Cannon Gray 

Analytics is not software.  Software is not analytics.  Software and analytics do not make management decisions.  They are tools that help managers make decisions.

Many analytic tools are now available in user friendly software packages. User friendly doesn’t mean easy to use skillfully, however.  That requires formal education in research methods and statistics and a lot of experience applying this classroom learning to real-world problems.

Of course technical proficiency by itself will not make one a competent marketing scientist.  Marketing science is social science.  It calls upon diverse skills, such as being able to partner with marketers to help them see the big picture and anticipate key decisions they'll have to make. 

Let's briefly sum up our philosophy.  The purpose of marketing science is to enhance decisions.  We first need to make sure we're tackling the right marketing issues.  Research design, data and analytic methods must be aligned with these concerns.  Finally, analysis requires thought, imagination and judgment as well as attention to detail.  It is not a series of assembly line tasks. 

Marketing science is a way of thinking.

To help illustrate our philosophy, below is a short overview of some types of research that have proven especially useful to clients.  Actual examples are given on the Case Studies page of this website.

Consumer Segmentation

We are each different in our own ways but also have more in common with some people than with others.  Similarly, as consumers, we are not all alike but none of us is entirely unique either.  Our differences and similarities tend to fall into patterns that are related to our potential profitability as customers for a brand or service.  Consumer Segmentation can take advantage of this.

In Consumer Segmentation, Cluster Analysis is typically used to group survey respondents into attitudinal segments which are then cross tabulated against demographics, consumption habits and other relevant marketing variables.  A newer and better way is full-profile segmentation, in which attitudes and other important marketing variables are clustered simultaneously.  Recent advances in statistics have facilitated this.  Though the modeling can be tricky, full-profile segmentation is sounder statistically and yields sharper, more actionable segments.

Key Driver Analysis

A skin care brand manager will want to know what is most important to shoppers when they are choosing a skin care brand.  This information can be used for communication purposes and new product development.  While we can ask consumers directly what is most important to them, more often than not the results will not be very discriminating - they will tell us everything is important. 

Alternatively, Key Driver Analysis can be used to derive importance indirectly through statistical analysis.  There are various ways to perform Key Driver Analysis.  Latent variable methods adopted from Psychometrics are often useful.  Sophisticated versions of these models able to uncover segments with different priorities have recently been developed.  Segmented driver analysis is complex but often preferable to running many models on pre-defined (and perhaps arbitrary) respondent subgroups. 

Market Response Modeling

Companies need to know how their marketing activity (e.g., advertising, in-store promotions, price…) connects with actual movements in sales or market share.  Market Response Modeling is a statistical way to accomplish this.  It can also be used to forecast sales and share (Demand Forecasting). 

Although standard regression analysis is sometimes employed for these kinds of models it is not the best choice.  The type of regression we normally use in marketing research is intended for cross-sectional data, not data that have been collected across time.  Specialized methods borrowed from Econometrics are more appropriate.

Data Mining and Predictive Analytics

Medical Researchers want to know what factors are associated with illness and health and marketers want to identify risks and opportunities.  Following methodologies pioneered in Biostatistics and other fields, Cannon Gray uses Data Mining and Predictive Analytics for Customer Relationship Management (CRM) and other marketing purposes. 

Data from internal company records, secondary sources and consumer surveys can be integrated and analyzed.  Frequently a very large amount of data is modeled, a consequence of which is a high risk of "fluke" results.  This of course can lead to incorrect decisions.  Competence in this area requires expertise in statistical science and a broad assortment of analytic techniques in addition to a solid understanding of the client's business.

AI and ML R&D

Some questions users of AI and ML and software developers frequently ask are:

How well does the AI or machine learner work?

How does it compare against competing methods, including established statistical techniques?

How does it actually work? In other words, can we open the black box?