Statistical Answers to Common Business Questions
Examples in Marketing, Pharmaceuticals, and Finance © 2012 Ceres Analytics

Business Question	Examples of Business Action	Statistical Answer*	Illustration
I tried something different with two groups of customers (or patients, or portfolios)—did it work? How do I know if a small group I selected is just like everybody else?	Compare an experimental group with a control group (drug administered vs. not, investment portfolio vs. benchmark) Measure the effect of a single marketing campagin or a group of campaigns (before vs. after) See if a small study sample represents the underlying population	T-test Komolgorov-Smirnov test
How can I construct segments of customers (or patients or portfolios) with distinctly different profiles?	Form customer groups that you can label Asses relative size and characteristics of opposing patient segments	Cluster Analysis Latent Class Analysis     (for surveys)
How do I predict a metric (like sales, weight loss or market return) from a variety of data?	Find out what drives sales level (big spending) Find what drives sales growth (acceleration of sales) Forecast sales level or growth See if you can really predict body weight from the amount of sugar and other carbohydrates people eat.	Regression Analysis
How do I predict a yes/no characteristic (like whether a stock pays a dividend, a customer adopts a new technology, or a prospect resonds to a credit card solicitation) from a variety of data?	Find out what drives the yes/no response to drug success (or a stock’s dividend payout or a customer’s choice of LCD TV) Predict probability of individual success or failure  (which stocks might pay dividends, which households will subscribe to broadband)	Logistic Regression Analysis
For all the data I have on borrowers (or students or members or customers), how do I find the interactions among data that determine the answer to a yes/no question? (or a multiple choice question, like always/never/sometimes)	From available data on club members, discover whether those most likely to quit either: - visited less than three times last year AND were late with their dues at least once AND never brought a guest OR - took a leave of absence greater than 3 months BUT less than 6 months	Decision Tree --CHAID or --Recursive Partitioning
For all the data I have on patients (or stocks or customers), what few, fundamental concepts define them?	From available metrics, determine key stock characteristics like: -  Growth -  Price Volatility -  Size Asses relative importance of individual metrics on the key characteristics (e.g., are income and unemployment all you need to know about a local economy, and does income count for 75%?)	Factor Analysis
How do I articulate a “bright line” between two groups (of customers, drug impacts or stocks) so that I know what makes them different—and how?	When some customers (patients, stocks) look great and some look downright awful… find out what makes them different. Determine relative importance of things that separate good customers from bad (like prices, advertising, their local economic conditions).  Then migrate “bad”  customers across the line to “good” by changing the most important thing that you can.	Linear Discriminant Analysis
* The techniques shown are intended as examples.  Different techniques may be suitable for specific business questions.