Business / Performance Forecasting
- built and implemented a predictive model for holiday sales
- built a cannibalization model for new B&M store locations
- designed a repository of all advertising campaigns and developed metrics to score the effectiveness of a campaign and the impact of subsequent targeting of the same prospective customer
- built a threat-based business model framework (a generalization of the Factor Analysis of Information Risk methodology, FAIR) to evaluate the mitigating measures to various threats by several different business domains
Design and evaluate experiments
- developed a test-and-control framework and a sampling design for ad campaigns
- developed, implemented, and delivered a design of experiments to test the effectiveness of digital and physical security services in stores
Identify new levers to help move key metrics
- introduced trend analysis to a business domain whose model was based on aggregated data, masking sensitive changes in the data
- built and implemented a similarity rank model for item substitutes (Coke/Pepsi) for the merchandising domain of large enterprise
- built a model for item deletion based on sales, seasonality, similarity/co-occurrence, and defined item stability based on the support in the network model including affiliation with best items (pagerank-like), shared customer profile, profit impact, and geo-location. Lowest ranked items with minimal impact can be removed from the catalogue at the right time (seasonal stability) to avoid adverse secondary impacts
Monitor key product metrics
- developed a dashboard to monitor changes in the customer segments according to consecutive visits during holiday discount holidays (Black Friday, Pre-Thanksgiving, Cyber-Monday, etc.)
Build and analyze dashboards and reports
- built and designed business performance reports on production pipelines, business KPIs, general BI in various business domains
- influence product teams through presentation of data-based recommendations based on key behavioral indicators, oversaw the design of KPI dashboards for services adoption and customer behavior
Build models of user behaviors for analysis or to power production systems
- designed and implemented anti-fraud models for omni-channel services to detect organized criminal behavior
- built loyalty model for a large loyalty-based retail enterprise
Hard skills
- produce thousands of lines of production-level code per year
- engineer solutions from idea to production in a lean and effective way to enable business stakeholders to "test-drive" the prototype solution before launching a production process at scale, ensuring the result is precisely what the end-user wants and needs
Evaluate and define metrics
- built a customer-segmentation model for growing the customer base, including geo-demographic metrics such as saturation, showing areas of population where the existing customer base would not grow and business would need to target a diversified customer profile
- built and delivered a model for the effectiveness of physical security services in B&M stores, the model includes metrics such as incident prevention and safety rank, which is based on the percent of stores where the services provided by a given company have been effective
Understand root causes of changes in metrics
- a key anti-fraud metric in an analytics product delivered was the percent of customers who enrolled (downloaded and installed an app) and immediately purchased (activated). System alerts arrived (through an email and text message) of a spike in the daily enrollments but there was no spike in the activated accounts that would indicate a possible setup for a coordinated fraud effort for immediate attack. A preliminary investigation eliminated the high likelihood of a coordinated effort, and analysis was delivered to management leads who would then make a business decision. Later it was revealed that it was a system-generated activity
- based on an attrition model in production, developed a methodology for monitoring changes
Communicate state of business, experiment results, etc to product teams
- daily direct interactions with business and executives require a well-shaped and very diplomatic approach to conveying critical pieces of technology
Spread best practices to analytics and product teams
- frequently acting as a lead data scientist and mentor to junior data science and analytics teams, training and implementing best coding practices, Github/Confluence/JIRA integration of workflow, which enables an individual contributor role on selected key designs, as well as a manager and product owner role overseeing the workflow of several team
- leading subject-matter seminars and mentoring fellow data scientists in code reviews, workshops, setting high standards for the code reproducibility, documenting the algorithmic approach, optimizing the code and models, and delivering the results effectively, telling the story
Ability to hire and retain top talent
- interviewed data science and analytics candidates regularly for large enterprise tech organization
- mentored interns from Master’s and PhD programs to help scope new talent for hire
- influenced a data scientist to stay with the company by working on a challenging analytics project with direct exposure to the business users
Build key data sets to empower operational and exploratory analysis
- for multiple new services and acquisitions, designed and implemented data models based on customer-centric, device-centric, tender-centric, and model-based data sets