Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Predictive analytics: only heard the good stuff?
Analytics 2017 20 April 2017
Nigel Harrison
Head of Customer Analytics
THIS PRESENTATION IS FOR INFORMATIONAL PURPOSES ONLY AND SHALL NOT BE CONSTRUED TO CONSTITUTE ADVICE.
AIG ASIA PACIFIC INSURANCE PTE. LTD. (“AIG SINGAPORE”) AND AIG APAC HOLDINGS PTE. LTD. MAKE NO REPRESENTATION OR WARRANTIES AS TO THE ACCURACY, COMPLETENESS OR TIMELINESS OF THE INFORMATION, TEXT, GRAPHICS OR OTHER ITEMS CONTAINED IN
2
THIS PRESENTATION. AIG SINGAPORE EXPRESSLY DISCLAIMS ALL LIABILITY FOR ERRORS OR OMISSIONS IN, OR THE MISUSE OR MISINTERPRETATION OF, ANY INFORMATION CONTAINED IN THIS PRESENTATION.
Apply findings
Build model
3
Build model
Get data
What problem are you trying to solve exactly?
4
exactly?
400
500
600
700
800
900
1,000
Num
ber
of S
ale
s
Marginal cost per
sale
Unit cost Number
of Sales
5
0
100
200
300
400
Num
ber
of S
ale
s
Marketing cost
Unit cost per sale
of Sales
Now that you know, what are you going to do about it?
6
going to do about it?
Predict something you can change
Identify key interactions
7
Identify key interactions
Consider what’s under your control
Is the most predictive model always the best one for the job?
8
the best one for the job?
9
AIG On the Go
The Intelligent Driving app for Singapore
10
Biased data is worse than no data.
11
Biased data is worse than no data.
On two occasions I have been
asked,
“Pray, Mr Babbage, if you put into
the machine wrong figures, will the
right answers come out?"
... I am not able rightly to
12
... I am not able rightly to
apprehend the kind of confusion of
ideas that could provoke such a
question.
Charles Babbage, 1864
Quality data
� Timely
� Accurate
� Complete
� Consistent
+ Contrasting
+ Unbiased
13
� Consistent
� Relevant
� Appropriate detail
You can’t infer
propensities by
looking only at
buyer profiles.
14
Non-buyers
buyer profiles.
Build contrasts
into the dataset.
You cannot change
the laws of physics.
Don’t accept
15
Don’t accept
leakage from the
future.
Put the science
into data science.
Experiment.
16
Experiment.
Solve problem
Build model
17
Build model
Curate data
18
American International Group, Inc. (AIG) is a leading global insurance organization. Founded in 1919, today AIG member companies provide a wide range of property casualty insurance, life insurance,
retirement products, mortgage insurance and other financial services to customers in more than 100 countries and jurisdictions. These diverse offerings include products and services that help businesses
and individuals protect their assets, manage risks and provide for retirement security. AIG common stock is listed on the New York Stock Exchange and the Tokyo Stock Exchange.
Additional information about AIG can be found at www.aig.com and www.aig.com/strategyupdate | YouTube: www.youtube.com/aig | Twitter: @AIGinsurance | LinkedIn:
http://www.linkedin.com/company/aig. These references with additional information about AIG have been provided as a convenience, and the information contained on such websites is not incorporated by
reference into this presentation.
AIG is the marketing name for the worldwide property-casualty, life and retirement, and general insurance operations of American International Group, Inc. For additional information, please visit our
website at www.aig.com. All products and services are written or provided by subsidiaries or affiliates of American International Group, Inc. Products or services may not be available in all countries, and
coverage is subject to actual policy language. Non-insurance products and services may be provided by independent third parties. Certain property-casualty coverages may be provided by a surplus lines
insurer. Surplus lines insurers do not generally participate in state guaranty funds, and insureds are therefore not protected by such funds.