srikanth banerjee ha2 imt-2013

Upload: srikanthiitb

Post on 04-Jun-2018

230 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    1/32

    Marketing AnalyticsAssignment - 2

    Group 13

    M V S Srikanth (1301-108)Rupayan Banerjee (1301-186)

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    2/32

    Developing composites By applying causal logic, Bencare managers identified custom

    satisfaction, trust and value as key causal predictors

    Customer satisfaction - SAT1(inter1), SAT2(inter2), SAT3(inter3) SAT = (SAT1 + SAT2 + SAT3)/3 TrustIt has 2 categories

    Representatives- Rep17(Trust-Agent1), Rep18(Trust-Agent2),Rep19(Trust-Agent3), Rep20(Trust-Agent4)

    REP = (Rep17 + Rep18 + Rep19 + Rep20)/4

    Management practices - Trust-Comp1, Trust-Comp2, Trust-Com

    Comp4 MGT = (prac17 + prac 18 + prac19 + prac20)/4

    ValueIt has 2 categories

    Short Term Value - ST-VALUE1, ST-VALUE2, ST-VALUE3

    STV = (val1 + val2 + val3)/3

    Long Term Value - LT-VALUE1, LT-VALUE2, LT-VALUE3

    LTV = (val4 + val5 + val6)/3

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    3/32

    Contd. Dependent Variable: Loyalty

    Beh-Loyalty1, Beh-Loyalty2, Beh-Loyalty3, Beh-Loyalty4, Cog-LoyaLoyalty2 (Ignored Cog-Loyalty3, Cog-Loyalty4 as per our results froanalysis)

    LOY = (loy1 + loy2 + loy3 + loy4 + loy5 + loy6)/6

    Independent variables Dependent variables

    SAT

    LOYREP

    MGT

    STV

    LTV

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    4/32

    Test of AssumptionsCountry: Germany*

    Test of NormalitySkewness/Std Error < 3 i.e. Data is normal

    * Data set has been divided into 3 sub sets based on the countries

    Test of HomoscedastiNo pattern is observed.

    homoscedastic

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    5/32

    Contd.

    The partia

    have beenand regres

    difference

    than 0.02.

    considere

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    6/32

    Test of AssumptionsCountry: USA*

    Test of NormalitySkewness/Std Error < 3 i.e. Data is normal

    * Data set has been divided into 3 sub sets based on the countries

    Test of HomoscedastiNo pattern is observed.

    homoscedastic

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    7/32

    Contd.

    The partial corhave been fit w

    and regression

    difference in R

    than 0.02. The

    considered to

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    8/32

    Test of AssumptionsCountry: Holland*

    Test of NormalitySkewness/Std Error < 3 i.e. Data is normal

    * Data set has been divided into 3 sub sets based on the countries

    Test of HomoscedastiNo pattern is observed. S

    homoscedastic

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    9/32

    Contd.

    The partial chave been fi

    and regressi

    difference in

    than 0.02. Th

    considered t

    B li i d l

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    10/32

    Baseline regression modelCountry: Holland

    Loyalty = 5.122 + 0.336 *

    TRUSTAGENT + 0.550 * LTVALUE

    Country: USA

    Country: Germany

    Loyalty = 5.235 + 0.156 * SAT + 0.250

    * TRUSTAGENT + 0.308 *

    TRUSTCOMPY + 0.374 * LTVALUE

    Loyalty = 5.026 + 0.438 * SAT + 0.233

    * TRUSTCOMPY + 0.338 * STVALUE +0.215 * LTVALUE

    Id tif i i fl i th d t b C

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    11/32

    Identifying influencers in the data by CodistancesCountry: Holland Country: USA Count

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    12/32

    C i b li i d l

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    13/32

    Comparing baseline regression modelsbefore and after removing influencers

    Since, there is an significant increase in Adjusted R-square value after removing influencers, t

    considered after removing the influencers for further analysis. Also, the models were changed aft

    from the data set (Significant independent variables are changed)

    Country: Holland Country: USA Countr

    Before

    After

    Loyalty = 5.122 + 0.336 *

    TRUSTAGENT + 0.550 *LTVALUE;

    R-Square: 0.351

    Loyalty = 5.157 + 0.462 *

    LTVALUE;

    R-Square: 0.408

    Loyalty = 5.235 + 0.156 * SAT +

    0.250 * TRUSTAGENT + 0.308 *TRUSTCOMPY + 0.374 * LTVALUE

    R-Square: 0.721

    Loyalty = 5.235 + 0.156 * SAT +

    0.250 * TRUSTAGENT + 0.308 *

    TRUSTCOMPY + 0.374 * LTVALUE

    R-Square: 0.721

    Loyalty = 5.038

    * TRUSTAGENT

    TRUSTCOMPY +

    R-Square: 0.670

    Loyalty = 5.026

    * TRUSTCOMPY0.215 * LTVALU

    R-Square:0.624

    LTVALUE coefficient decreased

    by 0.088

    The effect of SA

    LTVALUE has be

    significant IV, TDifference No change

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    14/32

    Test of nonlinear relationships

    Since, we already proved that the data for all the 3 couhaving linear relationships, our baseline regression modechange because of nonlinear relationships

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    15/32

    Moderator regression modelCountry: HollandModerating variable: GenderIt has 2 categories:

    1Male ; 2FemaleCreated a dummy variable Dgender:

    1Male, 0FemaleThe R-

    signific

    interac

    DTAge

    signific

    significmodel

    Since,

    is not s

    interac

    be con

    has no

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    16/32

    Moderator regression modelCountry: GermanyModerating variable: GenderIt has 2 categories:

    1Male ; 2FemaleCreated a dummy variable Dgender:

    1Male, 0Female

    The

    sign

    Only

    term

    Contd

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    17/32

    Contd.

    Regression model:

    Loyalty = 4.887 + 0.18 * Dgender + 0.184 * ZSTVAL + 0.148 * Dgender * ZSTVAL

    For Male (Dgender = 1)Loyalty = 5.067 + 0.332 * ZSTVAL

    For Female (Dgender = 0)

    Loyalty = 4.887 + 0.184 * ZSTVAL

    Effect of short term value on loyalty in male category

    female category. (0.332/0.184 = 2 (approx.))

    So, Short term value is an important factor in case of

    category

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    18/32

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    19/32

    Moderator regression modelCountry: HollandModerating variable: AgeIt has 5 categories:

    118-24 yrs ; 225-34 yrs;335-44 yrs; 445-54 yrs; 555+ yrs

    Created 4 dummy variables: Dage1, Dage2, Dage3and Dage4

    Dage1: 1 Category 2; 0 Else

    Dage2: 1 Category 3; 0 Else

    Dage3: 1 Category 4; 0 Else

    Dage4: 1 Category 5; 0 Else

    The R-

    change

    signific

    has no

    effect.

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    20/32

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    21/32

    Moderator regression modelCountry: GermanyModerating variable: AgeIt has 4 categories:

    118-24 yrsNo cases in this age category;225-34 yrs; 335-44 yrs; 445-54 yrs; 555+ yrs

    So, number of valid categories: 4

    Created 3 dummy variables: Dage1, Dage2, Dage3and Dage4

    Dage2: 1 Category 3; 0 Else

    Dage3: 1 Category 4; 0 Else

    Dage4: 1 Category 5; 0 Else

    The R-

    change

    signific

    has no

    effect.

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    22/32

    Moderator regression modelCountry: HollandModerating variable: EducationIt has 4 categories:

    1High school, 2Some college, 3College,4Graduate school

    Created 4 dummy variables Dedu1, Dedu2, and Dedu3

    Dedu1: 1 Category 2; 0 Else

    Dage2: 1 Category 3; 0 Else

    Dage3: 1 Category 4; 0 Else

    The R-

    change

    signific

    educat

    interac

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    23/32

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    24/32

    Moderator regression modelCountry: GermanyModerating variable: EducationIt has 4 categories:

    1High school, 2Some college, 3College,4Graduate school

    Created 4 dummy variables Dedu1, Dedu2, and Dedu3

    Dedu1: 1 Category 2; 0 Else

    Dage2: 1 Category 3; 0 Else

    Dage3: 1 Category 4; 0 Else

    The R-

    change

    signific

    educat

    interac

    Final regression model for Bencare and

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    25/32

    Final regression model for Bencare andinsights

    Country: Germany

    Country: USA

    Country: Holland

    Loyalty = 5.157 + 0.462 * LTVALUE;

    R-Square: 0.408

    People in Holland are loyal to the company, when they see long term va

    the model built is having a low R-square value, it is advised to conduct regres

    much bigger sample

    Loyalty = 5.235 + 0.156 * SAT + 0.250 * TRUSTAGENT + 0.308 * TRUSTCOMPY + 0.374 *

    R-Square: 0.721

    Companys focus on Short term value can be reduced, as it is not affecting its

    Loyalty = 5.038 + 0.365 * SAT + 0.192 * TRUSTAGENT + 0.343 * TRUSTCOMPY + 0.289 * STVALU

    R-Square: 0.670Companys focus on long term value is not needed as it does not affect its customer

    Interaction of gender is significant and should be kept in mind

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    26/32

    Other key insightsIn all the three countries, Age and Education has no interaction effects.

    Consumers cannot be segmented as per age and Education for m

    loyalty

    Company can go for generic advertising i.e. not based on the ag

    In Germany, gender factor is significant in loyalty. The company

    male with high short term value to build their loyalty

    Consumer loyalty is independent of education. It means that co

    introduce special insurance plans for consumers with higher edu

    increase their loyalty

    This can lead to consumer thinking of Bencare when ever he

    insurance

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    27/32

    Thank you

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    28/32

    AppendixSynta

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    29/32

    Removing missing data:

    DATASET COPY BencareNoMissing.

    DATASET ACTIVATE BencareNoMissing.

    FILTER OFF.

    USE ALL. SELECT IF (rep17 * rep18 * rep19 * rep20 * inter1 * inter2 * inter3 * prac

    prac18 * prac19 *

    prac20 * val1 * val2 * val3 * val4 * val5 * val6 * loy1 * loy2 * loy3 * loy* loy6 * age *

    sex * educ * loc ~= 0).

    EXECUTE. DATASET ACTIVATE DataSet1.

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    30/32

    Filtering by Country

    DATASET ACTIVATE BencareNoMissing. DATASET COPY Germany.

    DATASET ACTIVATE Germany.

    FILTER OFF.

    USE ALL.

    SELECT IF (loc = 3).

    EXECUTE.

    DATASET ACTIVATE BencareNoMissing.

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    31/32

  • 8/13/2019 Srikanth Banerjee HA2 IMT-2013

    32/32

    Calculating Dummy variables DATASET ACTIVATE Germanycook.

    COMPUTE DREPEDU1=ZREP * Dedu1.

    EXECUTE.

    DATASET ACTIVATE Germanycook. COMPUTE DREPEDU2=ZREP * Dedu2.

    EXECUTE.

    DATASET ACTIVATE Germanycook.

    COMPUTE DREPEDU3=ZREP * Dedu3.

    EXECUTE.

    DATASET ACTIVATE Germanycook.

    COMPUTE DREPEDU4=ZREP * Dedu4.

    EXECUTE.

    DATASET ACTIVATE Germanycook.

    COMPUTE DSATEDU1=ZSAT * Dedu1.

    EXECUTE.