Marketing Analytics Group Project
Marketing Analytics Group Project
Marketing Analytics Group Project
Marketing Analytics
Q1. Do different promotions differ in their effectiveness across whole market? Which promotion
do you recommend if you have to select one promotion across whole market? (You are expected
to answer this question treating whole market as a Single Market).
Ans-
Descriptives
Sales in thousands
95% Confidence Interval for
Mean
N Mean Std. Deviation Std. Error Lower Bound Upper Bound Minimum Maximum
1 172 58.10 16.551 1.262 55.61 60.59 31 100
2 188 47.29 15.127 1.103 45.12 49.47 17 89
ANOVA
Sales in thousands
Sum of Squares df Mean Square F Sig.
Between Groups 11532.133 2 5766.066 22.076 .000
Within Groups 142347.634 545 261.188
Total 153879.766 547
Multiple Comparisons
Dependent Variable:
Sales in thousands
Tukey HSD
Mean 95% Confidence Interval
(I) Three different types (J) Three different Difference (I- Std. Lower Upper
of promotion types of promotion J) Error Sig. Bound Bound
1 2 10.806 *
1.705 .000 6.80 14.81
3 2.737 1.705 .244 -1.27 6.74
2 1 -10.806 *
1.705 .000 -14.81 -6.80
3 -8.069 *
1.667 .000 -11.99 -4.15
3 1 -2.737 1.705 .244 -6.74 1.27
2 8.069* 1.667 .000 4.15 11.99
*. The mean difference is significant at the 0.05 level.
Interpretation -
To assess the market-wide impact of three different methods of promotion one way ANOVA is
used. In these sales (ratio) is dependent and promotion (nominal) is independent. This test does
not take the market's size into account.
The descriptive table shows the average (mean) for each type of promotion.
P1 - 58
P2 - 47
P3 - 55
This indicates that the three promotions had different effects on sales. When the significance
between the groups in the ANOVA table is checked, it is less than 0.05, indicating a significant
difference between the three promotions.
We can observe from the multiple comparisons table that Promotions 1 and 3 do not differ
significantly from one another, however Promotion 2 differs from 1 and 3 because the
significance threshold is less than 0.05.
We can infer from the boxplot graph that 1 and 3 are close to and more effective than 2, which
has the lowest effectiveness. Therefore, promotion type 1 would be proposed if only one sort
of promotion was needed. The graph indicates that it is most effective.
Q2. Does the effectiveness of promotion type varies across different weeks?
Ans-
Between-Subjects Factors
N
Three different types of 1 172
promotion 2 188
3 188
Week of a month 1 137
2 137
3 137
4 137
As
a result, we performed a two-way ANOVA in which the sales volume was the dependent
variable and the week and the promotion were the independent variables. For the first
promotion, we only have 172 data; for the second and third promos taken together, we have
188 data. A month consists of 137 weeks in total.
Dependent Variable:
Sales in thousands
DESCRIPTIVE STATISTICS
Three different types of
promotion Week of a month Mean Std. Deviation N
1 1 58.26 16.843 43
2 57.00 16.541 43
3 58.70 16.421 43
4 58.44 16.930 43
Total 58.10 16.551 172
2 1 47.72 14.404 47
2 47.53 14.845 47
3 47.66 15.962 47
4 46.26 15.689 47
Total 47.29 15.127 188
3 1 55.81 18.112 47
2 55.91 15.665 47
3 54.38 17.227 47
4 55.34 16.560 47
Total 55.36 16.789 188
Total 1 53.80 17.005 137
2 53.38 16.131 137
3 53.43 17.042 137
4 53.20 17.072 137
Total 53.45 16.772 548
Q3. Will you suggest use of different promotions across different market sizes (small, medium and
large)? Which promotion type do you recommend on the basis of size of the market (Use interaction of
‘promotion’ and ‘marketSize’ to answer this question)?
Ans-
ANOVA
Sales in thousands Sum of Squares df Mean Square F Sig. Bayes Factora
Between Groups 11532.133 2 5766.066 22.076 .000 2894978.503
Within Groups 142347.634 545 261.188
Total 153879.766 547
a. Bayes factor: JZS
Descriptive Statistics
Dependent Variable: Sales in thousands
Three different types of
promotion MarketSize1 Mean Std. Deviation N
1 SMALL 60.10 5.046 20
MEDIUM 47.66 8.048 96
LARGE 75.29 15.442 56
Total 58.10 16.551 172
2 SMALL 50.75 5.802 16
MEDIUM 39.07 8.807 108
LARGE 60.30 15.776 64
Total 47.29 15.127 188
3 SMALL 59.58 5.266 24
MEDIUM 45.46 8.112 116
LARGE 77.19 14.464 48
Total 55.36 16.789 188
Total SMALL 57.40 6.631 60
MEDIUM 43.96 9.057 320
LARGE 70.12 17.076 168
Total 53.45 16.772 548
Q4. Which of the variables (MarketSize, promotion, AgeOfStore, Week) have more influence on sale
(SalesInThousands).
Ans-
Variables Entered/Removeda
Variables Variables
Model Entered Removed Method
1 MarketSize1, . Enter
dw3, dp1, Years
of existence of
store, dw2,
dummy_promoti
on2b
a. Dependent Variable: Sales in thousands
b. Tolerance = .000 limit reached.
Model Summaryb
Adjusted R Std. Error of the
Model R R Square Square Estimate Durbin-Watson
1 .546a
.298 .290 14.131 .330
a. Predictors: (Constant), MarketSize1, dw3, dp1, Years of existence of store, dw2,
dummy_promotion2
b. Dependent Variable: Sales in thousands
ANOVAa
Model Sum of Squares df Mean Square F Sig.
1 Regression 45853.229 6 7642.205 38.272 .000b
Residual 108026.537 541 199.679
Total 153879.766 547
a. Dependent Variable: Sales in thousands
b. Predictors: (Constant), MarketSize1, dw3, dp1, Years of existence of store, dw2,
dummy_promotion2
Coefficientsa
Standardized
Unstandardized Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) 28.670 2.742 10.455 .000
dw2 -.161 1.479 -.004 -.109 .914
dw3 -.394 1.479 -.010 -.267 .790
dp1 -11.380 1.492 -.322 -7.628 .000
dummy_promotion2 -1.757 1.495 -.050 -1.175 .240
Years of existence of store .089 .092 .035 .964 .335
MarketSize1 13.049 .999 .478 13.060 .000
a. Dependent Variable: Sales in thousands
Excluded Variablesa
Collinearity
Partial Statistics
Model Beta In t Sig. Correlation Tolerance
1 dw1 . b
. . . .000
a. Dependent Variable: Sales in thousands
b. Predictors in the Model: (Constant), MarketSize1, dw3, dp1, Years of existence of store,
dw2, dummy_promotion2
Residuals Statisticsa
Minimum Maximum Mean Std. Deviation N
Predicted Value 30.03 69.96 53.45 9.156 548
Residual -31.903 32.004 .000 14.053 548
Std. Predicted Value -2.558 1.803 .000 1.000 548
Std. Residual -2.258 2.265 .000 .995 548
a. Dependent Variable: Sales in thousands
Th
e aforementioned regression model evaluates the strength of the association between sales, a
dependent variable, and independent variables like market size, the number of years the store
has been open, and dummy variables representing various types of sales promotions.
Multiple regression may only be used when the dependent variable has a ratio/interval scale, as
in the model above, but the independent variable, such as the sales promotion, is nominal in
nature. To accommodate this, dummy variables were generated.
As a result of performing multiple regression, we can observe that the Adjusted R square value
is only 0.290, indicating that the independent variables are not affecting sales in any way.
The coefficient table also leads us to the conclusion that the store's sales are impacted by the
size of the market. Therefore, in order to maximize sales and reduce losses, the sales manager
of the individual stores would like to expand their markets.