Similar to a test profile, we have the same phases for conducting a test with session preparation with the definition of the questionnaire, then data collect, and third, calculations with the possibility to ask a report generation.
The questionnaires are very open and composed of different types of questions linked together so that they can ask constitute a very flexible and friendly way to ask the panelists about their preferences.
We will resume for example the test carried out during the SIAL Exhibition to query visitors about fruit cocktails.
Other settings will define the product code, the presentation of these products, and a range of information needed to define the questionnaire.
This single definition will then be used to to constitute questionnaires fitted for our different types of data collection. These questionnaires may then appear in the input box's indoors, but may also, by that definition constitute Word forms that can be sent by mail attachments to PCs or mobile phones.
These preparations can also define paper questionnaires edited directly in Word©, which will then be scanned to automatically obtain responses from the panel.
The data collect is usually done by displaying the questionnaire to the panellists as shown in the picture below, but as we noted above, it can be done at home on a PC or laptop through Word© Forms via E-mails.
Once the entry complete, the data are stored in a database, and it is just needed to ask the calculation of these results, different parameters are then required with product selection, judges according to certain criteria to filter by example, and variables to process.
In the same way applied on profiles, options for handling missing data can also be applied if data are missing in these collected data.
Many statistical techniques can then be implemented and displayed in the results depending on the types of questions:
Visualizations of the results may represent graphs, with examples, the results in histogram, star, comparing products, pairs, all products, all these results and graphs can be resumed and rebuilt automatically in EXCEL©.
We may note also automatic transformations intensities in ranks, with tests on the rankings, or analysis on data distribution.
It is also suggested various tests such as analysis of variance and this analysis can be performed with two,
three or four factors, with all relevant interactions, analysis of normality are provided based khi2 test or
Kolmogorov Smirnov.
Different tests comparing means are also available: LSD, Newman and Keuls, Duncan. Different outputs display detailed
and synthetic results, and combinations of significant segments.
Multivariate analysis such as Principal Component Analysis can be applied with different available options:
additional variables or not, additional individuals or not. Three types of PCA can be done: products PCA (on product
means), panellists PCA (on panellist means), or detail PCA on the products x panelists values.
Graphs in three dimensions can be observed, together with details of values to help interpret the results.
Further multivariate analysis may also be cited as Correspondence Analysis on frequencies of the scores given by panelists or Ascending Hierarchical Classification of variables, products or consumers.
Tastel+ also provides capabilities for treatment preferences with specific example of the probabilistic analysis, this analysis allows, when protocols incomplete block, to recover from personal authorizations (blocks: "Panelist"), a collective order to the entire panel. These models are planned for pairs and trios. These calculations are based on the frequencies to be first, second or third.
In case of trios, two models are proposed:The techniques of internal preference mapping, also available in Tastel+ are designed to analyze consumer preferences individually to identify the varying attitudes of preference. This will thus represent the segmentation of consumers based on the appreciation of more than 5 products. The graphs obtained will see the names or the criteria of the panelists, but also counting functions will cut the space to learn, for densities of presence, the most interesting points to consider. Zoom possibilities, changes in observation areas, a grid locator, ... are offered.
Are also proposed different models External Preference Mapping, these statistical techniques are designed to highlight links between descriptive data from expert sensory and consumer preference.
These multidimensional techniques (joint PCA) can simultaneously represent groups of consumers according to their assessments of the products concerned with the ratings of experts on these products.
These graphs are displayed in Tastel+, but can be automatically issued in Excel format.
An ideal product modeling is also available, based on the maximum density of consumer preference. Then, verification of this modeling is performed by representing the maximum density and modeled the new product, the latter product is represented by point 'IDEAL PRODUCT'.
The results can then know what are the sensory characteristics required to best meet the preferences expressed by consumers.
These statistics provide different tools and provide decision rules on sensory enhancements to make products for a better appreciation by the selected target.
Different tables counting frequency choices and arrangements with percentages are calculated, and graphs
representing the results can be viewed.
It is also possible to have statistical significance to identify if different frequencies are significant
or not.
Tastel+ provides a textual analysis which has for aim to count all words in a text, free or codified, and to
define synonyms where appropriate between these terms, then, possibly after having clarified the meaning of these
words Tastel+ can count those terms specified in their direction with frequency of appearance.
Synthetic representations of textual data can then be obtained using CFA, and if more conventional frequency
histograms.
This analysis of the texts given by the panelists (experts or consumers) can most often offer a more qualitative description on products.
All these results can be formatted and interpreted automatically by requesting a report generation, a report
consists of a merger between a predefined and user text which can be changed, and results from calculations.
The user specifies the results he need to issue, knowing that this application may be a model that can be applied
to any further calculations:
The results are then always include orange juice, leaving the results in report generation can take place directly after the data entry, and does little more than one minute.
One can observe that the system automatically detects the type of statistic (parametric or nonparametric) to implement for the comparison of product synthesis, it is also possible the system automatically shows an interpretation of the results of the CPA.
All these results automatically formatted in Excel© format can then be picked up and handled for specific and personnalized reports and to under different media.