, Domino’s

At Topos, we are fascinated by exactly this type of variation and believe it provides a powerful view into the culture of a location. While data sources like the United States Census are useful for understanding broad demographic trends over decades, they give little insight into what defines the moment-to-moment culture of a city, a neighborhood, a street corner.

Walter Benjamin at work

Inspired by thinkers like Walter Benjamin, who, in his unfinished Arcades Project examined subjects as varied as fashion, construction materials, poetry, lighting, and mirrors in order to understand Paris in the 19th century, we are fascinated by the way seemingly simple, ubiquitous subjects like the coffee we drink or the concerts we go to define a place. However, unlike Benjamin, we are interested in constructing this understanding in a way that can dynamically scale across the globe, allowing us to understand how different locations relate to one another, and how locations evolve in real time. To achieve this, we use data from dozens of different sources and techniques from a wide range of technologies and disciplines including computer vision, natural language processing, statistics, machine learning, network science, topology, architecture and urbanism.

In this article we apply this multidisciplinary, multidimensional approach to pizza in order to understand what pizza can tell us about New York City, and — conversely — what other aspects of life in New York City can tell us about pizza.

Pizza History

Lombardi’s Pizza, established 1905 and still in action (credit: Ephemeral New York)

New York-style pizza, like bagels and the Statue of Liberty, has its roots in a foreign country. In particular, the original New York pie was closely based on the traditional pizza of Naples — known as Neapolitan Pizza — a small, round pie topped simply with tomato sauce and mozzarella cheese.

Pizza arrived in New York with the waves of Italian immigrants who came in the late 19th and early 20th century. In 1905 Lombardi’s became the first official NYC pizzeria when Gennaro Lombardi asked his employee Antonio Totonno to start making pizzas in the bakery oven in the back of his grocery store. Originally a to-go business, customers would buy whole pizzas at 5 cents a pie. When pizza sales inevitably eclipsed the rest of the grocery store’s business, Lombardi’s transitioned to running as a pizzeria full time. Lombardi’s also started the New York tradition of eating individual pizza slices; when customers couldn’t afford a whole pizza, Lombardi’s would sell slices by the inch depending on what the customer could pay.

Pizza in Italy was traditionally cooked in wood-fired ovens. Wood was harder to come by in NYC, so most of the first New York pizzerias used coal-fired ovens, a trend that continued as pizza’s popularity grew. Coal-fired ovens helped take pizza from its Neapolitan form — thin-crusted, lightly topped, individually sized and eaten with a knife and fork — to the Neapolitan-American style: larger (~16”) pies, that lacked the chewy center of wood-fired Neapolitan pies.

A slice of NYC pizza from…Ray’s… ;)

The true New York Style pizza wasn’t born until Manhattan banned coal ovens to limit the amount of heat they generated, forcing pizzerias to start using gas. Gas ovens cooked pizza more evenly than coal, unlocking new worlds of form and texture; pizzas could now be as large as 18” in diameter, enabling division into 8 triangular slices, and could support significantly more mozzarella cheese. Thus was the NYC style slice born.

This foundational cultural invention has happily remained ubiquitous throughout the city. At the same time, pizza culture in NYC has evolved in myriad ways over the past 100 years. Wood fired ovens are making a comeback; vegan slices are on the rise; new school ‘Brooklyn style pizza’ pioneered by restaurants like Roberta’s and Emmy Squared have influenced topping approaches not just in NYC but around the world — never again will slices be safe from honey and jalapeño peppers.

Pizza in New York has not evolved in isolation. Despite its global influence, the pizza scene in New York continues to absorb a variety of ‘imports’: pizzerias, toppings, pairings, and pie style. These imports range from national chains like Domino’s (established in Ypsilanti, Michigan in 1960) to burrata cheese (from Murgia in southern Italy); the popular pairing of pizza with chicken wings is also an import, albeit from another part of New York State (Buffalo); In addition to American pepperoni, pizza is now topped with a variety of cured meats from around the world: speck, prosciutto di Parma, sopressata, chorizo, and jamón ibérico have all found their way onto New York pies. In this sense, New York Pizza has become a forum for cultural exchange — a blank canvas where contemporary pizzaiolos can express themselves and the influences that formed them — a theme well captured by the inaugural episode of Ugly Delicious.

Pizza Dollars

Gold covered truffle slices aside, people often think of New York pizza as an inexpensive meal of convenience. The meteoric rise of the dollar slice over the past 20 years supports this perception; yet the economics of NYC pizza is not always as expected. For example, depending on location, plain cheese pies from national chains like Papa John’s can range from $8.99 (Bay Ridge) to $18.99 (Springfield Gardens), surpassing even nationally revered pies like Robertas’ $16 Margherita (Bushwick).

The Pizza Principle, graphed

It’s impossible to bring up the economics of NYC pizza without mentioning the delightfully titled “Pizza Principle”, a theory first proposed in 1980 by Eric Bram which observes the surprisingly tight correlation between a slice of pizza and the base fare for a subway ride. From 1960 to as recently as 2014 the principle has largely stood up, but at $2.50, the 2018 mean price for a slice has yet to catch up with 2015’s subway fare hike from $2.50 to $2.75.

Pizza locations in relation to walking time to subway entrances

There’s also a strong geographic connection between the subway and pizza: Of all the pizzerias in New York City, 79% (1438 / 2290) are within a 10 minute walk of a subway entrance. Looking only at pizzerias that sell individual slices strengthens the connection — we find 82% are within a 10 minute walk of a subway entrance. This relationship speaks to the foot traffic required to sell a whole pie by the slice to different customers, a condition that becomes even more geographically specific when considering the economic miracle of invention that is the NYC dollar slice.

Dollar Pizza

The dollar slice is a phenomenon largely unique to New York City (with some notable exceptions). Dollar slice pizzerias started popping up along 6th Avenue in Midtown in the early 2000s, gaining popularity in the wake of the 2008 recession. Despite 2012’s price wars which saw slices dip as low as 75¢, the $1 price point has stuck around even as the economy has recovered and prices for other things (like rents) have gone up.

The business model for dollar slices is precarious. Most dollar pizzerias source their ingredients from wholesaler Restaurant Depot, and use the cheapest low-moisture mozzarella (around $1.50 per pound), tomato sauce, and flour. Factoring in the cost of ingredients, rent and labor, the average pie costs $6 to make, or 75 cents per slice. Charging a dollar per slice leaves a slim profit of 25 cents, making one component of the business model — high sales volume — critical. Dollar slice pizzerias need to sell around 200 pies (1600 slices) a day to remain viable; some sell as many as 450 pizzas a day. In contrast, pizzerias that sell slices for $2.50 can remain profitable selling 50 pies a day.

90% of dollar slice pizzerias are within 3 minutes of a subway entrance

Given the high volume of customers needed for dollar pizzerias to stay afloat, it is unsurprising that they tend to be located in areas with large transient populations. The relation between subway entrances and dollar slices is the strongest yet — 98% of dollar slices are within a 10 minute walk of a subway entrance. However, some neighborhoods like the financial district have a conspicuous lack of dollar slices despite a large daytime population and proximity to a subway; conversely the East Village has a large concentration of dollar slices with a low daytime population. Looking at the nighttime population of the city helps explains these discrepancies: neighborhoods like the financial district dramatically slow down at night (prime pizza time) while the East Village heats up.

Left: daytime population map. Right: nighttime population map encompasses East Village dollar slice locations

Dollars and Truffles

Despite the obvious economic value proposition, and clear relation to daytime and nighttime population, dollar slices say surprisingly little about long term neighborhood economic indicators like median household income or average house value. The daytime and nighttime populations that pass through midtown Manhattan, where the highest concentration of dollar slices occur, is radically different from the resident population, with correspondingly different income brackets.

Dollar slices say surprisingly little about long term economic indicators; prosciutto and bufala mozzarella say more

Truffles, on the other hand, say more: having truffles available as a topping within a neighborhood corresponds to a 31% increase in median household income on average, and a 45% increase in average house value. In contrast, the presence of dollar slices within a neighborhood corresponds to a 5% average increase in median household income and a 24% increase in average house value.

As fancy as truffles are, there are fancier toppings yet: we found that prosciutto di Parma was the most powerful topping differentiator, representing a 48% increase in median household value on average, and a 47% increase in average house value. Prosciutto di Parma was closely followed by bufala mozzarella (24% increase in median household income, 42% increase in average house value) and speck (21% increase in median household income, 42% increase in average house value).

50 dimensions of Pizza

After looking at how certain pizza toppings partitioned NYC into distinct economic groups, we wondered what they could tell us, taken together as an ensemble, about the city. To this end, we constructed a 50 dimensional “pizza vector” for every NYC neighborhood. In addition to capturing the distribution of toppings like truffles and speck throughout the city, these pizza vectors encompass a range of other pizza dimensions including the availability of wings with pizza, the presence of national chains, the availability of dollar slices, the inclusion of pizzerias on Eater’s list of Iconic Pizzerias, and more.

Confusion Matrix for pizza predicting third wave coffee: accuracy = .84, recall = .89, precision = .88, F1 = .89

With our pizza vectors at hand, we set up a number of classification experiments to find out what pizza could tell us about various neighborhood amenities and establishments. To ensure we weren’t overfitting, we trained our classifiers using 10-fold cross validation, and computed confusion matrices averaged across the folds to determine model performance. Of these experiments, predicting the presence of third wave coffee shops, at 84% accuracy, was the most successful (most predictive feature: ‘density of pizzerias that have shiitake mushrooms available as a topping’). Following third wave coffee, we found our pizza vectors were able to predict whether a neighborhood had an art gallery (81% accuracy, most predictive feature: ‘per capita pizzerias that have guanciale available as a topping’), whether a neighborhood had a bookstore (75% accuracy, most predictive feature: ‘per capita pizzerias that have “margherita” on the menu’) and whether a neighborhood had a public playground (65% accuracy, most predictive feature: ‘density of pizzerias featured on an Eater list’).

Topping prediction

Flipping the script, we then wondered what the city could tell us about pizza toppings. Using Topos’ suite of neighborhood features — which encompasses everything from topological analysis of urban form, to ambient light levels, to religious diversity — we built a variety of ‘topping classifiers’ to predict the presence of a given topping within a neighborhood.

Confusion matrix for Topos features predicting speck: accuracy = .89, recall = .9, precision = .92, F1 = .91

Of these classifiers, at 89%, our ability to predict the availability of speck was most successful (most predictive feature: ‘density of trendy shopping’). Closely following speck was pancetta (87%, most predictive feature: ‘yoga studios per capita’), burrata (82%, most predictive feature: ‘density of children’s stores’) and gorgonzola (76%, most predictive feature: ‘visual appearance of modernist architecture’).

The Five Boroughs of Pizza

Having constructed neighborhood pizza vectors, we decided to ask one of our favorite data-enabled questions: “How does pizza partition the city?” or, in other words, “What are the five boroughs of pizza?”. Taking a similar approach to when we redefined the five boroughs of NYC to reflect the realities of the 21st century, we first applied dimensionality reduction to our 50 dimensional pizza hyperspace, resulting in 12 reduced dimensions (representing 87% of the original variance). We then applied K-means clustering (K=5) to this reduced 12 dimensional pizza space, thereby producing ‘five boroughs of NYC pizza’.

Two radically different zip codes, taken from the same zoom level

Unlike our earlier work redefining the five boroughs, in this attempt we decided to abandon zip codes, instead focusing on a more granular understanding of the city. In general we find this level granularity has several advantages over higher level zip codes, which are often wildly inconsistent, varying from single buildings (like 10118, The Empire State Building) to regions of up to 5500 square miles (like 89049, in Nevada). In the case of our five boroughs of pizza, working granularly allows us to distinguish less accessible areas like Brooklyn’s Navy Yard from the surrounding neighborhoods of Dumbo, Fort Greene, Clinton Hill and Williamsburg. It also provides a more nuanced view of neighborhoods like Williamsburg and Bay Ridge, which are often considered monolithically. Here we find Williamsburg has three distinct pizza pockets, while Bay Ridge has four.

Left: The Five Boroughs of Pizza; Right: The granularity of the clustering separates the navy yards from surrounding neighborhoods

While these five boroughs of pizza broke up Staten Island, the Bronx, Brooklyn and Queens in interesting ways, we were dissatisfied that Manhattan remained largely unified. To address this, we decided to zoom into the red cluster containing Manhattan, and further divide it into 10 subclusters.

10 Subclusters of pizza. Park Slope groups with the Upper West and Upper East side of Manhattan

Beyond breaking up Manhattan, we were happy with the way the subclusters produced a more nuanced texture for Astoria and Harlem, which were previously monolithic. Perhaps even more compelling are the synergies and continuities that emerge. The Upper West and Upper East Side stay largely continuous. Furthermore, they are part of the same subcluster, joined only by Park Slope, uniting what are often considered the most desirable neighborhoods in New York City for wealthy families with children. The large orange cluster is another interesting case: cutting across the trendiest parts of Williamsburg, Bed Stuy, Bushwick, Ridgewood and Crown Heights, this pizza pocket remains singular, defining a unique zone of hipster pizza culture within the city.

The orange zone demarcates a hipster pizza pocket cutting across Williamsburg, Bushwick, Ridgewood, Bed-Stuy and Clinton Hill

A Granular View of Culture

Spending so much time understanding the nuanced geography of pizza may seem excessive, if not downright silly — certainly 50 dimensions of pizza is over the top by most standards. Nonetheless, we believe constructing a granular view of culture has enormous potential to transform how we understand place.

Just as high level musical genres like ‘rock’, ‘jazz’ etc have little power to actually describe music, let alone recommend it, we believe high level descriptors of place like ‘urban’, ‘millennial’, ‘retail’, ‘park’, or even ‘bar’, say little about the real culture of a place. In our ongoing quest to construct a holistic understanding of location, we continue to search for more nuanced ways of understanding every aspect of what makes a particular neighborhood or street corner unique, whether that’s the way pizzas are topped or the architecture of the surrounding buildings. We believe this granular view unlocks radically new ways of understanding and recommending locations, whether that’s algorithmically determining the Bushwick of Boston, narrowing the search for a new apartment, or identifying the next ideal neighborhood for an expanding retailer.

This post is part of an ongoing series capturing different insights we generate while developing our platform. We would love to hear your feedback. If you enjoyed this article please share and 👏 a few times so other people can see it too.

To learn more about how we are transforming location intelligence, please contact us: info@topos.ai




Transforming the way we understand cities with Artificial Intelligence | @topos_ai