Using Regression to Determine the Impact of Media Placements on Net Sales

A client wants to launch an ad campaign. He has already made a similar campaign in the past and now wants to leverage historical data to assess the impact of  his ads before making new investments.

What data are going to use?

A spreadsheet with 21 rows and the ff. columns for our purpose:

1. Broadcast - number of placements in broadcast media
2. Out-of-Home - number of out-of-home ads
3. Print - number of ads in print
4. Net.Sales - net of sales 

Here's how it looks:

Short Answer:

I already ran a regression analysis on Tableau and found that broadcast has the most impact on net sales. 
Coeff: (Intercept) - 133108.8 Broadcast - 12141.9 P-value - 1.95e-10 Adj. R Sqaured - 0.8944

But let's try it using R.

R Code:

	> broadsales <- read.csv(file.choose())
	> head(broadsales)
	DATE GEO OUT.OF.HOME PRINT BROADCAST NET.SALES UNIQUE.WEBSITE.VISITORS SOCIAL.MEDIA.VOLUM SALES.INCREASE
	1 2/13/17 Kentucky 6 5 1 137452 3974 4 -3
	2 2/12/17 NI 8 6 1 147376 3732 8 7
	3 3/7/17 Virginia 1 1 2 148252 3127 5 2
	4 3/14/17 SIC 9 5 3 151842 3091 3 7
	5 2/5/17 DE 6 3 3 155207 4791 5 8
	6 1/3/17 IN 5 4 3 163855 5066 6 5
	U.W.VISITORS.INCREASE SOCIAL.MEDIA.VOLUME.INCREASE
	1 -2 0
	2 -3 7
	3 0 -3
	4 2 5
	5 7 -2
	6 -3 0
	> attach(broadsales)
	fit <- lm(NET.SALES~OUT.OF.HOME+PRINT+BROADCAST)
	> summary(fit)
	Call:
	lm(formula = NET.SALES ~ OUT.OF.HOME + PRINT + BROADCAST)
	Residuals:
	Min 1Q Median 3Q Max
	-18114.1 -8083.5 -712.5 8061.8 17124.8
	Coefficients:
	Estimate Std. Error t value Pr(>|t|)
	(Intercept) 132643.92 8859.37 14.972 7.86e-11 ***
	OUT.OF.HOME 126.76 1015.82 0.125 0.902
	PRINT -42.76 1072.75 -0.040 0.969
	BROADCAST 12128.39 1070.40 11.331 4.70e-09 ***
	---
	Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
	Residual standard error: 11870 on 16 degrees of freedom
	Multiple R-squared: 0.9001, Adjusted R-squared: 0.8813
	F-statistic: 48.04 on 3 and 16 DF, p-value: 3.169e-08
	#BROADCAST is the only channel with a low P-value
	> plot(BROADCAST, NET.SALES)
	> abline(NET.SALES~BROADCAST)

view raw advertising mult. regression hosted with ❤ by GitHub