predictive analytics: how it works and how to deploy it · eric siegel, ph.d., a senior consultant...
TRANSCRIPT
Prediction Impact
Predictive Analytics: How It Works and How to Deploy It
Predictive Analytics: Predictive Analytics: How It Works and How to Deploy ItHow It Works and How to Deploy It
Eric SiegelEric SiegelPrediction ImpactPrediction [email protected]@predictionimpact.com(415) 385 (415) 385 -- 13131313
© Copyright 2006. All Rights Reserved
Prediction Impact
About Prediction ImpactBased in San Francisco, Prediction Impact provides consulting services and training programs in predictive analytics, data mining, customer intelligence, and text mining.
To subscribe to our tri-annual newsletter, email [email protected]
To receive notifications of upcoming training sessions, email [email protected]
© Copyright 2006. All Rights Reserved
Eric Siegel, Ph.D.Eric Siegel, Ph.D.Prediction ImpactPrediction [email protected]@predictionimpact.com(415) 385 (415) 385 -- 13131313
www.PredictionImpact.comwww.BusinessPrediction.com
If you predict it, you own it.
Prediction Impact
About the speaker.Eric Siegel, Ph.D., a senior consultant at Prediction Impact, provides analytics services for marketing and strategy to medium through Fortune 100 companies. He is a 13 year veteran in data mining and predictive analytics and an award-winning teacher of graduate-level courses in the area. As a professor at Columbia University, Eric forwarded data mining technology in the realms of machine learning performance optimization, text mining and data visualization. He cofounded two New York City-based software companies for customer/user profiling and data mining. With data mining, Eric has also solved problems in computer security, fraud detection, computational linguistics and information retrieval. You can reach him at the contact information below.
© Copyright 2006. All Rights Reserved
Eric Siegel, Ph.D.Eric Siegel, Ph.D.Prediction ImpactPrediction [email protected]@predictionimpact.com(415) 385 (415) 385 -- 13131313
www.PredictionImpact.comwww.BusinessPrediction.com
If you predict it, you own it.
Outline of TopicsOutline of Topics
1. Predictive analytics: What it is and what it does for you2. How predictive models work
– How they are created– What they look like on the inside
3. Applications of predictive analytics– Deployment scenarios -- how models integrate– Example models
4. What it takes, organizationally, to employ predictive analytics
Prediction Impact© Copyright 2006. All Rights Reserved
Your Data is a Core Strategic AssetYour Data is a Core Strategic Asset
Data encodes your business’ collective experience• Sales records
• Customer profiles
• Campaign histories
• Operational logs
• Web logs
Prediction Impact© Copyright 2006. All Rights Reserved
Leverage Your Data as an AssetLeverage Your Data as an Asset
It is becoming increasingly imperative to:
Learn from your data.
Learn as much as possible about your customers.
Learn how to treat each customerindividually.
Prediction Impact© Copyright 2006. All Rights Reserved
Predictive Analytics:Predictive Analytics:
The IT and organizational techniques designed to:
• Learn from an organization’s experience
• Produce actionable predictions for each customer
• Respond accordingly with the right action
Prediction Impact
In doing so, you will:• Know your customers• Personalize the customer experience• Perform market segmentation• Discover what's been working
© Copyright 2006. All Rights Reserved
Predictive Analytics:Predictive Analytics:
Foresee the next action of each customer:– Purchase– Response– Cancellation– Conversion
A rose by any other name:Data mining – includes predictive analyticsMachine learning – the academic research fieldPredictive modeling – the core statistical method
Prediction Impact© Copyright 2006. All Rights Reserved
Predictive ModelingPredictive Modeling
Prediction Impact
Customerprofiles
Customerhistories
Data mining
Campaignhistories
Prediction model
© Copyright 2006. All Rights Reserved
Fine-Tuned Models for Your BusinessFine-Tuned Models for Your Business
• Prediction model created automatically from your data• According to your:
– Product– Prediction goal– Business model– Customer base
• Specialized for your business
Prediction Impact© Copyright 2006. All Rights Reserved
Predictive Insight for MarketingPredictive Insight for Marketing
Prediction Impact
Customerprofile
Customerhistory
Prediction modelProspective campaign
Predicted response
© Copyright 2006. All Rights Reserved
Taking Business ActionTaking Business Action
Prediction Impact
Predicted response
Business Logic Business action:•Mail a solicitation•Suggest a cross-sell option•Retain with a promotion
© Copyright 2006. All Rights Reserved
Response Modeling for Direct MarketingResponse Modeling for Direct Marketing
Lower campaign costs and increase response rates
• Make campaigns more targeted and more selective
• Identify segments five or more times as responsive
• Achieve 80% of responses with just 40% of mailing
• Increase campaign ROI & profitability
Prediction Impact© Copyright 2006. All Rights Reserved
Predictors: Building Blocks for ModelsPredictors: Building Blocks for Models
Prediction Impact
• recency– How recent was the last purchase?
• personal income– How much to spend?
Combine predictors for better rankings
recency + personal income
2*recency + personal income
© Copyright 2006. All Rights Reserved
Predictive Models Score Each CustomerPredictive Models Score Each Customer
• Each customer is scored with a response probability• The customers are listed in order of prediction score• The highest-scoring customers are targeted first
Prediction Impact
No �14%T. Mitchel674
Yes ☺22%G. Clooney256
No �31%D. Leong528
Yes ☺35%E. Siegel429
Response:Score:Name:ID number:
© Copyright 2006. All Rights Reserved
Campaign Profit CurveCampaign Profit CurveCampaign Profit Curve
Applications of Predictive AnalyticsApplications of Predictive Analytics
Prediction Impact
Attrition Prediction
Which current customers are going.
Customer Valuation
What each customer does in between.
Response Modeling
Which prospective customers are coming.
© Copyright 2006. All Rights Reserved
Specific Example ApplicationsSpecific Example Applications
Prediction Impact
Attrition Prediction
Telecom:Increase retention campaign efficiency by 300%; reduce churn 3% and recover $3.8 million.
Customer Valuation
Pharmaceutical:Invest in health providers most likely to convert prescriptions in the long-term.
Response Modeling
e-Commerce:Increase online ad take-rates by 28%
© Copyright 2006. All Rights Reserved
Applications of Predictive AnalyticsApplications of Predictive Analytics
Prediction Impact
Attrition Prediction
• Churn reduction• Customer retention• Customer renewal
Customer Valuation
• Hit-and-run prediction• Present lifetime value• First year revenue
Response Modeling
• Direct marketing• Acquisition• Conversion• Market optimization• Promotion targeting• Cross-selling• Up-selling
© Copyright 2006. All Rights Reserved
Customer Retention with Attrition PredictionCustomer Retention with Attrition Prediction
• Customer acquisition is more expensive than retention.• Retention campaigns are targeted with attrition models.• Small improvements in retention generate great returns:
– Decreasing churn from 10% to 5% increases net present value by 75%.*
– When monthly churn is high, reducing it by 3% can increase growth 12%.• Plus the compounding effect.• This translates to over $500,000 annually for a monthly online subscription service
with 200,000 customers.
*The Loyalty Effect, by Reichheld & Teal
Prediction Impact© Copyright 2006. All Rights Reserved
Retention Model: Business RulesRetention Model: Business Rules
Prediction Impact
IfIfIfIf the customer is rural, andandandand her monthly usage is high,
thenthenthenthen the customer will probably renew.
© Copyright 2006. All Rights Reserved
Decision Tree for Direct MarketingDecision Tree for Direct Marketing
Prediction Impact
Florida
Georgia Customer > 7 yrs
High Low High Low
2.6%
1.5%5.1%
7.4% 1.2% 9.5% 1.4%
© Copyright 2006. All Rights Reserved
Decision TreeDecision Tree
Prediction Impact© Copyright 2006.
A More Realistic Decision TreeA More Realistic Decision Tree
Prediction Impact© Copyright 2006. All Rights Reserved
Neural NetworksNeural Networks
� A more complexmathematical formula
Prediction Impact© Copyright 2006. All Rights Reserved
The “Bad” NewsThe “Bad” News
Predictive analytics is a business activity, not an IT activity.
Prediction Impact© Copyright 2006. All Rights Reserved
Good!
The Mining Business ProcessThe Mining Business Process
• Wholly collaborative process• Driven by business needs and marketing insight• Ensure actionable results
– Within your organization's operational framework– Greatest impact within your business model
Prediction Impact© Copyright 2006. All Rights Reserved
Analytics Process for Customer PredictionAnalytics Process for Customer Prediction
Prediction Impact
Customerprofiles
Customerhistories
Data mining
Campaignhistories
Prediction model
Prediction goal Evaluation withbusiness expertise
© Copyright 2006. All Rights Reserved
Predictive Analytics: SummaryPredictive Analytics: Summary
1. Actionable predictions for each customer
2. Tactical deployment yields strategic results
3. Requires a business process, driven by business pragmatics and business needs
Prediction Impact© Copyright 2006. All Rights Reserved
Closing Data Mining PunsClosing Data Mining Puns
Mine your own business!
What's mined is yours!
Prediction Impact© Copyright 2006. All Rights Reserved