To get started on the advanced analytical journey, marketers should know what advanced analytics entails, how to think about it and, at a high level, what types of marketing problems have been successfully addressed with advanced analytics. Advanced analytics is not a fully defined field. A variety of terms—analytics, predictive analytics, predictive modeling, advanced marketing research, advanced modeling, data mining, etc.—may all be used to essentially mean the same thing. Whatever the term used, the practice is more likely to be referred to as advanced when it:

1. Considers data as a strategic component in the search for actionable insights. For example, when designing surveys, advanced data collection approaches are selected that may go beyond the standard questioning approach. Examples include discrete conjoint choice tasks and laddering tasks. Advanced survey design also can connect the results of surveys to other data sources.

2. Use more advanced statistical modeling or data mining tools. We define modeling as any analysis that specifies a model and has a predictive element to it. To dig deeper into the data and uncover new and better insights, an advanced approach would go beyond standard predictive modeling to include more realistic and rigorous modeling of consumer or market behavior. For example, a standard approach to analyzing the effectiveness of consumer promotions would use a standard time series regression model to determine if promotions are correlated with increased sales. Such a modeling approach would not capture some critical consumer behavior dynamics, such as:  

• A lead effect, which exists if consumers start to anticipate a promotion by buying more during a promotion than normal so as to bridge the time to the next promotion (stockpiling). 

• A threshold effect, in which a specific promotion has no impact unless it is at least a certain size, say 10%.

If these effects are not captured via a more sophisticated model, the guidance that the model provides may be flawed and marketers could end up taking incorrect actions. There is a wide variety of effects that can capture specific behavioral consumer dynamics, but the key point is that there is value in looking beyond the standard analysis. 

3. Is executed iteratively, sometimes executed over multiple stages, using several different data sources. Frequently, the initial pass-through of the data does not give the desired results. It is not uncommon to analyze one specific data set, with one set of variables, in many different ways before one discovers the insights. Sometimes there are various components to an advanced analysis plan that are executed as stand-alone analyses, but the results of different types of analyses are later integrated. For example, conjoint analysis can be integrated with brand perception data, concept testing can be integrated with forecasting, etc. Combining different data sources often can bring unique perspectives and uncover new insights. For example, we can combine attitudinal, financial and transactional data or we can combine micro and macro data. An example of the latter is when survey data is mixed with population and macroeconomic data. A common scenario is to use internal transaction data and match this with external survey data. In this way, we would pull a sample of customers and transactions, both internal data sources, and then fuse them with external satisfaction data. Next, driver models could be estimated based on the fused data set.

The foundational element of advanced analytics is that it aims to produce deeper and more valid insights into consumers, often aimed at understanding what factors are causing marketing success or what factors are predictive of marketing metrics.

This blog post was written by Marco Vriens and Patricia Kidd and originally appeared on the AMA IH website. Click here to read the entire article.