The methods of hypothesis testing I learned as a scientist are applicable to many situations, including testing the effectiveness of advertising. The “null hypothesis” – the assertion that nothing has changed – is central to the method. This has the effect of putting me (the testing expert) in a position that feels adversarial, to those who are trying to show an effect of their work. In NASPL Insights June 2014 I recount how my first encounter with a client exposed both my methods, and their feelings.
Category: Mathematical Models
applied data science
Mega Millions: What is the New Normal?
Using math modeling techniques including thousand-year models, I forecast the changes in behavior of the Mega Millions game to be expected from structural changes that took effect in October 2013. This article appeared in NASPL Insights February 2014.
Automating the Lottery Instant Game Supply Chain
In NASPL Insights January 2012, I report on the Business Rule Test (BuRT) project. A six-month project at WA Lottery showed that a fully automated system supported retailers as well or better than the inside sales/telephone sales that was standard at the time of the test. One key to the success of this project was to provide the sales representatives a way to provide information to the system based on their local knowledge. Another key was to harvest higher-level data from the gaming system to understand which games were trending in the marketplace. I posited a set of business rules, implemented them as code on a generic computing platform (VBA/ Excel!), and compared consumption of instant tickets from retailers who participated in the test, to a control group. The test retailers saw a sales increase.
Forecasting Instant Game Sales
In Washington State, the lottery shares its sales forecasts with state economists, who are accustomed to forecasting all sorts of revenue of the basis of economic and population variables. Taking their lead, I found that I could account for a long, quarter-by-quarter history of instant game consumption on the basis of population, cost of living, unemployment, and so on- without taking into account anything that the lottery was doing. With forecasts of these driving variables, it is possible to make quarter-by-quarter forecasts of instant game consumption that are as accurate as other revenue forecasts. I apologize for the tight academic style of this article; I was still transitioning from hard science! NASPL Insights October 2011