Predictive Analytics as a Service- by MarketShare

MarketShare confidential and proprietary 1
Predictive Analytics as a Service
Prateem Mandal, Tech Lead Architect
October 28, 2015

Analytics Life Cycle
New Client
Onboarding
Scoring &
Attribution
Modeling
ETL
Scenario
Analysis &
Reporting
1
Stack
Generation

Analytics Workflow
Distributed
Cache
Tool
Application
Engines
Calculation
Engines
Execution
Systems
Elastic Load
Balancer
Client Onboarding
Model
Store
Config
Store
Attribution Funnel creationPost Processing
Service APIs
Metadata Store
Modeling Stack Transformation
StackETLConfigurationsETL
Model UAT
Attribution
Models
Evaluate
Automated Model Generation
1

Analytics Workflow
Distributed
Cache
Tool
Application
Engines
Calculation
Engines
Execution
Systems
Elastic Load
Balancer
Client Onboarding
Model
Store
Config
Store
Service APIs
Metadata Store
Model UAT
Attribution
Models
Evaluate
1

Analytics Workflow
Distributed
Cache
Tool
Application
Engines
Calculation
Engines
Execution
Systems
Elastic Load
Balancer
Client Onboarding
Model
Store
Config
Store
Service APIs
Metadata Store
Model UAT
Attribution
Models
Evaluate
1
Deployment Workflow
Modeling Workflow

Analytics Workflow
Distributed
Cache
Tool
Application
Engines
Calculation
Engines
Execution
Systems
Elastic Load
Balancer
Client Onboarding
Model
Store
Config
Store
Service APIs
Metadata Store
Model UAT
Attribution
Models
Evaluate
1
Modeling WorkflowModeling Workflow

New Client
Onboarding
Scoring &
Attribution
Modeling
ETL
Discovery
Scenario
Analysis &
Reporting
1
Stack
Generation

U S E R
U S E R S E Q U E N C E

L O O K B A C K
C A U S E E F F E C T U S E R

E F F E C T
L O O K B A C K
C A U S E
U S E R

E F F E C T U S E R
A G G R E G A T E D U S E R S E Q U E N C E
M E T R I C iM E T R I C 1 M E T R I C 2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

U S E RE F F E C TC A U S E
𝑮𝑯𝑻 𝒙𝒊 =
𝟏
𝟏 + 𝐦𝐢𝐧(𝒙𝒊)

E F F E C TC A U S E
𝟏

5
7
13
𝟏

5
7
13
E F F E C T U S E RM E T R I C iM E T R I C 1 M E T R I C 2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟏

5
7
13
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟏
𝟏 + 𝐦𝐢𝐧(𝟏𝟑, 𝟕, 𝟓)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟏

5
7
13
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟏
𝟔
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟏

5
7
13
U S E RE F F E C TM E T R I C iM E T R I C 1 M E T R I C 2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟎. 𝟏𝟔𝟔𝟔𝟔𝟔𝟕
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟏

New Client
Onboarding
Scoring &
Attribution
ETL
Discovery
Scenario
Analysis &
Reporting
1
Stack
Generation
Modeling

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
E F F E C TM E T R I C iM E T R I C 1 M E T R I C 2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟎. 𝟏𝟔𝟔𝟔𝟔𝟔𝟕
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟎. 𝟏𝟔𝟔𝟔𝟔𝟔𝟕
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟎. 𝟗𝟗𝟗𝟗𝟗𝟗𝟗 𝟎𝟏 𝟏𝟐𝟑𝟏. 𝟒𝟓𝟓𝟑

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13 7 5
𝟎. 𝟏𝟔𝟔𝟔𝟔𝟔𝟕
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟎. 𝟏𝟔𝟔𝟔𝟔𝟔𝟕
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
𝟎. 𝟗𝟗𝟗𝟗𝟗𝟗𝟗 𝟎𝟏 𝟏𝟐𝟑𝟏. 𝟒𝟓𝟓𝟑
E F F E C T U S E RM E T R I C 2
5713

New Client
Onboarding
Scoring &
Attribution
Modeling
ETL
Discovery
Scenario
Analysis &
Reporting
1
Stack
Generation

5713
C A U S E
13
7
5
E F F E C T U S E R A T T R A T T R
0.20
0.067
0.067
0.067

5713
C A U S E
13
7
5
E F F E C T U S E R A T T R A T T R
0.20
0.0
0.0
0.2

U S E R A T T RC A U S E E F F E C T

CID C1 . . . . Cm SID S1 . . . . SpEID E1 . . . . En
A T T R

A T T R
Reporting dimensions

A T T R
C1 . . . . Cm S1 . . . . SpE1 . . . . En
U S E RE F F E C TC A U S E A T T R
Reporting dimensions

Predictive Analytics as a Service- by MarketShare

stack_parameters.xml
stack_transforms.xml
stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<name>EVT_MARKETING_SEG_DIMS_SET</name>
<elements>
<elements><attribute>channel_type</attribute><value>“Display”</value></elements>
</elements>
<elements>
<elements><attribute>channel_type</attribute><value>“Paid Search”</value></elements>
</elements>
</set>
<select>
........
<param>METRICS_EVTSTACK</param>
........
</select>
........
<set name="EVT_MARKETING_SEG_DIMS">
<attr>channel_type</attr>
</set>
#!/bin/bash
.................
cmdstatus=`java -jar ${codegenerator_jar_dist_loc}/${code_generator_jar_name}
-s ${common_stages_loc}/$PHASE_NAME -d ${script_loc}/$PHASE_NAME -c
$input_schema_config_loc -g $code_generator_grammar_loc`
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
stack_gen_config.xml
2
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>HASDONEBEFORE</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
CREATE TABLE stackevtterms
AS
SELECT
.. AS stackmetric_hasdonebefore_event_display,
.. AS stackmetric_hasdonebefore_event_paidsearch,
......
FROM
......<set name="EVT_STACKMETRICS">
<val enable=“TRUE">HASDONEBEFORE</val>
<val enable=“FALSE">NUM</val>
</set>
#!/bin/bash
.................
cmdstatus=`hive –f $PHASE_NAME.hql`
.................
6
Input Config
Generated Config

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
</elements>
<elements>
</elements>
<elements>
<elements><attribute>channel_type</attribute><value>“Organic Search”</value></elements>
</elements>
</set>
<select>
........
........
</select>
........
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
AS
SELECT
.. AS stackmetric_hasdonebefore_event_display,
.. AS stackmetric_hasdonebefore_event_paidsearch,
.. AS stackmetric_hasdonebefore_event_organicsearch,
......
FROM
......
<set name="EVT_STACKMETRICS">
</set>
#!/bin/bash
.................
.................
6
Input Config
Generated Config
Data change

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements><attribute>evt_sub_channel</attribute><value>“Online Display”</value></elements> </elements>
<elements>
<elements><attribute>evt_sub_channel</attribute><value>“Direct Response Video”</value></elements>
</elements>
<elements>
<elements><attribute>evt_sub_channel</attribute><value>“Brand”</value></elements>
</elements>
<elements>
<elements><attribute>evt_sub_channel</attribute><value>“Unknown”</value></elements>
</elements>
</set>
<select>
........
........
</select>
........
<attr>evt_sub_channel</attr>
</set>
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
2
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
AS
SELECT
.. AS stackmetric_hasdonebefore_event_display_onlinedisplay,
.. AS stackmetric_hasdonebefore_event_display_directresponsevideo,
.. AS stackmetric_hasdonebefore_event_paidsearch_brand,
.. AS stackmetric_hasdonebefore_event_organicsearch_unknown,
......
FROM
......
</set>
#!/bin/bash
.................
.................
6
Input Config
Generated Config
Adding segmentation dimension
#!/bin/bash
.................
.................
5

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
<select>
........
........
</select>
........
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
AS
SELECT
.. AS stackmetric_num_event_display_onlinedisplay,
.. AS stackmetric_num_event_display_directresponsevideo,
.. AS stackmetric_num_event_paidsearch_brand,
.. AS stackmetric_num_event_organicsearch_unknown,
......
FROM
......
<val enable=“TRUE">NUM</val>
</set>
#!/bin/bash
.................
.................
6
Input Config
Generated Config
Selecting NUM metric

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
<select>
........
........
</select>
........
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
AS
SELECT
.. AS stackmetric_gsm90_event_display_onlinedisplay,
.. AS stackmetric_gsm90_event_display_directresponsevideo,
......
.. AS stackmetric_gsm365_event_paidsearch_brand,
.. AS stackmetric_gsm365_event_organicsearch_unknown,
......
FROM
......
<val enable=“TRUE">GSM</val>
</set>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
Input Config
Adding new metric
#!/bin/bash
.................
.................
6
Generated Config
<set name= “GSM_STEP">
<val>90</val>
<val>180</val>
<val>365</val>
</set>

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
AS
SELECT
.. AS stackmetric_gsm_event_display_onlinedisplay,
.. AS stackmetric_gsm_event_display_directresponsevideo,
.. AS stackmetric_gsm_event_paidsearch_brand,
.. AS stackmetric_gsm_event_organicsearch_unknown,
......
FROM
......
</set>
<val>3</val>
</set>
Input Config
Adding new metric
#!/bin/bash
.................
.................
6
Generated Config
stackevtterms_(<STCKGEN_RP><SDATE>i)
-- This query calculating all the NUM event metrics (stack variables or stack metrics).
-- stackevtterms_<RP_START_i> is created separately for each of the RP with RP_START in its name
DROP TABLE stackevtterms_(<STCKGEN_RP><SDATE>i);
CREATE TABLE stackevtterms_(<STCKGEN_RP><SDATE>i) AS
SELECT
<METRICS_EVTSTACK>
<OTHER_METRICS_EVTSTACK>
t2.rp_start AS rp_start,
t2.rp_end AS rp_end,
t1.userid AS userid,
t1.actuuid AS actuuid,
t1.<ACTIVITYGROUP> AS <ACTIVITYGROUP>,
t1.refdate_rp AS refdate_rp,
t1.response AS response
FROM
tmpUsersActivityDate_(<STCKGEN_RP><SDATE>i) t1 LEFT OUTER JOIN elbis_(<STCKGEN_RP><SDATE>i) t2
ON (t1.userid = t2.userid)
WHERE
(t1.sample_flag = 1)
GROUP BY
t2.rp_start, t2.rp_end, t1.userid, t1.actuuid, t1.<ACTIVITYGROUP>, t1.refdate_rp,
t1.response;
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
<select>
........
........
</select>
........

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
AS
SELECT
......
FROM
......
</set>
<val>3</val>
</set>
Input Config
Adding new metric
#!/bin/bash
.................
.................
6
Generated Config
SELECT
<METRICS_EVTSTACK>
FROM
WHERE
GROUP BY
t1.response;
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<select>
........
........
</select>
........

SUM(1 / (1 + DATEDIFF(
refdate_rp
CASE WHEN (
(channel_type IN ('Display') AND evt_sub_channel IN ('Online Display'))
AND
(evt_time > refdate_rp - (60 + 1) AND evt_time <= refdate_rp)
)
THEN evt_time
ELSE null END
)/365
)
) AS stackmetric_gsm365_event_display_onlinedisplay,
refdate_rp
CASE WHEN (
<and>
<xprod>
<in>
<xprodlist>EVT_MARKETING_SEG_DIMS_SET<element>attribute</element></xprodlist>
<xprodlist>EVT_MARKETING_SEG_DIMS_SET<element>value</element></xprodlist>
</in>
</xprod>
</and>
AND
)
THEN evt_time
ELSE null END
) / <val>GSM_STEP</val>
)
) AS stackmetric_gsm<val>GSM_STEP</val>_event_<xprod><xprodlist>EVT_MARKETING_SEG_DIMS_SET<element>value</element></xprodlist></xprod>,

stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
AS
SELECT
......
FROM
......
</set>
<val>3</val>
</set>
Input Config
Adding new metric
#!/bin/bash
.................
.................
6
Generated Config
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
SELECT
<METRICS_EVTSTACK>
FROM
WHERE
GROUP BY
t1.response;
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>
</defn>
</metric>
<select>
........
........
</select>
........

<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
<val>90</val>
<val>180</val>
<val>365</val>
</set>
refdate_rp
CASE WHEN (
<and>
<xprod>
<in>
</in>
</xprod>
</and>
AND
)
THEN evt_time
ELSE null END
)
12 different columns
will be generated from this
single expression
3 Steps
4 segments

<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
<val>90</val>
<val>180</val>
<val>365</val>
</set>
refdate_rp
CASE WHEN (
<and>
<xprod>
<in>
</in>
</xprod>
</and>
AND
)
THEN evt_time
ELSE null END
)
single expression
3 Steps
4 segments
......
......
......
......

New Client
Onboarding
Scoring &
Attribution
ETL
Discovery
Scenario
Analysis &
Reporting
1
Stack
Generation
Modeling

Modelling
Spec
• Initial set
of variables
• Business
rules
Model
Build N
Choice
Models
Model Evaluation and
Ranking
• Business Rules
• Goodness of Fit
• Statistical diagnosis
Finalized
Model
GA Iteration
• Top models cross –over to
generate offspring models
• Variable mutation
• Bayesian priors
• Coeff bounds
• Attribution bounds
Attribution
Stack
Automated Model Search

Max of modelScore Model Source
Iterations CrossOver Mutation New Max Score
1 0.814674133 0.814674133
2 0.840250827 0.874642259 -0.142370371 0.874642259
3 0.86051811 0.843359676 0.726339045 0.86051811
4 0.861854898 0.850294643 0.434102915 0.861854898
5 0.868137103 0.851901188 0.709897779 0.868137103
6 0.890517983 0.873505914 0.835870771 0.890517983
7 0.890538416 0.857600064 0.538250044 0.890538416
8 0.900386103 0.877312775 0.389444869 0.900386103
9 0.896563775 0.861402806 0.227565829 0.896563775
10 0.893281436 0.866702979 0.429202581 0.893281436
11 0.907564542 0.869666755 0.738092262 0.907564542
12 0.904108687 0.887093232 0.850036487 0.904108687
13 0.904159874 0.884415807 0.860378596 0.904159874
14 0.905516755 0.898474717 0.4425833 0.905516755
15 0.911024584 0.904605785 0.776345059 0.911024584
16 0.911007328 0.905694898 0.496509904 0.911007328
17 0.910692312 0.898231697 0.815998341 0.910692312
18 0.895302999 0.901081899 0.45632258 0.901081899
19 0.907471425 0.860367793 0.675483196 0.907471425
20 0.904312199 0.872822381 0.787050984 0.904312199
Max Score 0.911024584 0.905694898 0.860378596 0.911024584
GA Progression over Iterations
 Model with Max Score is generated from
{Cross-Over, Mutation} and not from
{New}

Model Score Improvement for Cross-over models over Iterations
-0.800
-0.600
-0.400
-0.200
0.000
0.200
0.400
0.600
0.800
1.000
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
53
55
57
59
61
63
65
67
69
71
73
75
77
79
81
83
85
87
89
91
93
95
97
99
101
103
105
107
109
111
113
115
117
119
121
123
125
ModelScore
Models over Iterations

Cross-over models over Iterations
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
ModelScore
Iterations
C - Average of modelScore C - StdDev of modelScore C - Max of modelScore

Distributed
Cache
Tool
Application
Engines
Calculation
Engines
Analytics Workflow – Performance – Large Client
Execution
Systems
Elastic Load
Balancer
Client Onboarding
Model
Store
Config
Store
Orchestrator
Metadata Store
Model UAT
Attribution
Models
Evaluate
2 h2 d
0-2
w
4 h .5 h 25 h
3 h
1.5
h
2 h
3 days
- 2
weeks
1 d
3

Result: significantly less effort for significantly larger deployments
2010 2012 2014 2015
Strategist Data Mgmt. Modeler Ops Support
Software
automation has
driven a decline in
service
requirements
Hours
4.5 B Data points
600MM Variables
170 MM Data points
90,000 Variables
20 MM Data points
7,000 Variables
2MM Data points
1,200 Variables
Total
hours
per
variable
Representative
deployment
Modeling scale
BANK CO. FINANCE CO. TELECOM CO. TRAVEL CO.
The size of models has grown significantly
Yet the deployment effort required
has decreased significantly
3

2010 2012 2014 2015
Revenue per resource
The revenue per deployment resource has 4X’d with scale gains
Representative
deployment
BANK CO. FINANCE CO. TELECOM CO. TRAVEL CO.
Revenue per service head for deployment increasing >4X over time
via greater automation
ACV Revenue/year ($ ‘000)
Deployment FTE Equivalent/year
$ ‘000
Revenue per
FTE
3

Questions &
Discussion

What’s Needed In A Decision Support System
MMM, Attribution, Targeting, Planning all need to share a common modeling platform
HIGHLY RESPONSIVE
Needs to provide responses in seconds not days or weeks
Must use robust models to calculate forward looking estimations and optimization
Needs the ability to integrate into the larger ecosystem
ANALYTICS RIGOR
INTEGRATIONS TO ECOSYSTEM
SINGLE PLATFORM
END-TO-END PLANNING
CMO, CFO,
Media planners, Digital Planners

Three pillars of technology

Decision Support Platform
Modeling Platform Results
2 34
Introduction :
Scenario Analysis
1

Large Scale Analytics requires configuration and metadata driven
modeling platform
2

TIME PRODUCT REGION # IMPRESSIONS SPEND ($$) GQV
TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
BI is well understood
FACT
Geo DIM
Product DIM
Time DIM
1
Data : Spend in $, Number of impressions bought and the corresponding Google Query Volume for 1st Quarter 2015

Reporting
What happened?
Analysis
Why did it
happen?
Monitoring
What’s
happening now?
Complexity
Business Value
BI can answer a lot of important questions about the past
Query, reporting & search tools
Dashboards, scorecards,
listening, real time reporting
OLAP and visualization tools
Business Intelligence
Complex event
processing; NLP;
Text mining
Time series analysis,
data mining,
clustering
1

TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
Impact of increasing spend by 10% to total spend in 3rd Quarter?
FACT
Geo DIM
Product DIM
Time DIM
1

TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
TID_3 PID_1 GID_1 0 4679.25
TID_3 PID_2 GID_1 0 40392
TID_3 PID_3 GID_1 0 37986.5
TID_3 PID_1 GID_2 0 15206.5
TID_3 PID_2 GID_2 0 260338
TID_3 PID_3 GID_2 0 249003.25
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
The Third Quarter spend could be replicated from the 1st Quarter
FACT
Geo DIM
Product DIM
Time DIM
1

TID_3 ALL ALL 419147300 607605.5
TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
TID_3 PID_1 GID_1 0 4679.25
TID_3 PID_2 GID_1 0 40392
TID_3 PID_3 GID_1 0 37986.5
TID_3 PID_1 GID_2 0 15206.5
TID_3 PID_2 GID_2 0 260338
TID_3 PID_3 GID_2 0 249003.25
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
Total Spend could be calculated
FACT
Geo DIM
Product DIM
Time DIM
To get the GQV value one has to build predictive models
1

Reporting
What happened?
Analysis
Why did it
happen?
Monitoring
What’s
happening now?
Complexity
Business Value
Query, reporting & search tools
Dashboards, scorecards,
listening, real time reporting
OLAP and visualization tools
Business Intelligence
Complex event processing;
NLP; Text mining
Time series analysis, data
mining, clustering
Prediction
What might
happen?
Marketers adopting next-gen analytics to drive decision making
Decision
What Should
I do now?
Predictive analytics
Decision support and
management
Time series analysis,
predictive modeling,
ensemble modeling,
machine learning
Constraint based
optimization; choice
modeling; decision trees
1

TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
To build models from data
FACT
Geo DIM
Product DIM
Time DIM
1

The Data Is Flattened Out
DENORMALIZED FLATTENED DATA
1/1/2015 COMP ALL 176843600 19885.75 8092237
1/1/2015 DIGI ALL 185465400 300730 3691062
1/1/2015 GAME ALL 56838300 286989.75 421021
1/1/2015 COMP GL 0 4679.25 8091825
1/1/2015 DIGI GL 0 40392 3684004
1/1/2015 GAME GL 0 37986.5 414270
1/1/2015 COMP MW 0 15206.5 412
1/1/2015 DIGI MW 0 260338 7058
1/1/2015 GAME MW 0 249003.25 6751
1

Data Is Grouped Into Sets Called Features
TIME PRODUCT REGION VARIABLE # IMPRESSIONS SPEND ($$) GQV
1/1/2015 COMP ALL TV_PRD_IM 176843600 19885.75 8092237
1/1/2015 DIGI ALL TV_PRD_IM 185465400 300730 3691062
1/1/2015 GAME ALL TV_PRD_IM 56838300 286989.75 421021
1/1/2015 COMP GL TV_LOCAL_PRD_SP 0 4679.25 8091825
1/1/2015 DIGI GL TV_LOCAL_PRD_SP 0 40392 3684004
1/1/2015 GAME GL TV_LOCAL_PRD_SP 0 37986.5 414270
1/1/2015 COMP MW TV_LOCAL_PRD_SP 0 15206.5 412
1/1/2015 DIGI MW TV_LOCAL_PRD_SP 0 260338 7058
1/1/2015 GAME MW TV_LOCAL_PRD_SP 0 249003.25 6751
DENORMALIZED FLATENNED DATA
Feature selection, also known as variable selection, attribute selection or variable
subset selection, is the process of selecting a subset of relevant features for use in
model construction.
1

Features Are Assembled Into An Equation
TIME PRODUCT REGION VARIABLE # IMPRESSIONS SPEND ($$) GQV
1/1/2015 COMP ALL TV_PRD_IM 176843600 19885.75 8092237
1/1/2015 DIGI ALL TV_PRD_IM 185465400 300730 3691062
1/1/2015 GAME ALL TV_PRD_IM 56838300 286989.75 421021
1/1/2015 COMP GL TV_LOCAL_PRD_SP 0 4679.25 8091825
1/1/2015 DIGI GL TV_LOCAL_PRD_SP 0 40392 3684004
1/1/2015 GAME GL TV_LOCAL_PRD_SP 0 37986.5 414270
1/1/2015 COMP MW TV_LOCAL_PRD_SP 0 15206.5 412
1/1/2015 DIGI MW TV_LOCAL_PRD_SP 0 260338 7058
1/1/2015 GAME MW TV_LOCAL_PRD_SP 0 249003.25 6751
FLATTENED DENORMALIZED DATA
Model Equation LOG(GQV_PD + 1) := TV_PRD_IM_LOGC*C(1) + TV_LOCAL_PRD_SP_LOGC*C(2)
1

Corresponding Coefficients Are Estimated
TIME PRODUCT REGION VARIABLE # IMPRESSIONS SPEND ($$) GQV COEFF_VALUE
1/1/2015 COMP ALL TV_PRD_IM 176843600 19885.75 8092237 0.045756241
1/1/2015 DIGI ALL TV_PRD_IM 185465400 300730 3691062 0.01985766
1/1/2015 GAME ALL TV_PRD_IM 56838300 286989.75 421021 0.007270448
1/1/2015 COMP GL TV_LOCAL_PRD_SP 0 4679.25 8091825 0.027113343
1/1/2015 DIGI GL TV_LOCAL_PRD_SP 0 40392 3684004 0.027113343
1/1/2015 GAME GL TV_LOCAL_PRD_SP 0 37986.5 414270 0.027113343
1/1/2015 COMP MW TV_LOCAL_PRD_SP 0 15206.5 412 0.027113343
1/1/2015 DIGI MW TV_LOCAL_PRD_SP 0 260338 7058 0.027113343
1/1/2015 GAME MW TV_LOCAL_PRD_SP 0 249003.25 6751 0.027113343
Model Equation
FLATTENED DENORMALIZED DATA
LOG(GQV_PD + 1) := TV_PRD_IM_LOGC*C(1) + TV_LOCAL_PRD_SP_LOGC*C(2)
1

Calculations Are Matrix Operations On The Model And The Coefficients
0.045756241
0.01985766
0.007270448
0.027113343
PRD=COMP & GEO = GL
PRD=DIGI & GEO = GL
PRD=GAME & GEO = GL
PRD=COMP & GEO = MW
PRD=DIGI & GEO = MW
PRD=GAME & GEO = MW
C1, PRD=COMP
C1, PRD=DIGI
C1, PRD=GAME
C2, PRD=ALL
GQV_PD
8091825
3684004
414270
412
7058
6751
TV_PRD_IM TV_PRD_IM TV_PRD_IM TV_LOCAL_PRD_SP
176843600 0 0 4679.25
0 185465400 0 40392
0 0 56838300 37986.5
176843600 0 0 15206.5
0 185465400 0 260338
0 0 56838300 249003.3
Model Input Coeff. Stack Outcome
The number of columns is the number of coefficients and
the number of rows is the number of distinct
combinations of dimensions
1

Impact of increasing spend by 10% to total spend in 3rd Quarter?
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
FACT
Geo DIM
Product DIM
Time DIM
TID_3 ALL ALL 419147300 607605.5
TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
TID_3 PID_1 GID_1 0 4679.25
TID_3 PID_2 GID_1 0 40392
TID_3 PID_3 GID_1 0 37986.5
TID_3 PID_1 GID_2 0 15206.5
TID_3 PID_2 GID_2 0 260338
TID_3 PID_3 GID_2 0 249003.25
1

TID_3 ALL ALL 419147300 668366.05
TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
TID_3 PID_1 GID_1 0 5147.18
TID_3 PID_2 GID_1 0 44431.2
TID_3 PID_3 GID_1 0 41785.15
TID_3 PID_1 GID_2 0 16727.15
TID_3 PID_2 GID_2 0 286371.8
TID_3 PID_3 GID_2 0 273903.6
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
FACT
Geo DIM
Product DIM
Time DIM
Distribute : Distribute The Spend At The Level Where The Model Is Defined
1

TIME PRODUCT REGION VARIABLE # IMPRESSIONS SPEND ($$)
7/1/2015 ALL ALL M_TV_N_BRD_SP 419147300 668366.05
7/1/2015 COMP ALL M_TV_N_PD_IM 176843600
7/1/2015 DIGI ALL M_TV_N_PD_IM 185465400
7/1/2015 GAME ALL M_TV_N_PD_IM 56838300
7/1/2015 COMP GL M_TV_L_P_SP 0 5147.18
7/1/2015 DIGI GL M_TV_L_P_SP 0 44431.2
7/1/2015 GAME GL M_TV_L_P_SP 0 41785.15
7/1/2015 COMP MW M_TV_L_P_SP 0 16727.15
7/1/2015 DIGI MW M_TV_L_P_SP 0 286371.8
7/1/2015 GAME MW M_TV_L_P_SP 0 273903.6
Calculate : Run The Model
TV_PRD_IM TV_PRD_IM TV_PRD_IM TV_LOCAL_PRD_SP
176843600 0 0 5147.18
0 185465400 0 44431.2
0 0 56838300 41785.15
176843600 0 0 16727.15
0 185465400 0 286371.8
0 0 56838300 273903.6
GQV_PD
88091838
3684114
414372.8
453.23
7763.86
7426.13 1

TID_3 ALL ALL 419147300 668366.05 92205968
TID_1 PID_1 ALL 176843600 19885.75 8092237
TID_1 PID_2 ALL 185465400 300730 3691062
TID_1 PID_3 ALL 56838300 286989.8 421021
TID_1 PID_1 GID_1 0 4679.25 8091825
TID_1 PID_2 GID_1 0 40392 3684004
TID_1 PID_3 GID_1 0 37986.5 414270
TID_1 PID_1 GID_2 0 15206.5 412
TID_1 PID_2 GID_2 0 260338 7058
TID_1 PID_3 GID_2 0 249003.25 6751
TID_3 PID_1 GID_1 0 5147.18 88091838
TID_3 PID_2 GID_1 0 44431.2 3684114
TID_3 PID_3 GID_1 0 41785.15 414372.8
TID_3 PID_1 GID_2 0 16727.15 453.23
TID_3 PID_2 GID_2 0 286371.8 7763.86
TID_3 PID_3 GID_2 0 273903.6 7426.13
ID CODE
PID_1 COMP
PID_2 DIGI
PID_3 GAME
ID CODE
GID_1 GL
GID_2 MW
ID CODE
TID_1 1/1/2015
TID_2 4/1/2015
TID_3 7/1/2015
FACT
Geo DIM
Product DIM
Time DIM
Aggregate : Increase the Total Marketing Spend by 10% Spend
1

Publisher
Marketing Driver
Tactic
Creative Concept
Geo
Time
Campaign
Year Quarter Month Week
Cube
Product
Non-Marketing Driver
Coupon redemption
Discounts
Macro economics
Pricing
Tabs
Weather
eCircular
Measures
Placement
Online Media Channel
Offline Media Channel
Social
Paid Social
Other
Affiliate
Display
Mobile
Video
Desktop
Other
Paid Search
Branded
Non Branded
Email
Audio
Magazine
Radio
TV
Leads
Tab
Product Listing
Services Directory
866
1590948
23910
336
15790
28
82
156
33
23
Assists OrdersRevenueClicks Events ImpressionsSpend
Last touches Click converting rate Converting click
131M cross sections, 10 KPIs
Dimension
1

INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION
ETL
Scenario
Analysis &
Reporting
CORE

STACK GENERATION MODELING SCORING & ATTRIBUTION
INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION

COEFFSTACK
COEFFSTACK
TRANSVAR
MODELING ATTRSTACK SCORING
EVENT
ATTRIBUTION
INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION

INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION
ETL
POST
PROCESSING
CORE

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
ACTUUID USERID REVENUE
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
MARKETING EVENTS IN LOOKBACK PERIOD
RECENT CONVERSIONS
DATA:
• ELBIS: Event table
• Contains marketing data.
• ALBIS : Activity table
• Conversions/Purchases

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
RECENT CONVERSIONS
DATA:
USER SEQUENCE:
• User history ending at conversion/purchase
• Marketing events shown to user
• Past conversions/purchases

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
RECENT CONVERSIONS
ATTRIBUTION:
• Share of each marketing event in a
sequence in deriving the conversion
DATA:
USER SEQUENCE:
• User history ending at conversion/purchase
• Marketing events shown to user
• Past conversions/purchases

INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION
ETL
POST
PROCESSING
CORE

INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION

COEFFSTACK
COEFFSTACK
TRANSVAR
EVENT
ATTRIBUTION
INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
ATTRIBUTION

MODEL.JSON
SYSTEMSETDEFINITION_GLOBAL
CLIENTSETDEFINITION.XML
<STAGE>_CONFIG.XML
CLIENTPARAMDEFINITION.XML
SYSTEMPARAMDEFINITION_GLOBAL
<SCHEMA>.XML
SYSTEMSETDEFINITION
<SCHEMA>.XML
<STAGE>_PARAMETERS.XML
COEFFSTACK
COEFFSTACK
TRANSVAR
EVENT
ATTRIBUTION
CONFIG/INPUTS/
CONFIG/SCHEMA_MAPPING/
COMMON/GLOBALCONFI
G/
CONFIG/TEMP/

COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
STACKMETRICS:
• Aggregated User behaviour information
• Aggregated marketing info per
sequence
• Aggregated past
conversion/purchase info
• Serves as base for variables for
Model building
• Calculated for each sequence

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
RECENT CONVERSIONS
RESPONSE
...........
1
1
1
1
...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE
STACKMETRICS:
sequence
• Aggregated past
Model building

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
NULL UUID4 NULL
NULL UUID5 NULL
........... ........... ...........
INCLUDING NON
CONVERTED USERD
RESPONSE
...........
1
1
1
1
...........
0
0
...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE
USER POOL:
• Collection of users used in model building
• Converted Users
• Non-converted Users
STACKMETRICS:
sequence
• Aggregated past
Model building

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
NULL UUID4 NULL
NULL UUID5 NULL
........... ........... ...........
USER-ACTIVITY POOL
NUM_IM NUM_CL
NUM_AFC
L
NUM_PS USERID
........... ........... ........... ........... ...........
2 0 0 0 UUID0
5 1 2 2 UUID1
1 0 0 3 UUID2
1 0 0 3 UUID2
........... ........... ........... ........... ...........
x x x x UUID4
x x x x UUID5
........... ........... ........... ........... ...........
STACKMETRICS
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE
RESPONSE
...........
1
1
1
1
...........
0
0
...........

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
NULL UUID4 NULL
NULL UUID5 NULL
........... ........... ...........
RECENT CONVERSIONS
MARKETING EVENTS IN LOOKBACK
(NUM_IM + NUM_CL
)
LOG( 1 +
NUM_AFCL)
LOG( 1 + NUM_PS
)
USERID
........... ........... ........... ...........
2 LOG(1+0) LOG(1+0) UUID0
........... ........... ........... ...........
x x x UUID4
x x x UUID4
........... ........... ........... ...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE
RESPONSE
...........
1
1
1
1
...........
0
0
...........
TRANSVAR
TRANSVAR:
• Universe of all model features
• Transforms on stackmetrics
• Input to GA, automated Model Selection
• The selected Model will have a subset these features

EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
NULL UUID4 NULL
NULL UUID5 NULL
........... ........... ...........
RECENT CONVERSIONS
(NUM_IM + NUM_CL
)
LOG( 1 +
NUM_AFCL)
LOG( 1 + NUM_PS
)
USERID
........... ........... ........... ...........
........... ........... ........... ...........
x x x UUID4
x x x UUID4
........... ........... ........... ...........
TRANSVAR
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
𝑷 𝑪𝒐𝒏𝒗 =
𝟏
𝟏 + 𝐞𝐱 𝐩 − 𝒊=𝟎
𝒏
𝜶𝒊 𝒙𝒊
M : a1*x1 + a2*x2
x1= (NUM_IM + NUM_CL )
x2= LOG( 1 + NUM_PS )
RESPONSE
RESPONS
E
...........
1
1
1
1
...........
0
0
...........

MarketShare confidential and proprietary
10
0
COEFFSTACK
COEFFSTACK
TRANSVAR
MODELING
INPUT CONFIGS
GENERATED
CONFIGS
MODEL BUILDING
MODEL.JSON

10
1
ATTRSTACK SCORING
EVENT
ATTRIBUTION
INPUT CONFIGS
GENERATED
CONFIGS
ATTRIBUTION
MODEL.JSON

10
2
EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
RECENT CONVERSIONS
NUM_IM NUM_CL
NUM_AFC
L
NUM_PS USERID
........... ........... ........... ........... ...........
2 0 0 0 UUID0
5 1 2 2 UUID1
1 0 0 3 UUID2
1 0 0 3 UUID2
........... ........... ........... ........... ...........
STACKMETRICS
RESPONS
E
...........
1
1
1
1
...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE

10
3
EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U0A1 UUID0 199
U1A1 UUID1 100
U2A1 UUID2 249
U2A2 UUID2 100
........... ........... ...........
RECENT CONVERSIONS
(NUM_IM + NUM_CL
)
LOG( 1 +
NUM_PS )
........... ...........
2 LOG(1+0)
6 LOG(1+2)
1 LOG(1+3)
1 LOG(1+3)
........... ...........
TRANSVAR
RESPONS
E
...........
1
1
1
1
...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
RESPONSE
FRAC_x1 FRAC_x2
........... ...........
0.35 0.65
0.46 0.54
0.57 0.43
0.39 0.61
........... ...........
MODELATTRIBUTION
𝟏
𝟏 + 𝐞𝐱 𝐩 − 𝒊=𝟎
𝒏
𝜶𝒊 𝒙𝒊
M : a1*x1 + a2*x2

10
4
EVENT
TYPE
EVTUUID USERID
........... ........... ...........
IM U0E1 UUID0
IM U0E2 UUID0
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
IM U2E1 UUID2
PS U2E2 UUID2
PS U2E3 UUID2
PS U2E4 UUID2
........... ........... ...........
........... ........... ...........
U1A1 UUID1 100
........... ........... ...........
TRANSVAR
RESPONS
E
...........
1
...........
COEFFSTAC
K
COEFFSTAC
K
TRANSVAR
EVENT
ATTRIBUTIO
N
GENERATED CONFIGS
INPUT CONFIGS
x FRAC_x FRACATTR
...... ......
...... ......
x1 0.46 .0766
x1 0.46 .0766
x1 0.46 .0766
...... ......
x2 0.54 0.27
x2 0.54 0.27
x1 0.46 .0766
x1 0.46 .0766
x1 0.46 .0766
...... ......
EVENT ATTRIBUTION
𝟏
𝟏 + 𝐞𝐱 𝐩 − 𝒊=𝟎
𝒏
𝜶𝒊 𝒙𝒊
EVENT
TYPE
EVTUUID USERID
........... ........... ...........
PS U1E1 UUID1
IM U1E2 UUID1
IM U1E3 UUID1
IM U1E4 UUID1
PS U1E5 UUID1
AFCL U1E6 UUID1
AFCL U1E7 UUID1
IM U1E8 UUID1
IM U1E9 UUID1
CL U1E10 UUID1
........... ........... ...........
M : a1*x1 + a2*x2

11
0
BaaS
Template Inputs
Orchestrator Scripts
stack_gen
RunGET
PUT


<description>ELBIS columns defining event segments</description>
</set>
<set name="STCKGEN_DAYSRANGE_EVTSTACK">
<description>List of time buckets used as T(max)_(min) for event stack
terms</description>
<elements>
<min>1</min>
<max>2</max>
</elements>
<elements>
<min>3</min>
<max>10</max>
</elements>
<elements>
<min>11</min>
<max>21</max>
</elements>
<elements>
<min>22</min>
<max>35</max>
</elements>
</set>
2

11
1
BaaS
Config
Generator
Template Inputs
stack_gen.sh
Phase Parameters
RunGET
PUT
<set>
<elements>
<elements><attribute>evt_sub_channel</attribute><value>“Online Display”</value></elements>
</elements>
<elements>
</elements>
<elements>
</elements>
<elements>
<elements><attribute>channel_type</attribute><value>“Paid Social”</value></elements>
</elements>
<elements>
</elements>
</set>
2

11
2
BaaS
Config
Generator
Template Inputs
stack_gen.sh
Phase Parameters
Phase Logic
RunGET
PUT
<param>NUM_METRICS_EVTSTACK</param>
<xprod order="1" typekey="hasdonebefore"
groupkey="event"
otherskey="0">
<expras>
<alias>
StackMetric_HasDoneBefore_Event
<xprod order="2">
_
<xprodlist order="2">
EVT_MARKETING_SEG_DIMS_SET
<element>
<element>value</element>
</element>
</xprodlist>
</xprodlist>
</xprod>
</alias>
.....................
.....................
.....................
.....................
</expras>
</xprod>
<xprod order="1" typekey="num"
groupkey="event" otherskey="0">
<expras>
<alias>
StackMetric_Num_Event
<xprod order="2">
_<xprodlist order="2">
EVT_MARKETING_SEG_DIMS_SET
<element>
<element>value</element>
</element>
</xprodlist>
</xprodlist>
</xprod>_T<xprodlist order="1">
STCKGEN_DAYSRANGE_EVTSTACK
<element>max</element>
</xprodlist>_<xprodlist order="1">
STCKGEN_DAYSRANGE_EVTSTACK
<element>min</element>
</xprodlist>
</alias>
.....................
.....................
.....................
.....................
</expras>
</xprod>
Phase Transform: https://git.marketshare.com:8443/projects/DIG/repos/msaction_backend/browse/common/BU3.0_core/stages/coeffstack/coeffstack_transforms.xml?at=refs%2Fheads%2Fautomatedmodeling
Phase Parameters: https://git.marketshare.com:8443/projects/DIG/repos/msaction_backend/browse/common/setup/config/globalconfig/systemparamdefinitions_global.xml?at=refs%2Fheads%2Fautomatedmodeling
Config Generator: https://git.marketshare.com:8443/projects/DIG/repos/msaction_backend/browse/common/BU3.0_core/util/configgenerator?at=refs%2Fheads%2Fautomatedmodeling
2

11
3
BaaS
AS
SELECT
...... AS stackmetric_hasdonebefore_event_display_onlinedisplay,
...... AS stackmetric_hasdonebefore_event_paidsearch_brand,
...... AS stackmetric_hasdonebefore_event_paidsocial_unknown,
...... AS stackmetric_hasdonebefore_event_organicsearch_unknown,
...... AS stackmetric_hasdonebefore_event_display_directresponsevideo,
...... AS stackmetric_num_event_display_onlinedisplay_t2_1,
...... AS stackmetric_num_event_paidsearch_brand_t2_1,
...... AS stackmetric_num_event_paidsocial_unknown_t2_1,
...... AS stackmetric_num_event_organicsearch_unknown_t2_1,
...... AS stackmetric_num_event_display_directresponsevideo_t2_1,
......
FROM
......
Config
Generator
Code
Generator
Template Inputs
stack_gen.sh
Phase Parameters
Phase Logic
stack.hql
RunGET
PUT
2

11
6
ETL
AttributionFunnel creation
ETL
customer
Data
(Raw Files)
Modeler
Attribution Models
Tool
Strategist
Elastic Load
Balancer
Application
Engines
Modelling
Spec
• Initial set
of variables
• Business
rules
Model
Build N
Choice
Models
Model Evaluation and
Ranking
• Business Rules
• Goodness of Fit
• Statistical diagnosis
Finalized
Model
GA Iteration
• Top models cross –over to
generate offspring models
• Variable mutation
• Bayesian priors
• Coeff bounds
• Attribution bounds
𝑃(𝐶𝑜𝑛𝑣) =
1
1 + exp(− 𝑖=0
𝑛
α𝑖 𝑥𝑖)
Attribution
Enhanced
Stack
Automated Model Search
2

11
9
stack.hql
Code
Generator
BaaS
Config
Generator
stack_gen.sh
<set>
<elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
<select>
........
........
</select>
........
</set>
#!/bin/bash
.................
.................
ORCHESTRATION
RUN
GET
PUT
ACTION
BACKEND
1
3
4
5
2
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
AS
SELECT
......
FROM
......
</set>
<val>3</val>
</set>
<metric>
<name>NUM</name>
<defn>.....</defn>
</metric>
<metric>
<defn>.....</defn>
</metric>
<metric>
<name>GSM</name>
<defn>.....</defn>
</metric>
Input Config
Adding new metric
#!/bin/bash
.................
.................
6
Generated Config

12
0
<transform>
T14: stackevtterms_(<STCKGEN_RP><SDATE>i)
1. DROP TABLE stackevtterms_(<STCKGEN_RP><SDATE>i);
2. CREATE TABLE stackevtterms_(<STCKGEN_RP><SDATE>i) AS
3. SELECT
4. <NUM_METRICS_EVTSTACK>
5. <OTHER_METRICS_EVTSTACK>
6. t2.rp_start AS rp_start,
7. t2.rp_end AS rp_end,
8. t1.userid AS userid,
9. t1.actuuid AS actuuid,
10. t1.<ACTIVITYGROUP> AS <ACTIVITYGROUP>,
11. t1.refdate_rp AS refdate_rp,
12. t1.response AS response
13. FROM tmpUsersActivityDate_(<STCKGEN_RP><SDATE>i) t1 LEFT OUTER JOIN elbis_(<STCKGEN_RP><SDATE>i) t2
14. ON (t1.userid = t2.userid)
15. WHERE (
16. (t1.sample_flag = 1))
17. GROUP BY t2.rp_start, t2.rp_end, t1.userid, t1.actuuid, t1. (<STCKGEN_RP><SDATE>i), t1.refdate_rp, t1.response
18. ;

12
1
<param>
COALESCE(SUM((case when (
t2.evt_event_type IN ('IM') AND
(TO_DATE(t2.evt_time) >= (DATE_SUB(TO_DATE(t1.refdate_rp), (2 - 1))) AND
(TO_DATE(t2.evt_time) <= (DATE_SUB(TO_DATE(t1.refdate_rp), (1 - 1))))
) THEN 1.0 else 0.0 END)), 0)
AS num_im_t2_1,
t2.evt_event_type IN (<xprodlist>STCKGEN_EVTTYPE</xprodlist>) AND
(TO_DATE(t2.evt_time) >= (DATE_SUB(TO_DATE(t1.refdate_rp),
(<xprodlist>STCKGEN_DAYSRANGE_ACTSTACK<element>max</element></xprodlist> - 1))) AND
(TO_DATE(t2.evt_time) <= (DATE_SUB(TO_DATE(t1.refdate_rp),
((<xprodlist>STCKGEN_DAYSRANGE_ACTSTACK<element>min</element></xprodlist> - 1))))
) THEN 1.0 else 0.0 END)), 0)
AS
num_<xprodlist>STCKGEN_EVTTYPE</xprodlist>_t<xprodlist>STCKGEN_DAYSRANGE_ACTSTACK<element>max</element></xprodlist>_<xprodlist>STCKGEN_DAYS
RANGE_ACTSTACK<element>min</element></xprodlist>,

12
2
t2.evt_event_type IN (<xprodlist>STCKGEN_EVTTYPE</xprodlist>) AND
(TO_DATE(t2.evt_time) >= (DATE_SUB(TO_DATE(t1.refdate_rp),
(<xprodlist>STCKGEN_DAYSRANGE_EVTSTACK<element>max</element></xprodlist> - 1))) AND
(TO_DATE(t2.evt_time) <= (DATE_SUB(TO_DATE(t1.refdate_rp),
((<xprodlist>STCKGEN_DAYSRANGE_EVTSTACK<element>min</element></xprodlist> - 1))))
) THEN 1.0 else 0.0 END)), 0)
AS
num_<xprodlist>STCKGEN_EVTTYPE</xprodlist>_t<xprodlist>STCKGEN_DAYSRANGE_EVTSTACK<element>max</element>
</xprodlist>_<xprodlist>STCKGEN_DAYSRANGE_EVTSTACK<element>min</element></xprodlist>,
Parameterization*

<set>
<name>STCKGEN_EVTTYPE</name>
<val>'IM'</val>
<val>'CL'</val>
<val>'PS'</val>
<val>'OS'</val>
<val>'EO'</val>
<val>'EC'</val>
<val>'ED'</val>
<val>'AFCL'</val>
<val>'SOCL'</val>
<val>'DR'</val>
</set>

<set>
<name>STCKGEN_DAYSRANGE_EVTSTACK</name>
<elements>
<min><val>1</val></min><max><val>2</val></max>
</elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
<elements>
</elements>
</set>
single expression
10 event types
5 time buckets

Predictive Analytics as a Service- by MarketShare

More Related Content

More from Sri Ambati (20)

Recently uploaded (20)

Predictive Analytics as a Service- by MarketShare

Editor's Notes