Академический Документы
Профессиональный Документы
Культура Документы
eting programs based on predictive models for customer segmentation, market mix
modeling and demand forecasting. Experience in integrating macro/business condi
tions with microeconomic analysis for financial impact analysis.
Expertise
Market Mix Modeling/promotion lift, multi-channel media optimization, campaign/c
hannel ROI
Demand/sales forecasting, macro/micro business conditions analysis and scenario
simulation
Pricing and elasticity, control tests, microeconomic analysis and cost accountin
g
Strategy and brand position, market opportunity and competitive analysis
CRM and Loyalty programs, segmentation, clustering, CHAID, response modeling
Professional History
Leading Edge Analytics - SAS Alliance R&D Partner - Jan 2001-Current
Assist clients with innovative methods to solve complex marketing problems.
- Supported revamping the consumer credit approval process for a major auto fin
ance company to cope with recalls and poor recent sales by prudently expediting
loan approvals. Analysis examines difference in attrition rates by segment based
on credit factors such as FICO and loan size relative to income. Other researc
h includes attrition forecasting using monthly curves to support earning and cas
h flow forecasting. Projects include extensive support for legacy systems using
SAS macros and Enterprise Guide projects in a server based environment.
- Prototyped an animated video content delivery via mobile phones using a collab
orative filtering survey research process based on fractional sampling and a mul
ti-step path model, to test content, predict consumer attitudes/purchase behavi
or and optimize ad message selection for mobile/online campaigns. Model was esti
mated using SAS macros and proc CALIS structural equations modeling. Concept was
demonstrated to China Unicom and presented at Chinese E Commerce Association to
the Ministry of Technology.
- Developed SmartCompsTM as a branded SaaS application to support Fox Studios po
sition as category manager for Walmart and Target entertainment centers in forec
asting new release DVD sales. SmartComps is an knowledge system combining user
interactive judgment inputs with an automated learning procedure to estimate eco
nometric models and predict DVD sales based on comparable movie theater and sale
s results. Architecture included a web server based SAS stored procedures to int
erface to Oracle database.
- Served as analytics director for Gillette during P&G acquisition to rebuild co
nfidence in Pulsar after the launch was aborted by merger disruptions. Used max
imum likelihood models with extracts from Nielsen sales and trial and repeat pan
el data/takeup curves along with advertising, price, competitive price and in-st
ore promotions to predict weekly sales. Modeling including adbank GRP/TRP and d
iminishing returns effects to drive new product sales. Other models included pr
ice and cross-price and size elasticities for price optimization by chain recomm
endations.
- Used SAS base and macros in a server environment to build a customer/asset ret
ention DM system for a major mutual fund manager to predict customer attrition b
ehavior with Bayesian type of inheritance model with individual "posterior" prob
ability linked to group level "prior" probabilities such as by financial plannin
g agent, portfolio characteristics such as allocation and Morning Star fund rati
ngs and economic measures such as stock market performance and economic conditio
ns. Models were estimated as multi-stage path models and forecast simulations w
ere used for evaluations. Results were provided via drill-down dash boards that
enabled campaign planners to clearly identify customer most-at-risk and type of
risk in order to create customized script themes for call center contacts.
- Used SAS base and regression/logistic regression to support a loyalty program
servicer in retaining retail and online clients, including restaurants, grocery/
drug stores, online and retail fashion merchandisers and services. Analytics in
cluded customer segmentation, advertising/adbank and promotion response modeling
, sales forecasting and assortment turn rates and seasonality. Results showed t
he value of the cross-merchant promotion strategies and end-of-season promotion.
Tools Used
SAS Base and Macros, and Stat: Regression, Logistic Regression
CALIS Structural Equations Modeling - Factor Analysis, Latent Variables
Enterprise Miner for CHAID and neural networks
ETS statespace with canonical correlations and distributed lag
SPSS
AMOS path modeling, hierarchical clustering, factor analysis
Wards Systems
Neural Networks for nonlinear path modeling
ANGOSS
Knowledge Studio - CHAID