New Statistical Applications in Research
Among the multitude of tools and methods available to researchers, statistics have long been a mainstay of the marketing research industry. While statistical methods have not changed over time the applications have changed. Some of the new and emerging uses of statistical methods include:
- Predictive analytics using structured big data
- Pattern recognition within unstructured big data
- Monte Carlo methodologies
- Segmentation
- Structural Equation Modeling (SEM)
- Value based pricing
PREDICTIVE ANALYTICS
Due to the explosion of “big data” one of the most frequently used methods is predictive analytics for analysis of structured big data. Corporations collect and warehouse structured data for analytic purposes. Predictive analytics include a number of statistical techniques used to make predictions for the future based on current and historical data. For example, Amazon predicts new customer product interests based on past behavior. Similarly, a mortgage company may store a list of good loan candidates or an insurance company may list possible fraud sources. As Eric Siegel of Predictive Analytics World states, Predictive Analytics holds “the power to predict who will click, buy, lie, or die.” Predictive analytics are very useful in predicting future markets and customer behaviors.
PATTERN RECOGNITION
Pattern recognition is being increasingly utilized, especially when deciphering unstructured big data. Unstructured big data is data that does not fit in pre-defined categories and is usually text heavy. For example, social networks such as Facebook, Twitter, LinkedIn, and Flickr are all a source of unstructured data. The sharing of comments and information between participants is free-flowing. These sources of unstructured data are quite valuable and companies such as Nike and Starbucks have responded by creating control rooms where they track unstructured data and extract meaning from it. This allows companies to monitor, or if possible, control the brand conversations. The statistical engine behind analysis of unstructured big data is pattern recognition which include a number of techniques used to find patterns in the text. Understanding the how, where, why, what and when of unstructured data provides companies with a wealth of actionable data.
MONTE CARLO SIMULATIONS
Monte Carlo simulations are used to predict probability distributions. Applications of Monte Carlo simulations include return on investment (ROI), loyalty, product development, customer satisfaction, and regression modeling. Another application of Monte Carlo simulations is market size and potential analysis. Market size and potential analysis can identify market opportunities and direct where to invest resources for the greatest return. This tool is not solely for enterprises entering a new market or new product development, is also valuable for determining new areas for potential expansion for existing brands.
The inputs to the simulations are typically derived from primary survey data and data from the end client. The power of Monte Carlo is the ability to run millions of simulations using different scenarios for each simulation.
SEGMENTATION
Segmentation involves separating a population (e.g. the population of the US or an entire customer base) into segments which have common needs and priorities. This process requires statistical methods to divide the audience into like segments depending on how they have answered a strategically developed questionnaire. Once the segments are developed, these segments can be applied to an entire customer database and the creation of targeted marketing campaigns can proceed. Sophisticated tools such as factor and cluster analysis are frequently used in segmentation.
STRUCTURAL EQUATION MODELING
Structural equation modeling (SEM) is a technique in which several statistical methods are used to test a conceptual or theoretical model. SEM allows researchers to hypothesize and test models of market behavior. These models allow researchers to visualize what variables cause changes in other variables. SEM also takes into account latent variables, that is, variables that are not observed but can relate to observed variables. SEM is often used to develop a causal model of satisfaction or loyalty.
VALUE BASED PRICING
Value based pricing is a pricing strategy where the price of a product or service is based on the perceived or estimated value to the customer rather than on the cost of the product, competitor prices, or historical prices. Once an appropriate survey has been designed, there are various statistical methods that can be employed to determine the optimal price depending on the value assigned to the product by the consumer. A number of statistical techniques may be employed to identify the optimal price for a product or service.
A NEW WORLD OF MULTIVARIATE STATISTICAL APPLICATIONS
Statistical methodologies in market research no longer need to consist solely of finding statistically significant differences, means, ranges, and standard deviations. While these are all still valuable tools, they are often best used in conjunction with relatively new applications of statistics.
Xavier Alvarez is a Project Manager at Q2 Insights, Inc., a research and innovation consulting firm with offices in San Diego and New Orleans. He can be reached at (760) 230-2950 ext. 4 or xavier.alvarez@q2insights.com.
This entry was posted in Tools and Techniques and tagged on December 14, 2016 by Kirsty Nunez