SlideShare a Scribd company logo
1 of 25
Data Science
in Business
Example from Debt Collection
Industry
1
Sergey Sviridenko Vladislav Sikach
VFG
Head of Process Development
Senior Process Manager
Process Manager
Quantitative Analyst
Junior Quant
2
3
Data
Science
4
AREAS
Valuation Segmentation Processing
5
Portfolio
Valuation
6
1.3M debtors
120+ clients
?
7
8
vs
9
Debtor
Segmentation
10
11
1. Value based
Customers grouping according
to their value.
2. Loyalty based
Used to separate loyal
customers (obedient payers)
from migrators/switchers.
Who will
bring
only
losses?
Who will
bring more
profit?
Debt becomes miniscule?
Any contact occurs?
Contact with Debtor occurs?
Agreement to pay given?
Promise given?
Started to pay?
Active Cases
Communication Channel Exists
Successful Action
Contact with Debtor
Agreement to Pay
Promise to Pay
Payment
Paid-off
Any address or phone added?
1. Value Chain
2. Loyalty
Convertible
(haven’t paid before)
Ongoing
(paid in last 60 days)
Revertible
(paid >60 days ago)
Optimal Segmentation
Initial + Operational Features
12
Optimal Segmentation
Initial + Operational Features
Known Best & Worst
Cases
3. Propensity based
Binning of customers in groups
according to their propensity scores.
Call list sorting by scoring, Known
Best/Worst Segmentation
5. Behavioral based
Used to develop customized product
offering strategies. It is about willingness
and ability to pay off debt (promise
fulfillment), attitude toward collectors
4. Socio-demographic
Used for promoting specific life-stage-
based products. Gathering people into
groups by Country, Region, Age, Sex for
special offers
13
CRISP-DM
14
Additional data structures and
procedures for cases (debtors,
phones, addresses and other
entities) attributes & statuses
changes
Part of segmentation and
flows are parameterized in
decision-making system
integrated in Qualco CRM
Prepare analyses/tests
results for presentation to
Management Board
OR
IN IN
Analyses,
Models,
Significance,
Hypotheses
testing
15
Optimal
Processing
16
1.3M debtors
2M actions daily
Multiple tools
How to process different segments
Optimally
17
Detailed List of
Areas to Improve
18
What is the best time to call depending on segment and phone type
Intraday recalls after successful (refusal to talk/pay, not identified, agreed but not promised etc.)
Optimal Calling Frequency after unsuccessful results
Optimal predictive dialer configuration (Technical contacts classifier, SCR, LT)
How to choose the best SMS and Letters templates and sequences
How to prioritize addresses for letter sending (~7 different addresses per debtor)
Targeted Skip-tracing (contact data search sources and optimal sequences)
When to deactivate the phone (counters for Gateway Disposition Codes & successful results)
Whom & how much to forgive, what product to offer (simple writeoffs, restructuring, hybrid schemes)
What segments should be visible for Legal & Field Collection Agents
Experimental
Framework
Champion/
Challenger
Testing
19
(Pi – P0)
additional marginal effect.
Pi and P0 are calculated
historically via regressions
Costi
Cost of ith instrument(s)
usage
(Cashn+1 – Cashn)
additional historical cash
increase gained from
transition
P0 – default probability, w/o
any instrument
Value Chain Stagen
Yields Cash per casen
Value Chain Stagen+1
Yields Cash per casen+1
Pi probability of transition
if ith-instrument is applied
vs.
Expected Profiti = (Pi – P0) ∙ (Cashn+1 – Cashn) – Costi
Maximize the profit at each value chain
step applying the most suitable
instrument
20
21
Profit Instr 1 Instr 2 Instr 3
Case 1 -0.5$ 0.5$ 0.1$
Case 2 0.1$ -0.5$ 0.5$
Case 3 0.5$ 0.1$ -0.5$
Case Instrument
1 2
2 3
3 1
ExpectedProfitpercase
Cases in
queue
Case #1 Case #2 … Casen #k … Case #n
$0
$3
$5
$9
-$2
Case queue for ith instrument
If there is a lack of capacity,
quick win is to use substitute and less profitable instruments
If this situation is repeated day after day,
then you need to increase capacity of particular instrument
If it is unprofitable to process them by any av
develop new/substitute instruments or transfe
Step 1
Step 2
Step 3
Calculate expected profit
for each case and instrument applicable
Assign the instrument,
which yields the maximum profit
Build case queues for each instrument
with expected profit as priority
For given case find the most suitable
instrument & apply it according to priorities
Additional data structures
and procedures for different
attributes changes (phones
and cases deactivation /
reactivation rules, dialing
logic, address priority
calculation etc.
Some workflows and rules
can be parameterized in
decision-making system
integrated in Qualco CRM
Prepare analyses/tests
results for presentation
to Management Board
IN IN
Used for analyses,
models,
check for
significance &
hypotheses testing
OR
New business processes,
algorithms, workflows, rules
and solution models are
designed as diagrams for
better process understanding
for business and further
development 22
Lessons
Learned
23
● Be curious & go into the field
● Speak domain language
● Deliver fast and frequent
● Show them money, not your Gini
● Do simple models
● Clarify assumptions
● Try something new, but
● Test everything
● Learn different languages
● Know system capabilities
● Know system constraints
● Borrow PM practices
24
25

More Related Content

What's hot

IGRP_EliminatingCost_Whitepaper (1)
IGRP_EliminatingCost_Whitepaper (1)IGRP_EliminatingCost_Whitepaper (1)
IGRP_EliminatingCost_Whitepaper (1)Karen Morgan
 
Financial Crime Projects
Financial Crime ProjectsFinancial Crime Projects
Financial Crime ProjectsDavid Allsop
 
Intelligence inthechannel
Intelligence inthechannelIntelligence inthechannel
Intelligence inthechannelManish Aurora
 
John_Lazcano_2016 (BHC)
John_Lazcano_2016 (BHC)John_Lazcano_2016 (BHC)
John_Lazcano_2016 (BHC)John Lazcano
 
2008 Investment Symposium Zhang 3 24 08
2008 Investment Symposium Zhang 3 24 082008 Investment Symposium Zhang 3 24 08
2008 Investment Symposium Zhang 3 24 08Frank Zhang
 
FS_StressTestingCapitalPlanning_BR_1213 v1
FS_StressTestingCapitalPlanning_BR_1213 v1FS_StressTestingCapitalPlanning_BR_1213 v1
FS_StressTestingCapitalPlanning_BR_1213 v1Sudip Chatterjee
 
Solving the FRTB Challenge: Why You Should Consider an Aggregation Solution
Solving the FRTB Challenge: Why You Should Consider an Aggregation SolutionSolving the FRTB Challenge: Why You Should Consider an Aggregation Solution
Solving the FRTB Challenge: Why You Should Consider an Aggregation SolutionFIS
 
John_Lazcano_2014-2PwC-1
John_Lazcano_2014-2PwC-1John_Lazcano_2014-2PwC-1
John_Lazcano_2014-2PwC-1John Lazcano
 
John_Lazcano_2014PwC
John_Lazcano_2014PwCJohn_Lazcano_2014PwC
John_Lazcano_2014PwCJohn Lazcano
 
Xuber Analytics
Xuber AnalyticsXuber Analytics
Xuber AnalyticsXuber
 
2. op risk and aml
2. op risk and aml2. op risk and aml
2. op risk and amlcrmbasel
 
Internal Auditor - Profile
Internal Auditor - ProfileInternal Auditor - Profile
Internal Auditor - ProfileKhushboo Jain
 
IFRS 9 IT Architecture : Go Strategic or Tactical ?
IFRS 9 IT Architecture  : Go Strategic or Tactical ?IFRS 9 IT Architecture  : Go Strategic or Tactical ?
IFRS 9 IT Architecture : Go Strategic or Tactical ?Sandip Mukherjee CFA, FRM
 
Early warning system_ white paper
Early warning system_ white paperEarly warning system_ white paper
Early warning system_ white paperFederica Tasselli
 

What's hot (20)

IGRP_EliminatingCost_Whitepaper (1)
IGRP_EliminatingCost_Whitepaper (1)IGRP_EliminatingCost_Whitepaper (1)
IGRP_EliminatingCost_Whitepaper (1)
 
Financial Crime Projects
Financial Crime ProjectsFinancial Crime Projects
Financial Crime Projects
 
rmarshall_pm
rmarshall_pmrmarshall_pm
rmarshall_pm
 
Feasible
FeasibleFeasible
Feasible
 
Intelligence inthechannel
Intelligence inthechannelIntelligence inthechannel
Intelligence inthechannel
 
My Resume
My ResumeMy Resume
My Resume
 
John_Lazcano_2016 (BHC)
John_Lazcano_2016 (BHC)John_Lazcano_2016 (BHC)
John_Lazcano_2016 (BHC)
 
Pramoth_R 2005-2015 (1)
Pramoth_R 2005-2015 (1)Pramoth_R 2005-2015 (1)
Pramoth_R 2005-2015 (1)
 
2008 Investment Symposium Zhang 3 24 08
2008 Investment Symposium Zhang 3 24 082008 Investment Symposium Zhang 3 24 08
2008 Investment Symposium Zhang 3 24 08
 
FS_StressTestingCapitalPlanning_BR_1213 v1
FS_StressTestingCapitalPlanning_BR_1213 v1FS_StressTestingCapitalPlanning_BR_1213 v1
FS_StressTestingCapitalPlanning_BR_1213 v1
 
Solving the FRTB Challenge: Why You Should Consider an Aggregation Solution
Solving the FRTB Challenge: Why You Should Consider an Aggregation SolutionSolving the FRTB Challenge: Why You Should Consider an Aggregation Solution
Solving the FRTB Challenge: Why You Should Consider an Aggregation Solution
 
John_Lazcano_2014-2PwC-1
John_Lazcano_2014-2PwC-1John_Lazcano_2014-2PwC-1
John_Lazcano_2014-2PwC-1
 
John_Lazcano_2014PwC
John_Lazcano_2014PwCJohn_Lazcano_2014PwC
John_Lazcano_2014PwC
 
Xuber Analytics
Xuber AnalyticsXuber Analytics
Xuber Analytics
 
L Escudero Resume62016
L Escudero Resume62016L Escudero Resume62016
L Escudero Resume62016
 
2. op risk and aml
2. op risk and aml2. op risk and aml
2. op risk and aml
 
Internal Auditor - Profile
Internal Auditor - ProfileInternal Auditor - Profile
Internal Auditor - Profile
 
IFRS 9 IT Architecture : Go Strategic or Tactical ?
IFRS 9 IT Architecture  : Go Strategic or Tactical ?IFRS 9 IT Architecture  : Go Strategic or Tactical ?
IFRS 9 IT Architecture : Go Strategic or Tactical ?
 
Early warning system_ white paper
Early warning system_ white paperEarly warning system_ white paper
Early warning system_ white paper
 
FRTB
FRTBFRTB
FRTB
 

Similar to Data science role in business

201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service
201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service
201306 Tech Decisions Webinar: Modernizing Claims for Better Customer ServiceSteven Callahan
 
Integrated Receivables: 5 Critical Factors For Adoption
Integrated Receivables: 5 Critical Factors For AdoptionIntegrated Receivables: 5 Critical Factors For Adoption
Integrated Receivables: 5 Critical Factors For Adoption3 Point Alliance
 
How Finance is driving growth in the Digital Age via OpenText
How Finance is driving growth in the Digital Age via OpenTextHow Finance is driving growth in the Digital Age via OpenText
How Finance is driving growth in the Digital Age via OpenTextOpenText
 
CTRM Value Survey and Analysis
 CTRM Value Survey and Analysis CTRM Value Survey and Analysis
CTRM Value Survey and AnalysisCTRM Center
 
Asset finance systems projects guide 101
Asset finance systems projects guide 101Asset finance systems projects guide 101
Asset finance systems projects guide 101David Pedreno
 
Ten Essentials of Treasury Technology TMANE 2009
Ten Essentials of Treasury Technology TMANE 2009Ten Essentials of Treasury Technology TMANE 2009
Ten Essentials of Treasury Technology TMANE 2009rthompson89
 
This is my test slideshare
This is my test slideshareThis is my test slideshare
This is my test slidesharepapdev
 
SPPM Clinical 7 Best Practices In Forecasting & Planning
SPPM Clinical   7 Best Practices In Forecasting & PlanningSPPM Clinical   7 Best Practices In Forecasting & Planning
SPPM Clinical 7 Best Practices In Forecasting & Planningguest1fe658d
 
CollectionOptimization
CollectionOptimizationCollectionOptimization
CollectionOptimizationMike Nguyen
 
Ppt Template
Ppt TemplatePpt Template
Ppt Templatepapdev
 
A G S006 Little 091807
A G S006  Little 091807A G S006  Little 091807
A G S006 Little 091807Dreamforce07
 
Nitin Singla_28072016
Nitin Singla_28072016Nitin Singla_28072016
Nitin Singla_28072016ntnsng
 
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...Perficient, Inc.
 
Account Management Handbook.pptx
Account Management Handbook.pptxAccount Management Handbook.pptx
Account Management Handbook.pptxVienVo15
 

Similar to Data science role in business (20)

Resume.Ram
Resume.RamResume.Ram
Resume.Ram
 
201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service
201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service
201306 Tech Decisions Webinar: Modernizing Claims for Better Customer Service
 
Integrated Receivables: 5 Critical Factors For Adoption
Integrated Receivables: 5 Critical Factors For AdoptionIntegrated Receivables: 5 Critical Factors For Adoption
Integrated Receivables: 5 Critical Factors For Adoption
 
How Finance is driving growth in the Digital Age via OpenText
How Finance is driving growth in the Digital Age via OpenTextHow Finance is driving growth in the Digital Age via OpenText
How Finance is driving growth in the Digital Age via OpenText
 
T S Kannan
T S KannanT S Kannan
T S Kannan
 
Claims
ClaimsClaims
Claims
 
CTRM Value Survey and Analysis
 CTRM Value Survey and Analysis CTRM Value Survey and Analysis
CTRM Value Survey and Analysis
 
Asset finance systems projects guide 101
Asset finance systems projects guide 101Asset finance systems projects guide 101
Asset finance systems projects guide 101
 
CRMS_Project-JF-edits
CRMS_Project-JF-editsCRMS_Project-JF-edits
CRMS_Project-JF-edits
 
Ten Essentials of Treasury Technology TMANE 2009
Ten Essentials of Treasury Technology TMANE 2009Ten Essentials of Treasury Technology TMANE 2009
Ten Essentials of Treasury Technology TMANE 2009
 
This is my test slideshare
This is my test slideshareThis is my test slideshare
This is my test slideshare
 
SPPM Clinical 7 Best Practices In Forecasting & Planning
SPPM Clinical   7 Best Practices In Forecasting & PlanningSPPM Clinical   7 Best Practices In Forecasting & Planning
SPPM Clinical 7 Best Practices In Forecasting & Planning
 
CollectionOptimization
CollectionOptimizationCollectionOptimization
CollectionOptimization
 
Ppt Template
Ppt TemplatePpt Template
Ppt Template
 
A G S006 Little 091807
A G S006  Little 091807A G S006  Little 091807
A G S006 Little 091807
 
Day 1 (Lecture 2): Business Analytics
Day 1 (Lecture 2): Business AnalyticsDay 1 (Lecture 2): Business Analytics
Day 1 (Lecture 2): Business Analytics
 
Nitin Singla_28072016
Nitin Singla_28072016Nitin Singla_28072016
Nitin Singla_28072016
 
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...
Increase Financial Firms' Sales Performance & Compliance with Compensation Ma...
 
Yellow belt
Yellow beltYellow belt
Yellow belt
 
Account Management Handbook.pptx
Account Management Handbook.pptxAccount Management Handbook.pptx
Account Management Handbook.pptx
 

Recently uploaded

原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档
原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档
原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档208367051
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubaihf8803863
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一F sss
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfJohn Sterrett
 
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样vhwb25kk
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPramod Kumar Srivastava
 
9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home ServiceSapana Sha
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024thyngster
 
Top 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In QueensTop 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In Queensdataanalyticsqueen03
 
Call Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceCall Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceSapana Sha
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationshipsccctableauusergroup
 
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理科罗拉多大学波尔得分校毕业证学位证成绩单-可办理
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理e4aez8ss
 
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一F La
 
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...dajasot375
 
B2 Creative Industry Response Evaluation.docx
B2 Creative Industry Response Evaluation.docxB2 Creative Industry Response Evaluation.docx
B2 Creative Industry Response Evaluation.docxStephen266013
 
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...Boston Institute of Analytics
 
Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxEmmanuel Dauda
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFAAndrei Kaleshka
 

Recently uploaded (20)

原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档
原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档
原版1:1定制南十字星大学毕业证(SCU毕业证)#文凭成绩单#真实留信学历认证永久存档
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdf
 
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
 
9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
 
Top 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In QueensTop 5 Best Data Analytics Courses In Queens
Top 5 Best Data Analytics Courses In Queens
 
Call Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceCall Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts Service
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships
 
E-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptxE-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptx
 
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理科罗拉多大学波尔得分校毕业证学位证成绩单-可办理
科罗拉多大学波尔得分校毕业证学位证成绩单-可办理
 
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一
办理(UWIC毕业证书)英国卡迪夫城市大学毕业证成绩单原版一比一
 
Call Girls in Saket 99530🔝 56974 Escort Service
Call Girls in Saket 99530🔝 56974 Escort ServiceCall Girls in Saket 99530🔝 56974 Escort Service
Call Girls in Saket 99530🔝 56974 Escort Service
 
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
Indian Call Girls in Abu Dhabi O5286O24O8 Call Girls in Abu Dhabi By Independ...
 
B2 Creative Industry Response Evaluation.docx
B2 Creative Industry Response Evaluation.docxB2 Creative Industry Response Evaluation.docx
B2 Creative Industry Response Evaluation.docx
 
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
NLP Data Science Project Presentation:Predicting Heart Disease with NLP Data ...
 
Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptx
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFA
 

Data science role in business

  • 1. Data Science in Business Example from Debt Collection Industry 1
  • 2. Sergey Sviridenko Vladislav Sikach VFG Head of Process Development Senior Process Manager Process Manager Quantitative Analyst Junior Quant 2
  • 3. 3
  • 8. 8
  • 11. 11 1. Value based Customers grouping according to their value. 2. Loyalty based Used to separate loyal customers (obedient payers) from migrators/switchers. Who will bring only losses? Who will bring more profit? Debt becomes miniscule? Any contact occurs? Contact with Debtor occurs? Agreement to pay given? Promise given? Started to pay? Active Cases Communication Channel Exists Successful Action Contact with Debtor Agreement to Pay Promise to Pay Payment Paid-off Any address or phone added? 1. Value Chain 2. Loyalty Convertible (haven’t paid before) Ongoing (paid in last 60 days) Revertible (paid >60 days ago) Optimal Segmentation Initial + Operational Features
  • 12. 12 Optimal Segmentation Initial + Operational Features Known Best & Worst Cases 3. Propensity based Binning of customers in groups according to their propensity scores. Call list sorting by scoring, Known Best/Worst Segmentation 5. Behavioral based Used to develop customized product offering strategies. It is about willingness and ability to pay off debt (promise fulfillment), attitude toward collectors 4. Socio-demographic Used for promoting specific life-stage- based products. Gathering people into groups by Country, Region, Age, Sex for special offers
  • 14. 14
  • 15. Additional data structures and procedures for cases (debtors, phones, addresses and other entities) attributes & statuses changes Part of segmentation and flows are parameterized in decision-making system integrated in Qualco CRM Prepare analyses/tests results for presentation to Management Board OR IN IN Analyses, Models, Significance, Hypotheses testing 15
  • 17. 1.3M debtors 2M actions daily Multiple tools How to process different segments Optimally 17
  • 18. Detailed List of Areas to Improve 18 What is the best time to call depending on segment and phone type Intraday recalls after successful (refusal to talk/pay, not identified, agreed but not promised etc.) Optimal Calling Frequency after unsuccessful results Optimal predictive dialer configuration (Technical contacts classifier, SCR, LT) How to choose the best SMS and Letters templates and sequences How to prioritize addresses for letter sending (~7 different addresses per debtor) Targeted Skip-tracing (contact data search sources and optimal sequences) When to deactivate the phone (counters for Gateway Disposition Codes & successful results) Whom & how much to forgive, what product to offer (simple writeoffs, restructuring, hybrid schemes) What segments should be visible for Legal & Field Collection Agents
  • 20. (Pi – P0) additional marginal effect. Pi and P0 are calculated historically via regressions Costi Cost of ith instrument(s) usage (Cashn+1 – Cashn) additional historical cash increase gained from transition P0 – default probability, w/o any instrument Value Chain Stagen Yields Cash per casen Value Chain Stagen+1 Yields Cash per casen+1 Pi probability of transition if ith-instrument is applied vs. Expected Profiti = (Pi – P0) ∙ (Cashn+1 – Cashn) – Costi Maximize the profit at each value chain step applying the most suitable instrument 20
  • 21. 21 Profit Instr 1 Instr 2 Instr 3 Case 1 -0.5$ 0.5$ 0.1$ Case 2 0.1$ -0.5$ 0.5$ Case 3 0.5$ 0.1$ -0.5$ Case Instrument 1 2 2 3 3 1 ExpectedProfitpercase Cases in queue Case #1 Case #2 … Casen #k … Case #n $0 $3 $5 $9 -$2 Case queue for ith instrument If there is a lack of capacity, quick win is to use substitute and less profitable instruments If this situation is repeated day after day, then you need to increase capacity of particular instrument If it is unprofitable to process them by any av develop new/substitute instruments or transfe Step 1 Step 2 Step 3 Calculate expected profit for each case and instrument applicable Assign the instrument, which yields the maximum profit Build case queues for each instrument with expected profit as priority For given case find the most suitable instrument & apply it according to priorities
  • 22. Additional data structures and procedures for different attributes changes (phones and cases deactivation / reactivation rules, dialing logic, address priority calculation etc. Some workflows and rules can be parameterized in decision-making system integrated in Qualco CRM Prepare analyses/tests results for presentation to Management Board IN IN Used for analyses, models, check for significance & hypotheses testing OR New business processes, algorithms, workflows, rules and solution models are designed as diagrams for better process understanding for business and further development 22
  • 24. ● Be curious & go into the field ● Speak domain language ● Deliver fast and frequent ● Show them money, not your Gini ● Do simple models ● Clarify assumptions ● Try something new, but ● Test everything ● Learn different languages ● Know system capabilities ● Know system constraints ● Borrow PM practices 24
  • 25. 25

Editor's Notes

  1. Who we are
  2. Кто знает про debt collection? Что это такое? Миссия - фин образование населения Баланс экосистемы (банки-люди-страна) Деятельность по взысканию долгов содействует долгосрочному развитию национальной экономики посредством возмещения и возврата в хозяйственный оборот значительных денежных сумм. Коллекторские агентства помогают также повысить финансовую дисциплину населения и экономических агентов.
  3. На что стоит обращать внимание разным ролям Вопрос - у кого какой бекграунд? Примеры: PM (Product owners, founders, business guys), Software Engineers, BI, Analysts,
  4. 3 основные задачи Оценка Кого и как эффективно и прибыльно обрабатывтаь Универсальность. Можно применять в любых других бизнес сферах, так как принципы одни и те же. (по предложению на описание задачи)
  5. Valuation: Business
  6. Valuation: Business
  7. Valuation: Business
  8. Даем описание проблемы в общем. Задача: кто прибыльный а кто убыточный, а кто потенциально интересных. Для каждого такого сегмента должен быть свой подход (разные группы операторов, сегменты выездного и судебного взыскания) СС: пример из другиъ бизнесов (випы, фиши, разные потребности) - сегментация будет всегда.
  9. Вообще в целом есть много классификаций по сегментациям. На примере одной из классификаций мы покажем как это устроено в DCA Главный вопрос Value, Loyalty, Propensity, Demographic, Behavioural Value Chain - ценность клиента или на каком этапе жизненного цикла находится клиент Loyalty - другое измерение, показывает клиентов с точки зрения плательщик, неплательщик и ушедший плательщик Соответственно каждому сегменту на разном этапе жизненного цикла будут применяться разные стратегии воздействия
  10. 3. Сегментация по уровню предрасположенности клиентов платить/идти на контакт и т.п. Как пример - сортировка очереди кол-листа с целью обзванивать более приоритетные дела в первую очередь, а менее приоритетные - по остаточному принципу (в случае undercapacity). Разбивка на заведомо перспективные и заведомо бесперспективные дела учитывая предыдущую историю по другим делам тех же должников 4. Социально-демографическая сегментация используется для формирование спецпредложений для определенных сегментов. Например, праздничные акции (страна, 8марта - пол, возраст - для пенсионеров специальные условия) 5. Поведенческая - здесь пример - в зависимости от того, как респондент отвечает в начале разговора по телефону используются различные скрипты разговоров и аргументация (тут может быть и признание задолженности и согласие платить, и весомые трудности с работой/фин.состоянием и жалобы - во всех этих случаях опертор ведет различные по содержанию переговоры с целью договориться, прийти к соглашению)
  11. Один из подходов к построению моделей в том числе моделей для сегментации - методология crisp-dm. Этот подход позволяет строить релевантные модели.
  12. Segmentation: Math Используются простые модели очень часто инсайты переводятся в простые правила (округляются числа - DPD 170 например будет округлен до 180 что равно 6 месяцам)
  13. Segmentation: Software Engineering
  14. Processing: Business
  15. Критерии ранжирования адресных записей одного должника и их приоритеты следующие: 1. Operator - правило наделяет большим приоритетом адрес, который был создан или изменен оператором. 2. Passport - правило наделяет большим приоритетом адрес, соответствующий паспорту. 3. Validity - правило наделяет большим приоритетом адрес с подтвержденной валидностью 4. Frequency - наделяет большим приоритетом адрес, улица, населенный пункт или город которого встречается чаще среди всех записей персоны, в сравнении с менее приоритетной записью. 5. HouseInMapAndStorage - приоритет адреса повышен, если его дом найден одновременно на карте и в адресном хранилище, в сравнении с менее приоритетной записью. 6. HouseInStorage - приоритет адреса повышен, т.к. его дом найден в адресном хранилище, тогда как у менее приоритетной записи дом найти не удалось. 7. HouseInMap - дом адреса найден на карте, а у менее приоритетной записи дом на карте не нашли. 8. CityAsRegion - адрес выиграл менее приоритетной записи, т.к. он принадлежит городу федерального значения, тогда как менее приоритетный адрес не принадлежит такому городу. 9. RegionCapital - адрес выиграл менее приоритетной записи, т.к. он принадлежит городу, являющемуся столицей региона, тогда как менее приоритетный адрес не принадлежит такому городу. 10. DistrictCenter - адрес выиграл менее приоритетной записи, т.к. он принадлежит городу или населенному пункту, являющемуся районным центром, тогда как менее приоритетный адрес не принадлежит такому городу или населенному пункту. 11. MapLevel - адрес выиграл менее приоритетной записи, т.к. уровень его детализации на карте больше, чем уровень детализации менее приоритетного адреса. 12. AddressType – наделяет большим приоритетом адрес, у которого приоритет типа имеет меньшее значение.
  16. Processing: Business Conversion to improve - Look for segments with low conversion rates or high drop-off rates that can be improved. Your conversion goals are the metrics that you are using to determine whether or not the variation is more successful than the original version. Once you've identified a goal you can begin generating A/B testing ideas and hypotheses for why you think they will be better than the current version. Once you have a list of ideas, prioritize them in terms of expected impact and difficulty of implementation. Create Variations - be prepared for test start - design and parameterize new caseflow or processing strategy, develop new letter/sms template, implement new rules etc. Kick off your experiment! At this point, cases/debtors will be randomly assigned to either the control or variation of your experience. Their interaction with each experience is measured, counted, and compared to determine how each performs. Once your experiment is complete, it's time to analyze the results.Look at the difference between how groups were performed, and whether there is a statistically significant difference. If your variation is a winner, congratulations! See if you can apply learnings from the experiment into production and continue iterating on the experiment to improve your results. If your experiment generates a negative result or no result, don't fret. Use the experiment as a learning experience and generate new hypothesis that you can test.
  17. Одна из методологий для определения оптимальной обработки различных сегментов дел - тесно связана с value chain
  18. Processing: Business