Marketing analyttics for daaaarr ch env roncenas
Wedel & Kannan (2016)
At the centre is the use of analytcs to support marketnn decisions. Key domains for analytcs
applicatons are:
Cusaocer relationsh i canagecena (CRM): methods that help acquisitonn retentonn and
satsfacton of customers to improee their lifetme ealue to the firm;
The carketing c x: methodsn modelsn and alnorithms that support the allocaton of
resources to enhance the efecteeness of marketnn efort;
Personal zation of the
marketnn mix to
indieidual consumers;
Pr vacyt anda secur ayt.
These domains lead to two pillars
of the successful deeelopment
and implementaton of marketnn
analytcs in firms:
The adopton of
structures and cultures
that foster data-drieen
decision makinn;
The educaton and
traininn of analytcs
professionals.
Daaa anda analyttics
Bin data is ofen characterized by the four (or fiee) “Vs”:
Volume: terabuytesn petabytes computnn;
Velocity (speed): snapshotsn hinh-frequency and streaminn data computnn;
Variety: numericn textn imanesn eideo analytcs;
Veracity: reliabilityn ealidity analytcs;
Value business.
Diferent ways of data collecton (bin data):
Observation: obseree customers (done mostly by Goonlen Amazon and Facebook);
Surveyts: haee become much easier to administer by online and mobile data collecton
(satsfactonn loyalty etc.);
F elda exier cenas: produce bin data and haee become powerful tools to explore the causal
efects of marketnn actons;
Lab exier cenas: nenerate smaller eolumes of datan but it can be used for online
administraton and collecton of audion eideon eye- and face-trackinn;
Do business decisions require more data or beter models? bias-eariance trade-of:
B as results from an incomplete representaton of the true data-neneratnn mechanism
(DGM) by a model because of simplifyinn assumptons. A less complex model ofen has a
hinher biasn but a model needs to simplify reality to proeide neneralizable insinhts;
Var ance results from random eariaton in the data due to samplinn and measurement error.
A larner eolume of data reduces the eariance. Complex models calibrated on smaller data
sets ofen oeer-fit the data (they capture random error rather than the DGM).
Wedel & Kannan (2016)
At the centre is the use of analytcs to support marketnn decisions. Key domains for analytcs
applicatons are:
Cusaocer relationsh i canagecena (CRM): methods that help acquisitonn retentonn and
satsfacton of customers to improee their lifetme ealue to the firm;
The carketing c x: methodsn modelsn and alnorithms that support the allocaton of
resources to enhance the efecteeness of marketnn efort;
Personal zation of the
marketnn mix to
indieidual consumers;
Pr vacyt anda secur ayt.
These domains lead to two pillars
of the successful deeelopment
and implementaton of marketnn
analytcs in firms:
The adopton of
structures and cultures
that foster data-drieen
decision makinn;
The educaton and
traininn of analytcs
professionals.
Daaa anda analyttics
Bin data is ofen characterized by the four (or fiee) “Vs”:
Volume: terabuytesn petabytes computnn;
Velocity (speed): snapshotsn hinh-frequency and streaminn data computnn;
Variety: numericn textn imanesn eideo analytcs;
Veracity: reliabilityn ealidity analytcs;
Value business.
Diferent ways of data collecton (bin data):
Observation: obseree customers (done mostly by Goonlen Amazon and Facebook);
Surveyts: haee become much easier to administer by online and mobile data collecton
(satsfactonn loyalty etc.);
F elda exier cenas: produce bin data and haee become powerful tools to explore the causal
efects of marketnn actons;
Lab exier cenas: nenerate smaller eolumes of datan but it can be used for online
administraton and collecton of audion eideon eye- and face-trackinn;
Do business decisions require more data or beter models? bias-eariance trade-of:
B as results from an incomplete representaton of the true data-neneratnn mechanism
(DGM) by a model because of simplifyinn assumptons. A less complex model ofen has a
hinher biasn but a model needs to simplify reality to proeide neneralizable insinhts;
Var ance results from random eariaton in the data due to samplinn and measurement error.
A larner eolume of data reduces the eariance. Complex models calibrated on smaller data
sets ofen oeer-fit the data (they capture random error rather than the DGM).
