Chili Wars

The Cincinnati Chili Wars

I’m a native of Cincinnati. I love everything about the town… especially chili. There are two dominant players in town - Skyline and Gold Star - polarizing brands and fanbases. I had Gold Star Chili as a client once, and know that they try to keep up with Skyline, but struggle to do so. While the main competitor - the general notion is that Skyline is “better” than Gold Star. I wanted to look at how Twitter talks about the brands and see if this is the case.

I imported 5000 rows of data from Twitter, focusing on a 6 month range (April-Sept), using both brands Twitter handles as the search criteria. There was far heavier volume of tweets for Skyline vs. GoldStar, so I combined the rows for both brands into one, then limited the rows by the oldest overlap date - which ended up being ~ April 1. I then removed roughly 70 columns that I wasn’t going to use in this analysis, and wrote as a csv.

Which brand tweets more frequently?

The primary variables to look at will be count of tweets and time, for each brand. I want a quick visual for easy comparison. I do not want to exclude retweets & replies, because an important part of a brand’s social media accounts is to build community - I’m not solely interested in just organic tweets. Since this is a six month period, I don’t want to look at days (will be too hectic) and I don’t want to group into months because it will be more difficult to read spikes - Weeks instead will be ideal.

I used a time series plot. I could have done a longer workaround on this, but it was a new approach and felt like it could simplify things. I did pull a group_by function into this, because I wanted to separate out the brands - I wasn’t interested in a total frequency over time. This grouping allowed me to compare performance. Other than that, it was about making the visual more readible.

Interpretation: Going into this, I knew Skyline had a high vollume of tweets since I had to clean the data. But, I didn’t realize the high variance on Skyline. Some weeks going around the 50 tweets range, while others are closer to 200. Digging in further, Skyline certainly has a large number of organic tweets, but they are also very involved in community management, which increases volume as replies.

Which brand is more popular?

Just tweeting alot, doesn’t necessarily mean your brand is more popular. I wanted to look at one of the closest Twitter metrics for popularity - favorite count, across both brands. Essentially each favorite is an indicator of one person finding that content worth praising(liking) Looking at this on a weekly basis felt like too much, so I backed into a monthly view. And used a bar chart. I started by mutating month against each row - so that we could easily see tweets against this parameter. I then placed that new “month” column as the x axis, and y was favorite count = tweet popularity by month. I decided to use a stacked bar here, where the stack was the screen_name (which brand it was).

Interpretation: Well… Not only does Skyline have more tweets (as we learned in visual 1), but they disproportionately higher in favorites compared to Gold Star. Skyline’s top favorited tweets take advantage of topical news, like covid, reds, and bengals. But.. the most favorited tweet out of all of these the past six months is the simplest plea for favorites… and it worked @skyline_chili said “f you’re the biggest Skyline fan to have walked the planet, like this tweet.” and they got 4700 favorites haha. Really though they are just more active and talking about things that their fanbase finds more interesting, as well as having a larger following, and more tweets. Hard for Gold Star to be close given that.

Do retweets have a relationship with favorites?

We know that Skyline wins the popularity contest for favorites, and instead of showing the same chart above for retweets, could we look at a correlation between favorites and retweets? Arguably, a retweet is more valuable than a favorite, because it exposes a message to a different audience. To do this relies mostly on our x and y : favorites and retweets. Just from cleaning the data, there are alot of tweets with very few favorites and retweets - so to control for that, need to adjust scales. Additionally, i want to add a regression. I

Interpretation: I initially ran GoldStar separate from Skyline, but it showed a similar pattern. So reading this as a total is valuable for both brands. There is a strong positive correlation between favorites and retweets for the brands. Favorites are a means to an end (retweets). If I was to advise GoldStar - I would recommend they focus solely on favorite driving tactics, as that will help grow their reach, and ultimately their audience acquisition for the platform.

#Summary We’ve seen that on Twitter, Skyline Chili is more active that Gold Star. They are heavily involved in community management as well as organic tweets. Additionally, their content drives favorites far exceeding that Gold Star. As far as chili brands tweets (gold star and Skyline), favorites are correlated to retweets.

tl;dr - Skyline dominates Cincinnati Chili on Twitter