Red Bull Predictive Modeling

Scenario:

You are an analyst at Red Bull. Your goal is to use predictive modeling to show which online teams’ channel should get a $100,000 bonus. You have been given data from each of the 7 subcategories of the online campaigns.

 The 7 categories are:

● Banner ads

● Facebook

● Instagram

● E-zine

● TV ads

● Twitter

● YouTube

Using the given excel sheets, complete the following to do list:

1. Run a Simple Linear Regression for EACH subcategory.

Run a regression on each data file in Excel:

§ Download the Red Bull files.

§ Run a regression (DV: sales). For example, for the Facebook file

o Click the Data tab>Data Analysis>Regression.

o Select the sales column as Input Y Range and the Facebook column as Input X Range.

§ Determine the predicted sales amount for spending $100.

§ Present your analysis in a profession way supporting your results with data visualization tools. Who should win the bonus?

2. Log Transform Sales for EACH category

Plot your simple linear regression.

§ Select Layout>Trendline>Linear Trendline

§ Create a new column with log sales

o In the first row of the new column type “=LN(“, then select the first row of the sales column. Click “Enter.”

o Drag that box down the entire column to calculate all log sales.

§ Run another regression but with log sales as Input Y Range instead of sales.

§ Plot the new regression.

§ Present your analysis in a profession way supporting your results with data visualization tools. Who should win the bonus?

3. Run a Multiple Regression

§ Open the redbull.xlsx file, which holds all of the subgroups’ data.

§ Run a multiple regression using all of the variables as controls. § This is the same process as before, but this time, select all of the columns besides sales for Input X Range.

§ Present your analysis in a profession way supporting your results with data visualization tools. Who should win the bonus now? Explain?

 4. Remove variables from the Multiple Regression Red Bull model based on:

1. Does it improve adjusted R2 ?

2. Does it have an insignificant p-value?

3. Is it associated with the response variable by itself?

4. Is it strongly associated with another explanatory variables? (If yes, then including both may be redundant.)

5. Does common sense (face validity) say it should contribute to the model?

6. What would you eliminate from the model?

§ Present your analysis in a profession way supporting your results with data visualization tools. Explain the results from each step.

redbull