DATA SCIENCE AND PREDICTIVE MODELING
My professional roots are in predictive modeling, data science, and machine learning and I have developed and deployed (or "productionized") many predictive models and forecasts using behavioral and econometric statistical approaches. I started with customer preference and timing models for direct mail marketing campaigns. Later I added website valuation, digital demand forecasting, customer segmentation and profiling, media investment modeling, and structural consumer preference analysis.
These types of analytics are typically more time and resource consuming to develop and are frequently outsourced for development or programming assistance. A good deal of care should be taken in planning and execution of these projects to ensure they work as intended and provide the maximum value and "interoperability" with other models and analytic efforts.
DIGITAL DEMAND
WEBSITE TRAFFIC INDICATES CUSTOMER INTEREST
Website traffic volumes and activities can be interpreted as demand (as can natural search volumes), driven by marketing and advertising, correlated to sales, and moderated by conversion efforts (human or website). Using weekly website metrics and sales with bundled time-series and econometric methods I got extended forecasts of demand headwinds and tailwinds. (Natural search trends also correlate to sales over time.)
MEDIA MIX MODELING
HOW MUCH? WHERE? WHEN?
Media mix models provide a baseline and planning tool for developing campaign media recommendations. There are many forms but the base inputs are media spend and impressions while what is being predicted is total sales or another KPI. Once the best model is developed an optimizer program is applied to provide "hit target" (minimum spend to achieve sales target) and "what if?" (scenario planning) capabilities, given specified budget constraints.
WEBSITE VALUATION
YES YOU CAN MONETIZE YOUR WEBSITE!
Behavioral modeling of website "events" (a/k/a customer actions) related to sales and other KPIs can provide a useful indicator of near-term or future sales, add information about how to interact with different visitors, and provide a better metric for comparing media impact and value. This model used log data, including partner site content, with an external sales match to look at all visitor behaviors within a three month window across devices and visits.
Glad to discuss these and other predictive and descriptive modeling projects!