In addition to a large variety of internationally validated methods, TNS INFO Research also has a lot of experience with the application of the following data analysis techniques:

• Regression analysis
• Analysis of variance
• Discriminante analysis
• Cluster analysis
• Factor analysis
• Multidimensional scaling (MDS)
• Correspondence analysis

Regression analysis
By means of regression analysis, the effect of one or more predictor variables (e.g. price, advertising expenditure) on a dependent variable (e.g. sales, market share) can be estimated. The main task is to find a (linear or non-linear) relationship between the variables which best reflects the underlying dependencies and can be used for forecasting. To do this, the least squares method is frequently applied.

Analysis of variance (ANOVA)
This technique is used to measure the effect of one ore more independent variables (e.g. packaging, placement in shelf, advertising sujet) on one or more dependent variables (e.g. sales). The main advantage compared to Regression analysis is that the independent variables can be measured on a nominal scale. Therefore, ANOVA is the most important statistical technique for analysing experimental research designs. By means of variance decomposition it can be determined whether the attribute levels of independent variables have a likely impact on the level(s) of the dependent variable(s).
In market research it is often necessary to take into consideration the impact of an external factor of influence (e.g. attitudes) on the dependent variable(s). In case such „Covariates“ are included in the analysis, ANOVA turns into ANCOVA (Analysis of covariance).

Discriminant analysis
The impact of one or more independent variables (measured on an interval scale, e.g. income, age) on a dependent variable (measured on a nominal scale, e.g. market segments) is expressed by a Discriminant Function. This function provides a most selective differentiation of attribute combinations in terms of the dependent variable. This enables the researcher to examine the impact of attributes on the group membership. More importantly, the function can be applied also to new elements (e.g. bank customers with certain attributes, asking for a credit), allocating them to pre-defined groups (e.g. high vs. low risk). 

Cluster analysis
Cluster analysis is used for grouping elements in different ways. The main task is to create groups consisting of elements with very similar attributes (e.g. age, gender, income, residence) while the created groups (e.g. market segments) should be as differentiated as possible.  Based on the outcomes of this often used research technique in the field of market segmentation, clients can decide whether it is reasonable to serve a particular segment in a different way.

Factor analysis
The principle of Factor analysis is to reduce a large number of directly measured attributes (e.g. crispy, healthy, aromatic, low-calorie, delicious) to a smaller number of latent „Factors“ (e.g. enjoyment, healthiness) which cannot be observed. The resulting factors are extracted to make sure they can be used to explain the attribute variables as accurately as possible. As the number of factors gets smaller than the number of initial variables (reduction of complexity), interpretations and comparisons are made easier.

Multidimensional Scaling (MDS)
MDS is used to determine the relative positions of objects (e.g. brands, products, politicians) in a persons’ (e.g. consumer, electors) mind. Based on global ratings (e.g. paired comparison, scale), objects are related to each other in a way that similar objects are more closely together than different ones. It´s important to mention that the relevant attribute dimensions need not be determined beforehand but will be defined in the course of analysis. This helps avoiding exclusion of important dimensions from the analysis, while less relevant ones might be included.

Correspondence analysis
Correspondence analysis is a powerful tool for analysing qualitative data (i.e. data measured on a nominal scale). The main purpose is to examine dependencies between attributes of a single object (Multidimensional Contingent Analysis). By means of reducing a large number of attribute dimensions (projection) to e.g. two or three, dependencies can be visualised graphically and interpreted more easily.