Chief Data Scientist

1/2014 – 6/2018

Supported their marketing efforts through the maintenance of 1,000+ equations globally. The data and models get refreshed on a monthly basis and include propensity-to-buy models, retention models, life-time value and satisfaction scores.

Statistiken

1/2005 – 1/2015

I supported IBM’s tactical, as well as strategic marketing in multiple ways. Key tasks include concept development and implementation of traditional B2B marketing through the use of predictive analytics. Other tasks included the creation of a dynamic CRM modeling framework that allows for real time scoring of IBM’s sales pipeline in order to predict time-to-close and likelihood-to-win for every sales opportunity world wide. Besides having very actionable predictions that have helped IBM meet its quarterly goals in all of its divisions (software, hardware, services), the dynamic CRM modeling environment also helped in the identification of emerging trends through the study of the changes in the model parameters and other key factors. We also developed the concept of eNurture, a way of nurturing leads through the sales cycle by predicting next marketing touches.

IBM Cognos, Spss, SAS (Software), Lotus Notes Script

1/2005 – 6/2007

Migrated a 4.5billion+ hedge fund from multiple systems, data bases, and languages into a single SAS environment. The system, developed within budget, allows for rapid sampling of all major data providers (IBES, CompuStat, Barra, etc) to create the required data sets for portfolio optimizations and evaluations. The system creates time-series, as well as cross-sectional data sets in seconds, compared to weeks using the old system. The data sets are split-adjusted, takeover cleaned, and have all ticker changes handled. To create the system, 100’s of scripts, ranging from pearl to oracle had to be translated (in some cases reverse engineered due to loss of source code). In addition, a common key was invented to make the merging of data sets transparent.

SAS (Software)

10/2002 – 12/2005

VCA provides veterinary services in a retail environment. I designed, sourced, and built a CRM system which utilizes 4-years of transactional data from Oracle warehouses and integrated SAS statistical/data mining solutions. Currently approx. 40 product specific statistical models are built and scored on a monthly basis to increase product penetration for various animal products and treatments. One product model is built for 1st-time and repeat buyers, separately. The system increases revenues by $25million on a $450million revenue baseline. The CRM system includes a budgeting mechanism based on potential revenue gains – making all channel marketing dollars focused and accountable. In addition to the internal up-selling and cross-selling through various channels, client acquisition models were built, and a routine monthly mailing established. This includes backend analysis and response feedback loop to refine modeling efforts. Currently the acquisition engine adds 50,000 new clients a year (or revenues of $20mill/year – first year revenues only).

Oracle Database, SAS (Software)

6/1996 – 12/2002

Documented the analytical process used for risk and later for marketing. This effort resulted in a 400-page document that outlines step-by-step instructions of the analytical process used in data mining. With the completion of the documentation, I was contracted to program the automation of the analytical process as much as possible. The SAS, SAS Enterprise Miner, and NeuralWare based, automated modeling environment reduced the model build time from 2 weeks to 2 days. This, and previous work at UBS/SBC, received the European SAS enterprise computing award for 2000. The addition of direct marketing to the retail client base at UBS turned the money-losing retail sector into a profitable business for UBS. The work has been the analytical backbone for UBS to date.
I was also contracted to plan, design, and manage the data feeds from multiple legacy systems to build the banks central analytical data mart, which supports risk and marketing at SBC (now UBS) for its retail and private banking units. The system is an analytical view of a 3-year transactional data history. The analytical work performed for UBS included risk and marketing models using advanced statistics, neural networks, and genetic algorithms. Additional consultation included the analytical department design, tools creation, tools evaluation, hardware sizing to support analytics, and other marketing issues such as how to analyze Internet data, performing ‘flows of funds’ analysis, and product portfolio optimizations. At UBS, I managed the entire analytical process from raw data to backend analysis to ensure that marketing efforts were successful. This work won the SAS Application of the year in 2000, and also became the foundation of SAS multiple SAS solutions now offered world wide

SAS (Software), C++

1/1994 – 2/1996

Built statistical and neural network models for American Express Cards Division and Customer Services using SAS. The analytical work included cluster and CHAID analysis, binary response, two-stage, and attrition models to support the personal (green, gold, platinum) and corporate cards. Other responsibilities included written documentation, and client consultation on analytical techniques, sampling issues, and analytical strategy.

SAS (Software), Deeplearning4j

8/1993 – 6/1995

Predicted call volumes for a 500+ people call center 6 months in advance to allow for proper staffing. The work utilized time series methods (ARIMA, Spectral) to predict daily call volumes and call distributions - 95% of the forecasts where within 3% of the actual values. Other work included applied queuing theory to prioritize workloads throughout the day, and predicting supply, demand, and utilization for inventory management using ARIMA. The improvements to the forecasts using ARIMA resulted in a 5% increase in inventory utilization (a $10million improvement).

Spss