Client Predictive Analytics Proposal PDF
Client Predictive Analytics Proposal PDF
Client Predictive Analytics Proposal PDF
Predictive
Analytics
Confidential Proprietary
Information of Innovation
Solutions
-
Not to be distributed to
“Client Proposal”
other parties without prior
written authorization
Sampling Distribution of Profitable Customers
0.10
0.08
Probability of 0.06 False Alarm:
Success 0.04 Pr (P<173 = 5 %)
0.02
0.00
140 150 160 170 180 190 200
Number of Successes
Confidential Page 1 of 7
Predictive Analytics Proposal
Table of Contents
1 Business Opportunity ............................................................................................... 3
2 Solution ...................................................................................................................... 3
2.1 Objectives....................................................................................................................... 3
2.2 Approach ........................................................................................................................ 4
2.3 Benefits .......................................................................................................................... 5
3 Implementation Plan ................................................................................................. 5
3.1 CRISP Work Breakdown Structure ................................................................................ 6
3.2 Resources and Roles ..................................................................................................... 6
3.3 Qualifications .................................................................................................................. 6
4 Costs .......................................................................................................................... 7
Confidential Page 2 of 7
Predictive Analytics Proposal
2.1 Objectives
What service 2.1.1. Summary statistical analysis of product, customer, marketing
characteristics are and financial data to:
responsible for • Identify past and current key measures
customer retention?
• Clearly define the extent and source of data available for
mining and predictive analytics
2.1.2. Develop a customer profile for specific product purchases
using either factor analysis or correspondence analysis.
2.1.3. Identify predictor variables for marketing campaigns using
regression analysis.
2.1.4. Develop a customer valuation classification, i.e. identify most
profitable customers and key indices of their retention using
ANCOVA (Analysis of Covariance)
2.1.5. Develop a wholesaler segmentation using cluster analysis.
Confidential Page 3 of 7
Predictive Analytics Proposal
2.2 Approach
The project delivery process will apply the industry standard data
mining implementation process called CRISP. (CRoss-Industry
Standard Process for Data Mining). The lifecycle of the CRISP
process is shown in Figure 1.0 below. The cycle consists of 6
phases which align with the major deliverable of each phase.
Confidential Page 4 of 7
Predictive Analytics Proposal
2.3 Benefits
The key benefits of combining these processes are:
1. Each ‘SPRINT’ will answer specific business questions.
2. Each ‘SPRINT’ will allow the evaluation of a specific analytic
technique.
3. Each ‘SPRINT’ will inevitably propose data needs whose
capture can be initiated during further iterations.
4. Each ‘SPRINT’ will introduce a new analytic technique as
well as reinforcing previous statistical terminology and
concepts.
5. Evaluation of competing technologies can be performed in
concert with each ‘SPRINT’. As each set of questions and
the corresponding model are created the ease of use and
cost associated with implementing these models in various
software technologies can be evaluated.
3 Implementation Plan
The CRISP methodology has an easy to follow and well defined
Project is scheduled Work Breakdown Structure that provides the basis for developing a
for 30 Weeks: refined project plan. Figure 2.0 below shows the high level tasks
within each phase of the project. Again, each ‘SPRINT’ would be
conducted over a 5 week period. Each objective listed in 2.1 above
May - November 2009
would be addressed in a single ‘SPRINT’. Therefore the complete
project including deployment of the final software architecture and
training would be conducted in 30 weeks. (6 objectives conducted
over 5 cycles).
Again, the incorporation of the Agile methodology allows for a
specific objective to be completed and presented to the business
stakeholders every 5 weeks, thereby answering specific business
needs, training personnel on the analytic techniques and reinforcing
value and reinforcing the predictive analytics methodology.
Confidential Page 5 of 7
Predictive Analytics Proposal
3.3 Qualifications
Dr. Peter Karpiuk has extensive training and experience developing
mathematical models and performing statistical analysis both in
academia and industry. Below are a list of available presentations
and publications:
Industry White Papers
Life Composite Rating Algorithm: A Single Parameter Approach:
BEST Life Insurance, 2008
Quality Assurance of Provider Directory Search Accuracy using a
Sequential Probability Ratio Test: Word and Brown, presented at
the Southern California Coalition Technology Conference, 2007.
Measuring the ROI of Host Interface Implementations: A Goodness
of Fit (Chi-Square) Methodology: Optum Software, 2001
Confidential Page 6 of 7
Predictive Analytics Proposal
Academic Publications
Clark, M. M, Karpiuk, P and Galef, B.J. Jr.(1993). Hormonally
mediated inheritance of acquired characteristics in Mongolian
gerbils. Nature, 364, p. 712.
Ginsburg, G. and Karpiuk, P. (1994) Random generation: Analysis of
responses. Perceptual and Motor Skills, 79, pp.1059-1067.
Karpiuk, P. (1994) Fechner and Stevens: Vive la difference! Les
Cahiers de Recherche de l'Ecole de Psychologie, 164 - 196.
Ginsburg, N., & Karpiuk, P. (1995). Simulation of human
performance on a random generation task. Perceptual and Motor
Skills, 81,. 1183–1186.
Karpiuk P., Jr., Lacouture Y., Marley Y A. A. J. (1997). A limited
capacity, wave equality, random walk model of absolute
identification. In Marley A. A. J. (Ed.), Choice, decision, and
measurement: Essays in honor of R. Duncan Luce (pp. 279-299).
Mahwah, NJ: Erlbaum.
4 Costs
The project costs listed below are for statistical consulting and
project management. All software and licensing costs will be
determined by MetLife based upon their technology selection.
Total Hours
Cost Schedule Weekly Hours Cost/Hour based upon 6 Total Cost
SPRINTS
Statistical
20 $ 600 $
Consulting
Project “Client”
8 240
Management 1 Employee
Business
“Client” * Costs to be
Subject Matter 10 300 provided by
Employee
Expert “Client” to
Data Base “Client” determine Total
10 300 Project Costs.
Analyst Employee
Table 1.0: Consulting Services Costs. Total Project Costs are based upon 20 hours per week.
1.
NOTE: Project Management duties can be performed as an adjunct to the statistical consulting
services at an industry rate of $.00/hour.
Confidential Page 7 of 7