All posts by florisbienfait

Disconfirmation Effect: How do our product opinions change over time?

Firms of all sizes are known to spend large portions of their budget on marketing campaigns through which to attract consumers. It has been found that 20-50% of all purchasing decisions are made primarily as a result of word-of-mouth (WOM) (Bughin et al., 2010). Online WOM (eWOM) offers consumers and companies certain advantages; eWOM increases persuasion and willingness to pay (WTP) in consumers, while offering consumers informedness and a sense of trust in a company’s products or services (Healey, 2016). Companies are able to improve products through an enhanced understanding of the consumer, while profiting from free/cost-effective advertisement (Buttle, 1998).

Disconfirmation Effect

In their article, Ho et al. (2017) introduce a new take on studying rating/reviews and purchasing decisions in the digital marketplace, which is the concept of disconfirmation, (formally expectation-disconfirmation theory, or EDT for short), and is defined as ‘the discrepancy between the expected and experienced assessment of the same product’ (Ho et al., 2017; Anderson and Sullivan, 1993). The paper aims to study the effect disconfirmation (the difference in prepurchase expectation and postpurchase experience) on the behaviour of consumers leaving online product reviews. A conceptual framework outlining the author’s views on prepurchase and postpurchase influences can be seen in the figure below:

Screen Shot 2018-03-11 at 20.16.18


Ho et al. (2017) made use of a Bayesian learning framework to study consumer rating behaviour from a dynamic perspective (Wang and Yeung, 2016). In this framework, a shift in consumer opinion over time is incorporated into the analysis of the data. While previous works have focused on static points of view, this paper effectively studies rating effects over a longer period. Data was provided by ‘an online e-commerce website similar to Amazon’. Purchasing information came from 10 major product categories, and was very specific (Ho et al., 2017. Specific data was gathered between March 2006 and November 2011 and included:

  1. Order information (product name, price, shipping and handling time, order placement date): useful for finding out if a user is posting a review or only lurking
  2. Customer-reported reviews (with review title, review body, submission date, overall rating on 5-star scale): Useful for recovering rating information (valence, volume, variance) available to consumer upon purchase (Ho et al., 2017; Flanagin and Metzger, 2013).


The most essential finding applicable to the goal of the study are the fact that a consumer is more likely leave a review when the disconfirmation is more severe; the sign of the rating (positive or negative) depends on the sign of disconfirmation (see Figure 2). Key moderating factors are that the disconfirmation effect is attenuated by an increased period of time between purchase and receipt, and that dissension in evaluations among peers accentuates the disconfirmation effect (Ho et al., 2017).

Screen Shot 2018-03-11 at 20.31.55

Managerial Implications

The authors outline some actionable strategies resulting from the aforementioned findings. Firstly, managers should ensure quick order fulfilment. Consumers contribute more positive reviews when products are received quicker, so more eWOM van be achieved through efficient order fulfilment. Secondly, managers should maintain a high level of service. Consumers are likely to leave negative reviews when service is poor, which could decrease purchase intent (Imrie, 2005).


This paper is strong for a number of reasons aside from the econometric robustness of the study, the long time period studied (approx. 5.5 years), and the extent of data gathered.

  1. The study’s dynamic perspective differentiates it from previous works studying EDT or online reviews. This allows the learning framework to develop as necessary, which is essential as perceptions change depending on the reviews posted.
  2. The authors collected data at the micro (individual) level (Clark and Avery, 2010). Most articles about EDT and post-consumption WOM use insights that have been aggregated (product-focused), while this paper studies cases from the individual consumer perspective. This results in stronger insights into consumer behaviour and the drivers of disconfirmation, possibly allowing for more actionable recommendations.
  3. In their paper, Ho et al. (2017) consider the fact that (potential) consumers may not trust the review system, and therefore hold a perception of the review system This consists of two components, namely system biasedness (relating to pre-purchase expectations of the product) and system preciseness (relating to the level of disconfirmation experienced after purchase). The authors also highlight that not all shoppers are the same. A distinction is made between frequent and occasional raters/reviewers. Occasional raters are observed to be more biased and susceptible to disconfirmation; frequent raters are more objective and should be relied on more by both future consumers and managers.


A number of weaknesses also exist in this paper:

  1. There is no clear distinction between written reviews and discrete ratings despite a short mention and usage of a Bayesian model. The authors should discuss the potential influence a written review may have on the disconfirmation effect relative to a rating on a 5-point scale.
  2. Online spending habits may have changed significantly between March 2006 and December 2011 (, 2018). Also, the extent of the popularity of the website used to gather data will likely have changed in the timeframe. The authors could have done a better job evaluating the pros and cons of such a timeframe, and better distinguished between consumers, rather than only labelling them as ‘occasional’ or ‘frequent’ raters.
  3. There is no mention of the influence of geographical location on the study. The authors wrote the paper at American and Chinese universities, which could be indicative of vast cultural differences in purchasing or reviewing behaviour and may also have influenced the handling and shipping times observed. This warrants an extensive review of where reviewers came from and the generalisability of the study, including outside of an Amazon-like business.

Works Cited

Anderson, E. and Sullivan, M. (1993). The Antecedents and Consequences of Customer Satisfaction for Firms. Marketing Science, 12(2), pp.125-143.

Bughin, J., Doogan, J. and Vetvik, O. (2010). A new way to measure word-of-mouth marketing. [online] McKinsey & Company. Available at: [Accessed 11 Mar. 2018].

Buttle, F. (1998). Word of mouth: understanding and managing referral marketing. Journal of Strategic Marketing, 6(3), pp.241-254.

Clark, W. and Avery, K. (2010). The Effects of Data Aggregation in Statistical Analysis. Geographical Analysis, 8(4), pp.428-438.

Flanagin, A. and Metzger, M. (2013). Trusting expert- versus user-generated ratings online: The role of information volume, valence, and consumer characteristics. Computers in Human Behavior, [online] 29(4), pp.1626-1634. Available at: [Accessed 11 Mar. 2018].

Healey, N. (2016). Electronic Word of Mouth (eWOM) – How can we can break it down?. [online] University of Brighton. Available at: [Accessed 11 Mar. 2018].

Ho, Y., Wu, J. and Tan, Y. (2017). Disconfirmation Effect on Online Rating Behavior: A Structural Model. Information Systems Research, [online] 28(3), pp.626-642. Available at: [Accessed 11 Mar. 2018].

Imrie, B. (2005). Beyond disconfirmation: The role of generosity and surprise. International Marketing Review, [online] 22(3), pp.369-383. Available at: [Accessed 11 Mar. 2018].

Wang, H. and Yeung, D. (2016). Towards Bayesian Deep Learning: A Framework and Some Existing Methods. IEEE Transactions on Knowledge and Data Engineering, 28(12), pp.3395-3408. (2018). E-commerce Worldwide. [online] Available at: [Accessed 11 Mar. 2018].

FlavorPrint: Personalizing your recipes through your tastes

How amazing would it be if you knew every meal you cooked would fit your tastes? McCormick & Company, a major player in the flavor industry, is reinventing traditional FMCG business models through its data-driven, customer-focused offerings. While the company generally manufactures and distributes spices, seasonings, and other products over 125 countries and territories (Amazon Web Services, n.d.), a shift has occurred from a product-centered company to a business model in which the entire customer value is achieved through a comprehensive consumer journey.

McCormick is continually moving towards innovative solutions to reach customers relative to competitors or FMCG companies in other sectors. The expected sales target of $5bn by the end of 2019 will come from e-commerce, innovation through platforms, and acquisitions of other companies (Nunes, 2017); evidently, digitization is driving the company’s growth. In 2014, McCormick created a spinoff company named Vivanda, through which a transformative product called FlavorPrint was developed (Nash, 2015).


FlavorPrint is ‘a technology that matches people with food they love’ (FlavorPrint, 2017). When users sign up to McCormick’s recipe platform, they are asked to fill out initial questions about their food preferences. Their recipe search behavior on the platform will continuously adapt the user’s ideal taste palate to recommend recipes that fit the user perfectly. FlavorPrint ‘combines sensory science and culinary science’ to ‘offer personalized recommendations for recipes, meals, and eventually wine pairings’ (Amazon Web Services, n.d.). FlavorPrint is able to change a person’s cooking habits by offering exciting alternatives that are customized to the user (while promoting McCormick’s products) (FlavorPrint, 2017).

Value to Consumers

Vivanda’s FlavorPrint follows a number of mass customization (MC) drivers while requiring little to no investment by the consumer, and consumers participate in the service because it offers them significant product utility. The extra costs for consumers are low; the quality of recommendations is high, no financial investment is necessary to use the service, and the effort of signing up to the platform is relatively low (Tsekouras, 2018). Furthermore, the FlavorPrint service works automatically, meaning that the consumer does not have to take any specific action to use the service, other than signing up to the platform. In short, FlavorPrint’s predictive analytics technology has made recipe selection much easier and more likeable, while demanding little time and effort from consumers.

Efficiency Criteria and the Future of Predictive Analytics in Food

In 2013, McCormick initiated a small beta program for its new technology. While a 1% increase in sales is very large in the industry, FlavorPrint quickly grew to 100,000 participants (while still in beta mode) and drove sales up by 4.9% (Amazon Web Services, n.d.). This was a sign that the company needed to ensure scalability for its platform, to allow millions of users to participate.

While financial data and statistics regarding platform usage have not been published, Vivanda has officially spun off from McCormick. In 2016, Vivanda announced a strategic partnership with and investment from German software giant SAP. This collaboration will ‘help our food industry partners to grow profitably by delivering increasingly personalized experiences and outcomes directly to customers’, according to E.J. Kenney, SVP Consumer Products Industry at SAP (SAP, 2016). The partnership indicates that Vivanda has shifted its strategy from focusing on McCormick customers to delivering its service to various players in the food and beverage industry; by targeting a wide range of food and beverage customers, Vivanda’s growth seems inevitable.


It will be interesting to see what the future will hold for Vivanda and the use of predictive analytics in food. McCormick evidently derives great value from the technology, but one has to wonder if the technology has its criticisms pertaining to a possible lack of understanding of consumer behavior or privacy issues. For example, while the technology takes into account various contextual factors such as consumer budget and nutritional objectives while recommending foods, changing lifestyle situations may prove it difficult for the technology to adapt fully to consumer’s lives.


Although FlavorPrint does not directly offer a new revenue stream, the new possibilities for consumer packaged goods firms to reach customers indicate a potential for significant impact on future sales for Vivanda clients. Customization/personalization lies at the heart of the service, which is why the business model provides companies with a way to target consumers much more directly than through traditional marketing.

Will you use FlavorPrint to find new recipes? Does the company have a bright future? Let me know in the comments!



Amazon Web Services. (n.d.). AWS Case Study: McCormick. [online] Available at: [Accessed 18 Feb. 2018].

FlavorPrint. (2017). FlavorPrint. [online] Available at: [Accessed 18 Feb. 2018].

Nash, K. (2015). Tech Spin-off from Spice Maker McCormick Puts CIO in the CEO Seat. [online] WSJ. Available at: [Accessed 18 Feb. 2018].

Nunes, K. (2017). Innovation central to McCormick’s growth strategy. [online] Food Business News. Available at:{CD115D1F-0E2B-4AE5-8295-8ED5DD8C1516}&page=1 [Accessed 18 Feb. 2018].

SAP. (2016). SAP and Vivanda Serve Up FlavorPrint Technology. [online] Available at: [Accessed 18 Feb. 2018].

Tsekouras, D. (2018). CCDC Lecture 3.