Techniques such as Generalized Linear Models (Regression being one of them) help marketers attribute sales to the various marketing inputs and environment. They help in isolating the inputs that work and those that don't. It is the essential first step to estimating Return on Investments.

It is a methodology where the impact of inputs are measured simultaneously across multiple success criteria; sales, market share, awareness, intention to purchase, etc.

A methodology that can incorporate a marketer's domain knowledge into the modeling process instead of depending only on the dataset at hand. Very useful in situations where there is sparse data. Effective in situations where the impact of a marketing input has to be broken into customer segments, product lines, different markets or brands.

The process of allocating scarce marketing resources, under several business constraints with the objective of maximizing sales or minimizing costs. Techniques employed here are Linear & Non Linear programing.

A classification technique that groups customers, markets or media channels into homogenous groups, based on some pre-agreed criteria. Customers could be grouped into buckets based on value or behavior, markets could be grouped together based on potential and marketing channels could be grouped together based on impact.

A popular predictive analytics technique that predicts the happening of an event. Logistic Regression is used to identify the customers who will purchase again for cross sell opportunities, loan borrowers who will default, insurance policies that may lapse, customers who may churn, etc..

Again a predictive model that predicts an event but incorporates an element of time into its prediction. Useful in creating a "when to sell what to whom" model for cross sell.

Association Analysis - A methodology that provides us a set of rules, in simple language, about products that are purchased together by consumers. The rules could be as mundane at "Bread and butter are bought together" or as magical as "Customers who buy premium shirts have a small probability of buying a Barbie doll".

Neural Network - To detect non-linear relationships between large number of variables and solve attribution,forecasting and prediction problems.

Random Forest - In addition to solving prediction and attribution problems it can also solve classification problems.