Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea

Transcription

1 Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea Ensuring the Business Relevance of a Company s Internal Sensory Program Presented by Dr. Benoît Rousseau The Institute for Perception, Richmond, USA Benoit.Rousseau@IFPress.com June 28, / 19 Ensuring the Business Relevance of a Company s Internal Sensory Program Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea Ensuring the Business Relevance of a Company s Internal Sensory Program Background 2 / 19 1

2 Methods of Sensory and Consumer Research Can you tell the difference? Triangle What is the difference? Is the difference important? How sweet? How bitter? How fruity? How astringent? P1 P2 Which sample do you prefer? Which sample is different from the other two? 3 / 19 Why Use Discrimination Testing? Measure the size of the difference between products Two main objectives - Prove products are different New and improved, Fresher, crisper taste - Prove products are similar Ingredient modification, cost reduction, change of supplier, government regulation (e.g., lowering of salt or sugar content) 4 / 19 2

3 Saltiness 6/29/2018 Sensory Science and Quality Insurance Upper limit Target Lower limit Time 5 / 19 Studying Product Similarities Product formulation change Preference test Internal Sensory Panel Question Will the consumer perceive a difference and potentially reject this new formulation? - Lower $$ - Longer shelf life - Faster process - Better ingredients - New product not released If significant sensory difference If no significant sensory difference New product good candidate for release 6 / 19 3

4 Effect of Experimental Sample Size Statistical hypothesis testing is futile because with enough replications, the null hypothesis will always be rejected A B A Triangle test Victor Chew (1977) N=20 9 correct (45% correct) 27 correct (45% correct) N=60 Binomial not significant (p = 0.19) Binomial significant! (p = 0.04) 7 / 19 Apparent Discrepancy in Discrimination Testing Results Case Illustration A Preference, but No Significant Difference 8 / 19 4

5 Ishii, O Mahony, Rousseau (2014) A Preference, but No Significant Difference Comparison of apple juices of different concentrations A Results B A B A B - vs. A 100% B 95% Two experimental protocols - Discrimination: Tetrad (N=228) A B A B - Hedonic: Paired preference (N=104) A B Discrimination # tests correct in tetrad test (out of 228) 84 p > 0.05 No diff. N 5% = 89 / Preference # choices for A in preference test (out of 104) 69 p < 0.01 Pref. N 5% = 63 / / 19 The Need for Information on Consumer Relevance Essential fact: - When comparing two products for similarity - Assuming that the sample size is large enough - A statistically significant result will always be found What is the optimal sample size? 12? 100? 20? 1,000? An optimal sample size can only be set if the size of the relevant difference is known (δ R ) 10 / 19 5

6 Ensuring the Business Relevance of a Company s Internal Sensory Program Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea Ensuring the Business Relevance of a Company s Internal Sensory Program How to Estimate and Use Consumer Relevance 11 / 19 Type I error Size of the difference Type II error (Power = 1-β) Sample size 12 / 19 6

7 % choices preferred product 6/29/2018 Measuring the Size of the Sensory Difference Thurstonian Modeling d = Standardized measure of sensory difference d = Experimental estimate of d X d Y In sensory testing, d typically varies between 0 and 2 13 / 19 Option 1 Estimating d R from Consumer Preference Tests 70% 65% 60% 55% 50% 45% 40% d' The relevant threshold can be set at d R =1.1 Using this value, the program s risk profile can be established 2-AFC test α=5% Power=90% d R =1.1 N=28 14 / 19 7

8 Option 2 - Estimating d R from using the Same-Different Method Are the two apples the same or different? 1.2 corresponds to the consumer threshold for difference Different A Same t B Different The program s risk profile can then be established Tetrad test α=5% Power=80% d R = 1.2 N=39 Different 15 / 19 Power 5 linked components: : Probability of a Type I error b : Probability of a Type II error (1-power) δ R : Size of the differenceof interest N : Sample size 90% 80% Protocol 23% 65 5% α Power δ R N Protocol Triangle Tetrad 2-AFC 16 / 19 8

9 Ensuring the Business Relevance of a Company s Internal Sensory Program Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea Ensuring the Business Relevance of a Company s Internal Sensory Program Conclusions 17 / 19 Conclusions Sensory evaluation and consumer science: Invaluable tools to the food and beverage scientist Need to establish the threshold above which a sensory difference is meaningful to the consumer Internal sensory program: Beneficial to companies to increase testing speed and decrease costs However: Internal findings must be predictive of consumer perception and opinions Theories available to quantify the size of sensory difference and estimate consumer relevance A sensory program can then be designed to optimize its relevance and maximize the likelihood of success of products released in the market place 18 / 19 9

10 June 28, 2018 Sensory and Consumer Science Division 2018 KoSFoST International Symposium and Annual Meeting Busan, Korea Ensuring the Business Relevance of a Company s Internal Sensory Program Presented by Dr. Benoît Rousseau The Institute for Perception, Richmond, USA Benoit.Rousseau@IFPress.com Thank You 19 / 19 10