We Asked Google To Explain Its Bonkers Super Bowl Food Map — And They Did
Like almost everyone else, we did a double take when we saw Google's 2019 "Super Bowl Foods" map.
Paella in Maine? Gluten-free pretzels in Massachusetts? Lentil soup in Montana? "Pea and peppercorn mash"? What even is that? I am an Official Food Journalist and I have never seen this dish although I suspect it's delicious because there's a whole state of people — New Mexico — searching for it.
In California, our top pick was supposedly "baked chicken breasts." On Twitter, someone tried to blame "mommy bloggers." Ummm, no. I'm a mom and a writer. I have many friends who are also moms and writers. I can assure you, I have never once in the history of ever (and I'm old!) seen "baked chicken breasts" served at a Super Bowl party. You can blame mothers for everything else (you're going to anyway) but do not blame us for this.
What's the deal? Either A) people are lying, B) the Google Trends team has gone bananas, or C) I am misunderstanding something.
I am not a data scientist, but in my role as an Official Food Journalist, I decided to ask Google about it. To my delight, a spokesperson for one of the most powerful corporations in the world took time to explain how they came up with this bizarre map.
First, I straight-up asked: "Is Google Trends just trolling America with this map?"
"We are not :) but we understand how stories that mislabel these as 'top searches' or 'most searched' can create confusion. We do our best to explain this, but the methodology doesn't always make it into stories," a Google spokesperson replied, via email.
Here's the deal, the reason the map looks so kookoopants is because it DOES NOT depict the most frequently searched (and, presumably, the most popular) Super Bowl foods. It's showing "uniquely searched" foods.
"The reason we focus on uniquely searched terms v. top searched terms (which would be an absolute measure of search interest) is that top searches would likely not vary much state by state," Google's spokesperson writes.
The top Super Bowl food searches would be what you'd probably expect.
HOW THEY DID IT
Google wanted to do something different. For the week of January 17 - 24, they looked at all searches associated with the topic "Super Bowl," which accounts for the various ways people might search for the game, then honed in on the Food & Drink category to identify the most uniquely searched terms in each state.
GREAT. NOW EXPLAIN IT LIKE I'M A CHILD
I literally asked, "In very simple layperson's terms, like simple enough for a 12-year-old, can you further clarify the difference between 'uniquely searched' vs. 'top searched' or 'most searched'?" Here's the Google spokesperson's reply.
I've pasted it in full, below, because I'm still reading it.
All Google Trends data is indexed on a scale of 1-100, with 100 being the highest level of interest in a topic or keyword over a given time period.
Let's use the broad category of recipes as an example. If you looked at top searched recipe topics, "chicken" recipes would likely be near the top across the country, because chicken recipes are very popular by volume of searches, both at the national and the individual state level. So if you created a map, it could be something like 48 (or even 50!) states of chicken — not that interesting.
For illustration purposes, let's say nationally chicken recipes have a search interest of 100, and they also have a search interest of 100 at each state level.
Now, let's say you wanted to understand what's unique to a state like California. While chicken recipes may be "top searched" in California, most people nationally are searching for chicken recipes — they're popular here, but they're also popular everywhere, so they're not unique.
To determine uniquely searched recipes, we look at searches that people in California do proportionally much more than they do around the country. For example, perhaps we search for avocado recipes more than the country at large. Again for illustration purposes, let's say in California avocado recipes have an indexed search interest of 70, whereas the U.S. has a search interest of 20. That search would be more unique to CA. To get at these lists, we look for queries with the highest search interest unique to an individual state.
Here's simpler analogy that illustrates the above. You have two groups: one has 1000 people, the other has 100. Each group does a handful of searches for recipes. Across both groups — chicken is by far the most popular. 900/1000 people search for chicken in one group, 90/100 in the other. Proportionally, interest in chicken is equal across both groups. (A map of these two groups' top search interest would just be two squares of "chicken".)
Now let's say that in the smaller group, 70 people search for avocado, and 200 people in the larger group do. Even though there are a larger absolute number of people in the larger group that search for avocado, there's a disproportionate amount of interest in the smaller group. If you looked at the proportion of avocado interest overall across both groups (270/1100) v. the smaller group (70/100), you notice that "avocado" is a unique search interest to that smaller group compared to everyone. This is the same principle as unique state interest compared to overall U.S. interest.
Okay, I think I get it.
"The methodology is what accounts for this," Google's spokesperson explains. "Perhaps this tells us that people in Maine are more likely to make paella on Super Bowl Sunday than anywhere else in the country... We're hopeful that if people searched for these recipes, that they found one they liked and were able to make it, regardless of the occasion."
But seriously, paella in Maine?