1. Using R for Marketing Research Dan Toomey 2/23/2015 dan@dantoomeysoftware.com

2. Outline Overview of R – Language Basics – Basic Operations – Tools – R, R Studio, others Validate Data Analyze Ad Campaign Effectiveness Sales Impact Drivers Determine Optimal Pricing

3. Language Basics Uses ‘library’s of open source code for specific applications Mostly statistics processing Simple syntax Free tools to use Multiple platforms

4. Basic Operations Load data – variety of formats Compute metrics – Again, mostly interested in statistics Display metrics – Textual – Graphical

5. Load Data ># libraries used >library(s20x) >library(car) ># load and display data >df <- read.csv("http://www.dataapple.net/wp- content/uploads/2013/04/grapeJuice.csv",header=T) >head(df) sales price ad_type price_apple price_cookies 1 222 9.83 0 7.36 8.80 2 201 9.72 1 7.43 9.62 3 247 10.15 1 7.66 8.90 4 169 10.04 0 7.57 10.26 5 317 8.38 1 7.33 9.54

6. Validate Data >par(mfrow = c(1,2)) >boxplot(df$sales,horizontal = TRUE, xlab="sales") ># histogram to explore the data distribution shape >hist(df$sales,main="",xlab="sales",prob=T) >lines(density(df$sales),lty="dashed",lwd=2.5,co l="red")

7. Data Validation No outliers from box plot Normal distribution

8. Ad Effectiveness > sales_ad_nature = subset(df,ad_type==0) > sales_ad_family = subset(df,ad_type==1) > # graph the two > par(mfrow = c(1,2)) > hist(sales_ad_nature$sales,main="",xlab="sales with nature production theme ad",prob=T) > lines(density(sales_ad_nature$sales), lty="dashed",lwd=2.5,col="red") > hist(sales_ad_family$sales,main="",xlab="sales with family health caring theme ad",prob=T) > lines(density(sales_ad_family$sales),lty="dashed",lwd=2.5,col="red")

9. Ad Effectiveness Normal distributions

10. Test Null Hypothesis(Normal) >shapiro.test(sales_ad_nature$sales) Shapiro-Wilk normality test data: sales_ad_nature$sales W = 0.9426, p-value = 0.4155 >shapiro.test(sales_ad_family$sales) Shapiro-Wilk normality test data: sales_ad_family$sales W = 0.8974, p-value = 0.08695

11. Test Means > t.test(sales_ad_nature$sales,sales_ad_family$sales) Welch Two Sample t-test data: sales_ad_nature$sales and sales_ad_family$sales t = -3.7515, df = 25.257, p-value = 0.0009233 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -92.92234 -27.07766 sample estimates: mean of x mean of y 186.6667 246.6667

12. Sales Drivers > pairs(df,col="blue",pch=20) > pairs20x(df)

13. Regression Model > sales.reg<-lm(sales~price+ad_type+price_apple+price_cookies,df) summary(sales.reg) Call: lm(formula = sales ~ price + ad_type + price_apple + price_cookies, data = df) Residuals: Min 1Q Median 3Q Max -36.290 -10.488 0.884 10.483 29.471 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 774.813 145.349 5.331 1.59e-05 *** price -51.239 5.321 -9.630 6.83e-10 *** ad_type 29.742 7.249 4.103 0.000380 *** price_apple 22.089 12.512 1.765 0.089710. price_cookies -25.277 6.296 -4.015 0.000477 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 18.2 on 25 degrees of freedom Multiple R-squared: 0.8974,Adjusted R-squared: 0.881 F-statistic: 54.67 on 4 and 25 DF, p-value: 5.318e-12

14. Test Data Check distribution of residuals

15. Multicolinearity > vif(sales.reg) price ad_type price_apple price_cookies 1.246084 1.189685 1.149248 1.099255

16. Price Elasticity Sales = 774.81 – (51.24 * price) + (29.74 * ad_type) + (22.1 * price_apple) – (25.28 * price_cookies) PE = (ΔQ/Q) / (ΔP/P) = (ΔQ/ΔP) * (P/Q) = -51.24 * 0.045 = -2.3 P/Q = 9.738 / 216.7 = 0.045 The PE indicates that 10% decrease in price will increase the sales by 23%

17. Optimal Pricing Substituting means for price changes in above model we come down to Sales = 772.64 – 51.24*price estimate profit(Y) = (price – C) * Sales Quantity = (price – 5) * (772.64 – 51.24*price) reduces to Y = – 51.24 * price2 + 1028.84 * price – 3863.2 > f<-function(x) { result <- -51.24*x^2 + 1028.84*x - 3863.2 return(result) } #have R optimize the function >optimize(f,lower=0,upper=20,maximum=TRUE) $maximum [1] 10.03942 $objective [1] 1301.28

18. Predict Sales >inputData <- data.frame(price=10,ad_type=1,price_apple=7.6 59,price_cookies=9.738) >predict(sales.reg,inputData,interval="p") fit lwr upr 1 215.1978 176.0138 254.3817

19. References Analyze sales pricing http://www.r- bloggers.com/data-analysis-for-marketing- research-with-r-language-1/