A product improvement is often necessary if ingredients are substituted or the production processes are changed. In this case it is often sufficient to maintain the same quality. Another situation where improvement of a product is needed is when a product is challenged by the competitive products on the market. The general measure of improvement is when consumers have a higher liking and preference for a product in consumer testing.

The improvement will often be quantified by a higher liking score from a consumer test. Consumer testing such as liking and preference tests will answer the question “Is your product liked better than the samples to which it is compared?” To make efficient experiments, for the purpose of finding the optimum taste, it is highly recommended to work systematically using an experimental design.

Gerner Hansen, Head of Sensory and Consumer Research Section at Danish Technological Institute, explains: “Manufacturing companies are often pressed for time to optimize a product and therefore haven’t got the time to plan and execute a systematic development strategy. But from an efficiency point of view, both time and money can be saved by working with experimental design where the variation of the sensory attributes are varied with the knowledge from pre-testing and structured by an experimental design. For insight in the importance of different sensory attributes the External Preference Mapping, PREFMAP, is a method used by many companies. In product development and innovation projects this method gives valuable insight, but requires expertise in statistical analysis of sensory and consumer data”.

However product optimization is a very broad issue and we will focus here on two situations:

1. Finding taste drivers using the PREFMAP method and
2. Optimizing a product by changing the recipe or process

Finding taste drivers using PREFMAP
To understand the importance of different sensory attributes for the over-all liking External Preference Mapping, PREFMAP can be used. Often the PREFMAP method is used to obtain insight of consumer’s liking in an early stage of product development. In literature, the statistics of the method has been discussed extensively since 1992, but more important many references show practical examples of the PREFMAP method. Briefly preference mapping can be described as a method, where a set of products are responded to, by a number of consumers. The respondents give their immediate liking score of a product, on a liking scale. Simultaneously the same products are described using sensory profiling method with a trained descriptive panel. The preference mapping is obtained by correlating the latent variables from each of the two data-sets. By using multiple regression analyses, like e.g. partial least squares regression or procustes analysis, both a statistical and more important a graphical presentation of the sensory drivers can interpreted. The respondents (their liking scores for all products) can be sub-divided into segments using a cluster analysis. Each segment will represent a certain pattern of liking of the products and the sub-set of respondents and their  data can be investigated. However, although the method is a rapid and effective way of getting insight to the sensory drivers of an existing product category, the method needs some expertise in making the statistical calculations and some experience in interpretation of the data. To make the interpretation of the results easier the set of products in the study should preferably have different sensory profiles.

Optimizing a product by changing the recipe or process
Product improvement, regarding sensory attributes, is relevant whenever the recipe or manufacturing process is changed. It may be a recipe or manufacturing change of an existing product or for optimization during product development. The optimization can be done systematically using appropriate experimental designs and descriptive sensory analyses or consumer testing, or the two methods in combination. The improvement will normally be measured as a higher liking score for the optimized product from consumer tests.

The first phase in optimization is defining the range of sensory variation to be tested. The range of variation for each of the factors should be within a range, so it is not extremely disliked and on the other hand the variation should be noticeable to have an effect in the test. It is recommended to do a pretest with a small number of respondents or colleagues to find out the appropriate range for the factors and to test if respondents will understand the questions. When the factors (recipe, process or other variables) are set the experimental plan can be made. Most often factorial designs are used, or reduced (fractional) factorial design.  

The experimental design, factorial design, can easily be made on paper if the number of factors is relatively small and there are no constraints to the levels of a factor. Calculations are equally simple and can be done with a calculator or in a spreadsheet. However, it will often happen that certain combinations of the factor levels are not possible or are not relevant, and then a more advanced experimental design may be necessary. There are a great number of software options for design of experiments and statistical calculations. Courses are available for design of experiments and experimental design for product optimization.

For further reading:
- Design and Analysis of Sensory Optimization
Maximo C. Gacula, Jr. ISBN: 978-0-470-38479-4. Wiley-Blackwell
- External preference mapping
McEwan JA. 1996. Preference mapping for product optimization. In: Naes T, Risvik E, editors. Multivariate analysis of data in sensory science. Amsterdam: Elsevier. p 71–80

For further information please contact:
Gerner Hansen
Danish Technological Institute