SlideShare a Scribd company logo
1 of 9
Download to read offline
1 © Experian 25/10/2019
Sip and Solve
Signs you need to adjust your
decision model
Featuring:
Marsha Silverman
Lead Analytic Consultant
Experian Commercial Decision Sciences
2 © Experian Sip and Solve: Signs you need to adjust your decision model
What is a model?
An engine or mathematical formula that distinguishes the relationships between “random
information” and an event.
• Used to guide a business area to make decisions at key points in the customer life cycle.
• Helps give the likelihood of particular ‘outcome’ for individual accounts or transaction.
• Based on the historic behavior of similar types of accounts.
• Examples of use cases:
• Will an account go delinquent within the next 12 months?
• What is the likelihood a customer will respond to a marketing offer?
• Who is more likely to pay its delinquent debt?
3 © Experian
K-S or Kolmogorov-Smirnov measures performance of classification models
• It is a measure of the degree of separation between the positive and negative distributions.
• In most classification models, the K-S will fall between 0 and 100, where the higher the value, the better the model is at separating
the positive from negative cases.
GINI is similar to KS and measures the area under the curve
Bad Capture rate measures how many “bads” are captured in the bottom percentage of the population
Cumulative%bads
Cumulative % goods
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
How do we know if a model “works”?
4 © Experian Sip and Solve: Signs you need to adjust your decision model
Why do models deteriorate?
One of the chief assumptions in modelling is that the future will be a lot like the past.
However,
• Change is inevitable
• Change is continuous
• Businesses change how they conduct business
• There are changes in external factors such as the economy, politics, regulations, etc.
Therefore,
• Scorecards will degrade over time as a byproduct of other changes
• If changes are gradual, so, too, will be the process of degradation
5 © Experian
Starting population – Accounts opened in Q1 2018
Jan – Mar 2018
Observation period
Jan – Mar 2019
12-month rolling performance window from opening
Performance definitions
Bad
Businesses with C/O or
91+ delinquency
Good
Businesses without Bad
statuses
End of performance period
Accounts would be retro-scored as of the observation point to determine how well the score
predicts the actual behavior experienced during the performance window.
Sip and Solve: Signs you need to adjust your decision model
Performing periodic score validations will show if a model deteriorates
• For example, if a model was developed using new accounts from Q1 2017, a validation may be
done on accounts from Q1 2018
How do we know if a model still “works”?
6 © Experian Sip and Solve: Signs you need to adjust your decision model
Comparing current year (CY) to prior year (PY) demonstrates if deterioration has occurred
Experian performs annual score validations
7 © Experian Sip and Solve: Signs you need to adjust your decision model
A population shift can mask the effectiveness of a model. Assessing the population stability index (PSI) can
be helpful to determine if the model still works but the population has shifted.
The Population Stability Index (PSI) quantifies shifts in score distributions for two populations. Each score
distribution is separated into buckets and the percent of records in each bucket is calculated.
The PSI is calculated as: PSI=Sum of { LN (percent/percent2)*(percent-percent2) } over all buckets.
Results are interpreted with the following parameters:
PSI > 0.25 – Significant change in population
PSI between 0.1 and 0.25 – Some minor changes
PSI<= 0.1 – No significant changes in the population
Population stability
8 © Experian Sip and Solve: Signs you need to adjust your decision model
If a model has been found to have deterioration and/or there is a large shift in population running
through the model:
• Models may be recalibrated using the same predictors but with different co-efficients
• Models may be fully redeveloped on a new sample population
What should be done?
9 © Experian 25/10/2019
Thank you!
Have questions or want to learn more about when to adjust your model?
• Send us a message in the “Chat” function or via bit.ly/sip-solve-chat
Upcoming Sip and Solves -
• Join the list to stay in the loop: bit.ly/sip-and-solve

More Related Content

More from Experian

Choosing The Right Credit Decisioning Model
Choosing The Right Credit Decisioning ModelChoosing The Right Credit Decisioning Model
Choosing The Right Credit Decisioning ModelExperian
 
Make Smarter Collections Decisions with Analytics
Make Smarter Collections Decisions with AnalyticsMake Smarter Collections Decisions with Analytics
Make Smarter Collections Decisions with AnalyticsExperian
 
Best Practices to Identify Companies in the Market for Business Credit
Best Practices to Identify Companies in the Market for Business CreditBest Practices to Identify Companies in the Market for Business Credit
Best Practices to Identify Companies in the Market for Business CreditExperian
 
Collections prioritization using a scorecard
Collections prioritization using a scorecardCollections prioritization using a scorecard
Collections prioritization using a scorecardExperian
 
Visualizing Portfolio Data
Visualizing Portfolio DataVisualizing Portfolio Data
Visualizing Portfolio DataExperian
 
Using Blended Business Owner Data in Credit Decision Making
Using Blended Business Owner Data in Credit Decision MakingUsing Blended Business Owner Data in Credit Decision Making
Using Blended Business Owner Data in Credit Decision MakingExperian
 
Alternative Data in Decisioning Models
Alternative Data in Decisioning ModelsAlternative Data in Decisioning Models
Alternative Data in Decisioning ModelsExperian
 
Experian Women in Business Credit Study
Experian Women in Business Credit StudyExperian Women in Business Credit Study
Experian Women in Business Credit StudyExperian
 
Measuring the Business Impact of Hurricane Harvey
Measuring the Business Impact of Hurricane HarveyMeasuring the Business Impact of Hurricane Harvey
Measuring the Business Impact of Hurricane HarveyExperian
 
The Face of Small Business Infographic
The Face of Small Business InfographicThe Face of Small Business Infographic
The Face of Small Business InfographicExperian
 
8 Easy Steps to Report Data to Experian
8 Easy Steps to Report Data to Experian8 Easy Steps to Report Data to Experian
8 Easy Steps to Report Data to ExperianExperian
 
The State of Minority Owned Small Business
The State of Minority Owned Small BusinessThe State of Minority Owned Small Business
The State of Minority Owned Small BusinessExperian
 
2016 Election: Will the political affiliation of small-business owners have a...
2016 Election: Will the political affiliation of small-business owners have a...2016 Election: Will the political affiliation of small-business owners have a...
2016 Election: Will the political affiliation of small-business owners have a...Experian
 
Turning Data Into Insight
Turning Data Into InsightTurning Data Into Insight
Turning Data Into InsightExperian
 
The State of Women Owned Businesses
The State of Women Owned BusinessesThe State of Women Owned Businesses
The State of Women Owned BusinessesExperian
 
Men Business Owner Study
Men Business Owner StudyMen Business Owner Study
Men Business Owner StudyExperian
 
Women Business Owner Snapshot
Women Business Owner SnapshotWomen Business Owner Snapshot
Women Business Owner SnapshotExperian
 
Big Data Being Used For Good
Big Data Being Used For GoodBig Data Being Used For Good
Big Data Being Used For GoodExperian
 
10 Common Reasons For Small Business Failure
10 Common Reasons For Small Business Failure10 Common Reasons For Small Business Failure
10 Common Reasons For Small Business FailureExperian
 
Case Study - Wurth Louis and Company
Case Study - Wurth Louis and CompanyCase Study - Wurth Louis and Company
Case Study - Wurth Louis and CompanyExperian
 

More from Experian (20)

Choosing The Right Credit Decisioning Model
Choosing The Right Credit Decisioning ModelChoosing The Right Credit Decisioning Model
Choosing The Right Credit Decisioning Model
 
Make Smarter Collections Decisions with Analytics
Make Smarter Collections Decisions with AnalyticsMake Smarter Collections Decisions with Analytics
Make Smarter Collections Decisions with Analytics
 
Best Practices to Identify Companies in the Market for Business Credit
Best Practices to Identify Companies in the Market for Business CreditBest Practices to Identify Companies in the Market for Business Credit
Best Practices to Identify Companies in the Market for Business Credit
 
Collections prioritization using a scorecard
Collections prioritization using a scorecardCollections prioritization using a scorecard
Collections prioritization using a scorecard
 
Visualizing Portfolio Data
Visualizing Portfolio DataVisualizing Portfolio Data
Visualizing Portfolio Data
 
Using Blended Business Owner Data in Credit Decision Making
Using Blended Business Owner Data in Credit Decision MakingUsing Blended Business Owner Data in Credit Decision Making
Using Blended Business Owner Data in Credit Decision Making
 
Alternative Data in Decisioning Models
Alternative Data in Decisioning ModelsAlternative Data in Decisioning Models
Alternative Data in Decisioning Models
 
Experian Women in Business Credit Study
Experian Women in Business Credit StudyExperian Women in Business Credit Study
Experian Women in Business Credit Study
 
Measuring the Business Impact of Hurricane Harvey
Measuring the Business Impact of Hurricane HarveyMeasuring the Business Impact of Hurricane Harvey
Measuring the Business Impact of Hurricane Harvey
 
The Face of Small Business Infographic
The Face of Small Business InfographicThe Face of Small Business Infographic
The Face of Small Business Infographic
 
8 Easy Steps to Report Data to Experian
8 Easy Steps to Report Data to Experian8 Easy Steps to Report Data to Experian
8 Easy Steps to Report Data to Experian
 
The State of Minority Owned Small Business
The State of Minority Owned Small BusinessThe State of Minority Owned Small Business
The State of Minority Owned Small Business
 
2016 Election: Will the political affiliation of small-business owners have a...
2016 Election: Will the political affiliation of small-business owners have a...2016 Election: Will the political affiliation of small-business owners have a...
2016 Election: Will the political affiliation of small-business owners have a...
 
Turning Data Into Insight
Turning Data Into InsightTurning Data Into Insight
Turning Data Into Insight
 
The State of Women Owned Businesses
The State of Women Owned BusinessesThe State of Women Owned Businesses
The State of Women Owned Businesses
 
Men Business Owner Study
Men Business Owner StudyMen Business Owner Study
Men Business Owner Study
 
Women Business Owner Snapshot
Women Business Owner SnapshotWomen Business Owner Snapshot
Women Business Owner Snapshot
 
Big Data Being Used For Good
Big Data Being Used For GoodBig Data Being Used For Good
Big Data Being Used For Good
 
10 Common Reasons For Small Business Failure
10 Common Reasons For Small Business Failure10 Common Reasons For Small Business Failure
10 Common Reasons For Small Business Failure
 
Case Study - Wurth Louis and Company
Case Study - Wurth Louis and CompanyCase Study - Wurth Louis and Company
Case Study - Wurth Louis and Company
 

Recently uploaded

Generative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfGenerative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfEmmanuel Dauda
 
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...Amil baba
 
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...ssuserf63bd7
 
Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Jon Hansen
 
Pre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxPre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxStephen266013
 
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Valters Lauzums
 
The Significance of Transliteration Enhancing
The Significance of Transliteration EnhancingThe Significance of Transliteration Enhancing
The Significance of Transliteration Enhancingmohamed Elzalabany
 
一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理cyebo
 
Fuzzy Sets decision making under information of uncertainty
Fuzzy Sets decision making under information of uncertaintyFuzzy Sets decision making under information of uncertainty
Fuzzy Sets decision making under information of uncertaintyRafigAliyev2
 
basics of data science with application areas.pdf
basics of data science with application areas.pdfbasics of data science with application areas.pdf
basics of data science with application areas.pdfvyankatesh1
 
社内勉強会資料  Mamba - A new era or ephemeral
社内勉強会資料   Mamba - A new era or ephemeral社内勉強会資料   Mamba - A new era or ephemeral
社内勉強会資料  Mamba - A new era or ephemeralNABLAS株式会社
 
Artificial_General_Intelligence__storm_gen_article.pdf
Artificial_General_Intelligence__storm_gen_article.pdfArtificial_General_Intelligence__storm_gen_article.pdf
Artificial_General_Intelligence__storm_gen_article.pdfscitechtalktv
 
一比一原版麦考瑞大学毕业证成绩单如何办理
一比一原版麦考瑞大学毕业证成绩单如何办理一比一原版麦考瑞大学毕业证成绩单如何办理
一比一原版麦考瑞大学毕业证成绩单如何办理cyebo
 
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理pyhepag
 
2024 Q1 Tableau User Group Leader Quarterly Call
2024 Q1 Tableau User Group Leader Quarterly Call2024 Q1 Tableau User Group Leader Quarterly Call
2024 Q1 Tableau User Group Leader Quarterly Calllward7
 
Exploratory Data Analysis - Dilip S.pptx
Exploratory Data Analysis - Dilip S.pptxExploratory Data Analysis - Dilip S.pptx
Exploratory Data Analysis - Dilip S.pptxDilipVasan
 
Easy and simple project file on mp online
Easy and simple project file on mp onlineEasy and simple project file on mp online
Easy and simple project file on mp onlinebalibahu1313
 
Formulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfFormulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfRobertoOcampo24
 

Recently uploaded (20)

Generative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfGenerative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdf
 
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
 
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
 
Abortion pills in Dammam Saudi Arabia// +966572737505 // buy cytotec
Abortion pills in Dammam Saudi Arabia// +966572737505 // buy cytotecAbortion pills in Dammam Saudi Arabia// +966572737505 // buy cytotec
Abortion pills in Dammam Saudi Arabia// +966572737505 // buy cytotec
 
Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)
 
Pre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxPre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptx
 
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
 
The Significance of Transliteration Enhancing
The Significance of Transliteration EnhancingThe Significance of Transliteration Enhancing
The Significance of Transliteration Enhancing
 
一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理
 
Fuzzy Sets decision making under information of uncertainty
Fuzzy Sets decision making under information of uncertaintyFuzzy Sets decision making under information of uncertainty
Fuzzy Sets decision making under information of uncertainty
 
basics of data science with application areas.pdf
basics of data science with application areas.pdfbasics of data science with application areas.pdf
basics of data science with application areas.pdf
 
社内勉強会資料  Mamba - A new era or ephemeral
社内勉強会資料   Mamba - A new era or ephemeral社内勉強会資料   Mamba - A new era or ephemeral
社内勉強会資料  Mamba - A new era or ephemeral
 
Artificial_General_Intelligence__storm_gen_article.pdf
Artificial_General_Intelligence__storm_gen_article.pdfArtificial_General_Intelligence__storm_gen_article.pdf
Artificial_General_Intelligence__storm_gen_article.pdf
 
一比一原版麦考瑞大学毕业证成绩单如何办理
一比一原版麦考瑞大学毕业证成绩单如何办理一比一原版麦考瑞大学毕业证成绩单如何办理
一比一原版麦考瑞大学毕业证成绩单如何办理
 
Machine Learning for Accident Severity Prediction
Machine Learning for Accident Severity PredictionMachine Learning for Accident Severity Prediction
Machine Learning for Accident Severity Prediction
 
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
 
2024 Q1 Tableau User Group Leader Quarterly Call
2024 Q1 Tableau User Group Leader Quarterly Call2024 Q1 Tableau User Group Leader Quarterly Call
2024 Q1 Tableau User Group Leader Quarterly Call
 
Exploratory Data Analysis - Dilip S.pptx
Exploratory Data Analysis - Dilip S.pptxExploratory Data Analysis - Dilip S.pptx
Exploratory Data Analysis - Dilip S.pptx
 
Easy and simple project file on mp online
Easy and simple project file on mp onlineEasy and simple project file on mp online
Easy and simple project file on mp online
 
Formulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfFormulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdf
 

Signs you need to adjust your decision model - Sip and Solve

  • 1. 1 © Experian 25/10/2019 Sip and Solve Signs you need to adjust your decision model Featuring: Marsha Silverman Lead Analytic Consultant Experian Commercial Decision Sciences
  • 2. 2 © Experian Sip and Solve: Signs you need to adjust your decision model What is a model? An engine or mathematical formula that distinguishes the relationships between “random information” and an event. • Used to guide a business area to make decisions at key points in the customer life cycle. • Helps give the likelihood of particular ‘outcome’ for individual accounts or transaction. • Based on the historic behavior of similar types of accounts. • Examples of use cases: • Will an account go delinquent within the next 12 months? • What is the likelihood a customer will respond to a marketing offer? • Who is more likely to pay its delinquent debt?
  • 3. 3 © Experian K-S or Kolmogorov-Smirnov measures performance of classification models • It is a measure of the degree of separation between the positive and negative distributions. • In most classification models, the K-S will fall between 0 and 100, where the higher the value, the better the model is at separating the positive from negative cases. GINI is similar to KS and measures the area under the curve Bad Capture rate measures how many “bads” are captured in the bottom percentage of the population Cumulative%bads Cumulative % goods 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 How do we know if a model “works”?
  • 4. 4 © Experian Sip and Solve: Signs you need to adjust your decision model Why do models deteriorate? One of the chief assumptions in modelling is that the future will be a lot like the past. However, • Change is inevitable • Change is continuous • Businesses change how they conduct business • There are changes in external factors such as the economy, politics, regulations, etc. Therefore, • Scorecards will degrade over time as a byproduct of other changes • If changes are gradual, so, too, will be the process of degradation
  • 5. 5 © Experian Starting population – Accounts opened in Q1 2018 Jan – Mar 2018 Observation period Jan – Mar 2019 12-month rolling performance window from opening Performance definitions Bad Businesses with C/O or 91+ delinquency Good Businesses without Bad statuses End of performance period Accounts would be retro-scored as of the observation point to determine how well the score predicts the actual behavior experienced during the performance window. Sip and Solve: Signs you need to adjust your decision model Performing periodic score validations will show if a model deteriorates • For example, if a model was developed using new accounts from Q1 2017, a validation may be done on accounts from Q1 2018 How do we know if a model still “works”?
  • 6. 6 © Experian Sip and Solve: Signs you need to adjust your decision model Comparing current year (CY) to prior year (PY) demonstrates if deterioration has occurred Experian performs annual score validations
  • 7. 7 © Experian Sip and Solve: Signs you need to adjust your decision model A population shift can mask the effectiveness of a model. Assessing the population stability index (PSI) can be helpful to determine if the model still works but the population has shifted. The Population Stability Index (PSI) quantifies shifts in score distributions for two populations. Each score distribution is separated into buckets and the percent of records in each bucket is calculated. The PSI is calculated as: PSI=Sum of { LN (percent/percent2)*(percent-percent2) } over all buckets. Results are interpreted with the following parameters: PSI > 0.25 – Significant change in population PSI between 0.1 and 0.25 – Some minor changes PSI<= 0.1 – No significant changes in the population Population stability
  • 8. 8 © Experian Sip and Solve: Signs you need to adjust your decision model If a model has been found to have deterioration and/or there is a large shift in population running through the model: • Models may be recalibrated using the same predictors but with different co-efficients • Models may be fully redeveloped on a new sample population What should be done?
  • 9. 9 © Experian 25/10/2019 Thank you! Have questions or want to learn more about when to adjust your model? • Send us a message in the “Chat” function or via bit.ly/sip-solve-chat Upcoming Sip and Solves - • Join the list to stay in the loop: bit.ly/sip-and-solve