Introduction
Cameo is a website where people can pay celebrities to record a personalized video that’s roughly a minute long. Many celebrities have signed up for Cameo from sports players to reality tv personalities to various cast members from the Office. If you’ve had a cameo on Friends, you can have a Cameo account. One of the most challenging parts of Cameo for the celebrities (besides keeping their energy up to record 10 birthday videos back-to-back) is figuring out what price to charge for their videos.
The prices vary dramatically from celebrity to celebrity. For example, it’ll only cost you pay $40 for the actress that played Paige Michalchuk in the Canadian teen melodrama, Degrassi: The Next Generation. Or you can pay $1200 for Kevin O’Leary aka that winemaker and photographer from Shark Tank that calls himself, Mr. Wonderful.
My goal is to help celebrities figure out just how much a minute of their fifteen minutes of fame is worth.
Data
For this project, I will be using scraped data from two sources:
1. Cameo.com — The website where you can purchase videos from celebrities
2. FameFlux.com — A website that ranks celebrities and companies by their online popularity
The Cameo data was scrapped from January 21st, 2021, and contains the name of the celebrity, the cameo video price, and the star rating for the celebrities. Fame Flux includes all the data below that begins with FF including the full name of the celebrity, the celebrity’s celebrity rating. It was also gathered on January 21st, 2021.
Data source: https://www.netcredit.com/blog/cost-effective-celebs-on-cameo/
Cameo price is the price a customer has to pay for the ~1minute video from the celebrity. This will be the column I’m trying to predict in my model. The variables I’ll be using to try and accurately predict the Cameo price are:
- FF Celebrity rating – is the ranking of the celebrities compared to another. A ranking of 1 would be the most famous of celebrities. The higher number, the less famous the celebrity is.
- FF Main category rating – is the rating of the celebrity within their category. For example, if the person is a musician then they might have a general ranking (FF Celebrity ranking) of 202, but a musician category ranking (FF Main category rating) of 90.
- Wikipedia Traffic\n(Avg. daily) – is the daily number of visitors that go to the celebrity’s Wikipedia page per day on average.
All of these indicate how famous a particular person is to the general population. Each of these numeric values will be used to determine the potential price a celebrity should charge for a Cameo.
Methodology
First, I explored the data using histograms. Looking at the Celebrity Rating, Main Category Rating, and Cameo Price. Each histogram made it apparent that there was a non-normal distribution of data and the data was largely skewed to the right.
In order to see if there were any stronger correlations, I one-hot encoded the data. Luckily this didn’t make the data Too Hot to Handle. For $50 you can get Sharron from Too Hot to Handle to explain One-hot encoding to you, but I’ll do it here for free. I changed the category of the celebrity, like musician, chef, or actor to a matrix of binary values. For each category, a new column was created and for the category column that matched the celebrity’s denoted column, the value was set to one. For all other category columns, the value was set to zero. One-hot encoding makes it easier for categorical variables to be used in the various types of regression.
Since the price of a cameo is a continuous variable, I initially explored using linear regression to predict the price, however, I realized this was not the best option after analyzing the data using single variable linear regression. Both Celebrity Rating and Main Category Rating had a weak correlation with Cameo Video Price. Furthermore, Wikipedia daily traffic had a stronger correlation, but upon plotting the data it was clear that most of the data points did not follow this correlation and the strong correlation was tied to a few outliers in the data. Additionally, I tried using non-linear regression using polynomial regression and the highest R-squared value was below 0.20. A low R-Square value means that the data points don’t fit very well on the regression line. If the data wanted to get more fit they could hit up Peloton Girl on Cameo – yes the actress that plays the wife in that sad, upsetting Peloton Christmas ad.
It was clear after plotting the data on a scatter plot that there was just a large clump of data, which I thought K-means might be the best approach. I also thought it would be the method the Kardashians would use, and I desperately want to keep up with them.
Since K-means clustering doesn’t work with continuous variables, I created Cameo video price ranges based on how I saw the prices be distributed and then created a new column based on the video price ranges. The ranges are:
- $0.00-5.00
- $5.01-10.00
- $10.01-25.00
- $25.01-50.00
- $50.01-100.00
- $100.01-250.00
- $250.01-500.00
- $500.01-1000.00
- $1000.01-5000.00
Then to make the new cameo price groups work for KNN, I converted the cameo price groups to integers. Once all the data was converted to integers, I used data standardization to give the data zero mean and unit variance. Then I randomly split the data into a training dataset (80% of the initial dataset) and a testing dataset (20% of the initial dataset).
After that I used the elbow method, which would be my signature move in the WWE if I could fight against it to determine the best value of K. The best value of K was determined to be 15 with an error rate of 0.392.
Results
For the final results, I joined the predicted values to see the predicted and actual cameo price ranges.
Based on the k-means clustering, the celebrities that are predicted to have the highest cameo video prices include:
- Snoop Dogg – At first I was surprised he had the time to make cameos between his football coaching and hanging out with Martha Stewart. But then I watched a few of the cameos and it looks like he does them as he smokes pot. Actual was $1200
- Caitlyn Jenner – Caitlyn Jenner being on the app was not a shock. She has time for Masked Singer she definitely has time for a Cameo. Actual was $2500
The celebrities that are predicted to have the lowest cameo video prices ($0-$5) include:
- Chase Crawford – This was really surprising until I realized it’s not THAT Chase Crawford. This actor has never appeared in Gossip Girls or The Boys. But he might be able to give you some gossip while he’s out with the boys. Actual was $1.00
- Ben Robinson – Who is a man on TikTok…but he will create a cameo about how to cook a pork roast at the correct temperature. Actual was $4.
Discussion
Because Cameo video prices are chosen by the individual celebrity, it’s a difficult price to predict even when predicting within a range of cameo video prices. A celebrity might be delusional about how famous they are and price themselves far above what most people would pay for them. Additionally, a celebrity might be particularly famous to one subculture or niche community. Those within the community would be willing to pay a large sum of money for the celebrity. However, their Fame Flux celebrity ratings and Wikipedia page daily averages may not reflect these zealous followers.
One way I could potentially cut through this confusion in the future is by importing data from other sources like social media sites. Adding data from social media sites like Instagram, TikTok, and Twitter will allow the clusters to become more accurate.
Additionally, I performed the data analysis based on a static data pull from cameo and FameFlux. The fame ratings and Cameo video prices are all dynamic. They fluctuate as celebrities start receiving a backlog of videos to record or they are recently in the news. In the future, this data could be accessed dynamically using daily web scraping tools.
Conclusion
Cameo Prices are as fickle as the fifteen minutes of fame that lead you to be able to sell Cameo videos. Celebrities and random dudes on TikTok can better price their videos based on existing celebrity prices. That way they can spend the hours they would spend deciding on their cameo price doing something truly useful, like dressing up as a robot porcupine and singing.
Want to see the final results of the predictions click here to see the final version in Google Sheets.