Finding the Best Chicago Pizza in San Francisco
Chicago prefers the strong; it’s architecture, winter cold, summer heat, and, most of all, the flavors. As a Chicago expat living in San Francisco, I was excited to see a Chicago style pizza joint open in my sleepy neighborhood. When I shared the news about Patxi’s new Noe Valley location with Daniel Raffel, a friend and also former Chicagoan, a debate about the best local deep dish pizza ensued. Daniel argued feverently for Little Star. Very soon we were at an impass, the result of a fiery conversation comparing the San Francisco ‘Chicago facsimiles’ to the likes of Lou Malnati’s, Ginos, and Giradanno’s. There was, in fact, one way to settle it: a show down.
By nightfall Daniel and I had mail-ordered par-cooked frozen pizzas from Chicago; his from Gino’s, mine from Lou’s. We set a date for the next week and found 10 taste testers from the local area. We also called upon the services of Dr. Elizabeth F. Churchill (a scientist and cognitive psychologist specializing in human-computer interaction) to design and run this experiment. You see, we did not want a contest. We wanted to see what people liked. How they felt. What they disliked. If they liked two pizzas the same, then they should be able to rate them the same. More so, we needed a fair experiment design.
On December 12th, we gathered our 12 testers in total. Entrepreneurs, product managers, engineers, scientists, designers, a pizzaiolo, a librarian and a lawyer rounded out our collection of people from around the bay area; half of these testers identified themselves as those who prefer New York pizza over Chicago. Three cheese Chicago-style pizzas were brought in locally from: Little Star, Patxi’s, and Zachary’s. The 2 shipped frozen and reheated from Chicago were Lou’s and Gino’s. All these pizzas were of Chicago Deep Dish style—not the “stuffed” variety. Dr. Churchill, who was not taking the test herself, designed a ‘pizza score card’ to measure peoples judgement on each pizza on a 1-5 scale: Appearance, Smell, Texture, Taste, Finish, and Overall-loving-feeling. Each pizza was presented unbranded on a standard pizza board labeled only with a number. Each taster drew numbers from a hat to determine the order in which they tasted pizzas. This helps to control the error of the pizzas getting cold as they sat out.
As each taster sampled a slice, they rated the pizza numerically on the dimensions listed and left write-in comments—a qualitative measure. Some did this in a more controlled manner than others.
Tasters were informed to keep their conversations and discussions to a minimum; focus only on your pie pieces. Bowls of coffee beans were avaliable for resetting nosmic/smell sensitivity and a selection of beverages including beer/champagne/water and Mexican Coca-Cola were on offer for palate cleansing.
Once each person was finished, one task remained. Each tester received a small amount of Hanukkah gelt to put next to the pizza’s they liked or not distribute at all for an ‘overall ranking’. Scientifically speaking, this is used to verify the validity of the detailed score cards.
Eating this much pizza is no easy task. Each taster finished approximately 1/2 of a 14” Chicago pizza. This left most of us docile and motionless while Dr. Churchill and Daniel did a quick data study. They presented us with their preliminary finding: without a doubt, Patxi’s was preferred the most by the 12 participants.
Later, I statistically examined the distribution of ratings. Hands down, Patxi’s emerged on top. In fact, the worst Patxi’s rating received was about the average rating of the other San Francisco pies.
Here, above, is the distribution of ratings for each pizza as a box plot. This shows the ‘bulk’ of the 12 testers ratings (the range within the boxes, called the IQR). The lines, or whiskers, that extends out shows you the overall range of scores. For example, Lou’s received scores from 1 to 5, with the bulk of scores between 2 and 3.25. The line in the middle of the box (at 2.25) shows the median score for Lou’s. Patxi’s received the best score distribution with a IQR between 3.874 and 4.625. Its lowest score was 3.5 and the highest score was a 5 (given by 3 testers). Zachary’s was generally favored next (with a median score of 4.0). Little star follows third (last in the San Francisco facsimiles).
The Chicago par-cooked and frozen pizza’s didn’t fare as well in the averages by the testers. Lou’s had a slightly higher average (2.625) versus Gino’s (2.458) although Gino’s showed a higher median. Not that one was significantly better than the other. This is perhaps an experimental confound—being frozen and then reheated does not make for a fair test; the frozen shipped pizzas from Chicago didn’t perhaps show Lou’s and Gino’s in their best light.
Oh, we didn’t forget about the gelt. It showed the same ranking: Patxi’s, Zachary’s, Little Star, Lou’s, and Gino’s. Yay Science! You can check out all the photos of the madness for yourself on Flickr.
Now, of course, there are some more sophisticated methods of analysis. Perhaps we could model a simulation? Or develop a scoring function across the features? Or dive into the qualitative data? We’re working on it. If you wanna help out, you’re welcome to throw our data set (get the comma separated value file) at your favorite stats program. We’ll update it later once we get the write-in responses. In the meantime, congrats Patxi’s!
