The drivers of the diffusion process of new
products within the fast moving consumer goods
industry
by
Sterre Peters
January 15, 2018
Master Thesis
The drivers of the diffusion process of new
products within the fast moving consumer goods
industry
January 15, 2018 Master ThesisMSc Marketing Intelligence and Marketing Management
University of Groningen Faculty of Economics and Business
Department of Marketing 9700 AV Groningen
First supervisor: dr. T. Bijmolt Second supervisor: dr. H. Risselada
MANAGEMENT SUMMARY
Lately, the number of new product introductions of fast moving consumer goods has increased a lot. The number of stock keeping units is higher compared to several years ago. Nowadays, competition has grown and consumers’ demands are more diverse. As a result, companies are trying to find new ways to achieve the goals of the consumers, for instance by investing in new product development. However, the number of products that do not meet their commercial objectives is also high. In other words, the product failure rate is large. As a consequence, managers are interested in studies about new product introductions and the process of how consumers adopt these products. Moreover, the past recession has led to an increase in private label share at the expense of national brands because consumers have become more price-sensitive. Studies argue that national brands can fight against these private labels by investing in new product launches. This study contributes to the question that rises from managers about the diffusion process of new product introductions. The diffusion process deals with the number of adopters of new products, relating to the first-purchase sales. Also, it focuses on the timing of adoption, so whether there are a lot of innovators that adopt products soon in the product lifecycle. Moreover, internal influences is a pillar of the diffusion process, that focuses on the imitators who are influenced internally, for instance by word-of-mouth communication.
This thesis investigates the drivers of the diffusion process of new products within the fast moving consumer goods industry. The data consists of 3,130 products that were introduced in the period between 2007-2015. The drivers that are studied are the product category (food or non-food), the degree of newness (new variant or new product), and the influence of the economic situation at the time of the product launch (contraction or expansion). Eleven hypotheses were derived from previous literature and research, which will help answering the research question. Diffusion models are commonly used in marketing, where the representative diffusion model of Bass forms the basis of this thesis. The number of adopters is parameter ! of the Bass model, the timing of adoption is parameter ! and the internal influence is parameter !. First, the Bass model parameters (!, ! and !) were estimated for all new product introductions using the ordinary least squared (OLS) method, followed by three linear regression models that show the differences between the product category, the degree of newness, and the business cycle, per dependent variable.
PREFACE
This Master Thesis is the final work of my double-track Master degree in Marketing Intelligence & Marketing Management at the University of Groningen. It presents the results of my research regarding the diffusion process of new products within the fast moving consumer goods industry. Marketing has been a passion of mine since a long time. This passion started already some years ago. While doing courses in marketing during my minor in Hong Kong, I found out that I am highly interested in this field and wanted to learn a lot more. During the courses of Marketing Management & Marketing Intelligence as well as writing my thesis, I learned a lot, for which I am grateful.
In truth, I could not have achieved this without a strong support group. I would like to thank my parents and boyfriend a lot for their support and help during my academic journey. I couldn’t have done it without you. Second, I would like to thank my supervisor dr. T. Bijmolt for his useful feedback and support while writing my thesis. Also, I would like to thank dr. H. Risselada and his student assistant Sjoerd van Haaren for providing me the right data and some useful insights about the datasets. Then, I would like to thank Keyvan Dehmamy for helping me with some issues regarding R. Finally, I would like to thank my friends and fellow students for their support and cooperation during the master.
Sterre Peters
TABLE OF CONTENTS
MANAGEMENT SUMMARY 3 PREFACE 5 1. INTRODUCTION 8 1.1 Important definitions 8 1.2 Research question 9 1.3 Relevancy 10 1.4 Contribution 10 1.5 Outline 11 2. THEORETICAL FRAMEWORK 12 2.1 Conceptual framework 122.2 The product category 14
2.2.1 Food and non-food products 14
2.2.2 External- and internal influences 15
2.3 The degree of newness 16
2.4 The business cycle 17
2.5 The interaction effect between the product category and the business cycle 18
3. METHODOLOGY 19 3.1 Data collection 19 3.2 Data description 20 3.2.1 Sample size 20 3.2.2 Aggregation 21 3.2.3 New variables 21 3.2.4 New dataset 22 3.3 Research method 22 3.3.1 Model classification 23 3.3.2 Bass model 23
3.3.3 Linear regression model 25
3.4 Data cleaning 25
3.4.1 Missings 26
3.4.2 Outliers 26
4. RESULTS 28
4.1 Descriptive statistics 28
4.2 Bass model estimation 29
4.3 Linear regression models 31
4.4 The product category 32
4.4.1 Hypothesis 1 32
4.4.3 Hypothesis 3 33
4.5 The degree of newness 33
4.5.1 Hypothesis 4 33
4.5.2 Hypothesis 5 33
4.5.3 Hypothesis 6 33
4.6 The business cycle 34
4.6.1 Hypothesis 7 34
4.6.2 Hypothesis 8 34
4.6.3 Hypothesis 9 34
4.7 The interaction effect between the product category and the business cycle 35
4.7.1 Hypothesis 10 35 4.7.2 Hypothesis 11 35 4.7.3 Hypothesis 12 35 4.8 Validation 37 4.8.1 Multicollinearity 37 4.8.2 R-squared 37 4.8.3 Other assumptions 37 5. CONCLUSION 38 5.1 Discussion 38 5.2 Managerial implications 40
5.3 Limitations and suggestions for further research 41
6. REFERENCES 43
7. APPENDICES 48
Appendix 1: Importance of innovation per product category 48
Appendix 2: Product Categories 48
Appendix 3 50
A) GDP in the Netherlands corrected for inflation 50
B) Economic growth based on GDP 50
Appendix 4: Outliers 51
Appendix 5: ANOVA output 52
Appendix 6: Linear regressions output 52
1. INTRODUCTION
In recent years, there is an enormous rise in new product introductions for fast moving consumer goods (FMCGs). The number of stock keeping units (SKUs) in US supermarkets was increased by 50% in the period 1980-2004 (Bhattacharya and Innes, 2016). More recently, the Netherlands scores the highest number regarding product introductions (9,5, adjusted to population size), compared to other countries such as England (6,6), Germany (3,4), and the US (2,9) (Innova Market Insights, 2011). The FMCG industry’s growth over the past quarter-century has been nothing short of breath taking (Chatterjee, Küpper, Mariage, Moore and Reis, 2010). However, also many new product introductions in the FMCG industry fail. The estimations of product failure lie between 45% (Castellion and Markhan, 2013) and 75% (Schneider and Hall, 2011). Next, the past economic recession led to an enormous rise of private label (PL) share in the expense of national brands (NB), e.g., 40% of the FMCG market in the US is controlled by PLs (Lamey, Deleersnyder, Steenkamp, and Dekimpe, 2012). Consequently, new product development is becoming more and more important in today’s world.
As already mentioned, there has been an enormous rise in the market share of PLs at the expense of national brands (Baltas, Doyle and Dyson, 1997). During times of an economic downturn, this increase of interest in PLs is even higher (Lamey, et al. 2012). Consequently, manufacturers of NBs are trying to find new strategies against their PL rivals. Many studies argue that NBs should come up with innovations to gain market share (Kumar and Steenkamp, 2007). Because of the high failure rate of new product launches as well as the rise in private label market share, many manufacturers of national brands are interested in studies about the innovation of new products.
1.1 Important definitions
“an idea, practice, or object that is perceived as new by an individual or another unit of adoption” (Rogers 1983, p. 11). The number of adopters and the timing of adoption are measured across different product categories. Adoption of a new product is the number of customers that buy the new product for the first time. The adoption processes show the sequence of stages by which consumers progress from unawareness to eventually the adoption of a product (Leeflang et al. 2017). The timing of adoption is related to the rate at which the innovation is diffused, or the speed at which members of the social system adopt the new product (Mahajan and Peterson, 1985). Within this research, there are two different types of new product introductions, namely new variants and new products. To prevent misunderstanding, new product introductions are (1) new variants of an already existing product, such as a new package or a unique flavour, or (2) new products that are new in the market, such as a cheese brand introducing cheese spread to its product line.
1.2 Research question
The following research question is proposed: “Are there differences in the diffusion process of new product introductions between food and non-food product categories within the FMCG industry, and does the newness of the product and the business cycle affect the diffusion?”
To answer the research question, the following sub-questions are defined:
1. Are there significant differences in the number of adopters, the timing of adoption, and the degree of internal influences between food and non-food products?
2. What is the influence of the degree of newness on the number of adopters, the timing of adoption, and the degree of internal influences?
3. What is the influence of the business cycle on the number of adopters, the timing of adoption, and the degree of internal influences?
or non-food products are more affected by internal influences. Then, the influence of the degree of newness per product is looked at to find out whether the degree of innovation impacts the number of adopters, the timing of adoption, and degree of internal influences. Finally, the importance of the business cycle is taken into consideration.
1.3 Relevancy
This thesis differs from other literature in two ways. First of all, this thesis treats products as separate cases rather than common effects to make a distinction between different product categories. Many previous studies were focusing on average results of new product introductions (Dean, Griffith, & Calantone, 2016); however, it is also vital to show differences in product categories concerning product-life cycle aspects. Secondly, this thesis shows the number of adopters for customer packaged goods, where most literature about the diffusion of innovation as well as the adoption of new product launches is focusing on durable goods (Mahajan, Muller and Bass, 1990).
1.4 Contribution
1.5 Outline
2. THEORETICAL FRAMEWORK
Figure 1 shows the conceptual framework that guides this research. First, some additional information will be provided to explain the motive of this study. Then, each arrow of the conceptual model will be discussed, and hypotheses are formulated.
Market launch of new products is a critical step in the new product development process (Beard and Easingwood, 1996). It is the stage where customers experience the new product for the first time, but also where many products fail. Product failure is the percent of new products introduced to the market that fail because of commercial goals that are not met. The product failure rate of consumer goods is studied by different researchers and lies somewhere between 45% (Castellion and Markhan 2013) and 75% (Schneider and Hall, 2011). As a result, new product development process is crucial but also risky, especially the introduction of new products. The buying behaviour of consumers has been changed a lot; therefore new product development for fast moving consumer goods (FMCG) is crucial. The Nielsen Company (2015) found some elements that play a major role in the purchase of new products: the fact that the product is something new (26%), and getting to know the product via word-of-mouth (21%), also studies in this thesis.
Additionally, in the period from 1998-2008, PLs had a more prominent growth than NBs in the United States on a yearly basis (Steenkamp, van Heerde, and Geyskens, 2010). In the Netherlands, the market share of private labels is steadily increasing by 1.3% in 2013, 0.3% in 2014, and 0.7% in 2015. In 2015, the market share of private labels in the Netherlands was estimated to be 37.4% (The Nielsen Company, 2016). Steenkamp, et al. (2010) mentioned in their study the factors that made people buy the price premium for NBs rather than PLs. One of these factors was the degree of innovation for national-brand goods. Private-label goods are known for its lower quality in comparison to national-brand goods. Product innovation helps to increase the quality gap between the two types of products, which eventually increases the willingness to pay of NBs. To add on that, these innovations do not have to be major innovations, only minor changes already make a difference (Gielens and Steenkamp, 2007). 2.1 Conceptual framework
adopters is being shown on a cumulative basis over time (Rogers, 1976). This is known as the diffusion model of innovation (Bass, 1969). The purpose of the diffusion model developed by Bass is to portray the increases in the number of adopters and forecast the development of a diffusion process. To be more specific, the diffusion models for product innovations focus on the development of a lifecycle curve to predict the first-purchase sales (Mahajan, Muller and Bass, 1990). The diffusion of innovation has three elements: the number of ultimate adopters (m), the coefficient of innovation (p), and the coefficient of imitation (q) (Bass, 1969). The coefficient of innovation measures the timing of adoption. When the coefficient of innovation is high, the number of innovators is high and products are adopted soon. The coefficient of imitation indicates the degree of internal influence, so whether the product is affected by internal influences. Consumers are for instance influenced by word-of-mouth communication. Therefore, there are three dependent variables: the number of adopters (m), the timing of adoption (p), and the internal influences (q) Besides, this thesis tries to find out whether the type of product category, the degree of newness of the product, and the business cycle are drivers of the dependent variables. These drivers are explained in the following sections.
FIGURE 1: Conceptual framework New product introductions (FMCG)
Note: FMCG = fast moving consumer goods
2.2 The product category 2.2.1 Food and non-food products
To see whether the product category affects the diffusion process, the product categories are divided into food and non-food categories. The food category consists of the product categories: mineral water, coffee, ice cream, sports drinks, yoghurt, cereal, soft drinks, biscuits, juices, and beer. The product categories dental care, household cleaners, pet food, hair care, and laundry detergent belong to the non-food class. To find out whether product categories show differences regarding innovation, several product categories, which are also available in the dataset, are explained in the following part.
In September 2014, the Nielsen Company provided a “breakthrough innovation report” in Europe, where different innovations of FMCG were studied. This study showed the importance of innovation to a category versus the category performance of the previous year. In table 1, the outcomes of several product categories are provided. This is also presented in figure 4 of Appendix 1. As can be seen, the food products show a different performance compared to non-food products, where food products show on average a better performance. Furthermore, the food products mineral water, coffee and ice cream stand out from the rest of the products concerning performance. Finally, the performance of the product categories within the non-food category is similar. Studies that found reasons for this difference in performance between food- and non-food products are described later in this section.
Product category Value % growth of category 2013 vs. prior year
% of new product sales from total category sales
Dental care -1 9
Household cleaners -4 10
Pet food cat and dog -4 8
Hair care -4 11
Laundry detergent -7 11
Table 1: importance of innovation to a category versus the category performance of the previous year
2.2.2 External- and internal influences
Communication can be divided into two complementary forms that can play a different role: external and internal. External communication (for instance advertising) can be very useful in the beginning phase of the adoption process, and internal communication (interpersonal communication within a social system) is effective when the adoption process has already started. The intensity of both forms of communications differs per innovation (Ruiz-Conde, 2005). Based on both communication forms, there are also two types of customers in the diffusion process of new products: “Innovators” influenced by mass-media communication (external influence) and “Imitators” influenced by word-of-mouth communication (internal influence) (Bass, 1969). As already mentioned before, in a study by the Nielsen Company was found that 21% of the people get to know the product via word-of-mouth and plays a major role in the adoption of new products. The innovators are the first customers that adopt the new product. Imitators, unlike innovators, are influenced by the decision of other consumers (Bass, 1990).
A study focusing on the most important factors that determine the buying behaviour of new products measured that the curiosity factor, as well as the interest factor, are high for the food product categories (Ashokkumar and Gopal, 2009). That means that consumers are curious to try out new things within the food product category. Moreover, Baker, Donthu, and Kumar (2016) did a study relating to word-of-mouth communication, in which they found that the total percentage of word-of-mouth communication was considerably higher for beverages and food/dining products (food category) than for household products and beauty/ personal care (non-food category).
Since the word-of-mouth percentage is high for food products, it is expected that food products are more affected by internal influences than non-food products.
Hypothesis 1: The number of adopters of food products is higher than for non-food products. Hypothesis 2: Food products are adopted sooner than non-food products.
Hypothesis 3: Food products are more affected by internal influences than non-food products.
2.3 The degree of newness
Several studies (Henard and Szymanski, 2001) mentioned that innovation and performance are positively related. However, the degree of the uniqueness of a product is often different. Product newness can be defined as the degree to which the product is new to the firm and new to the market (Olson, Walker, and Rucker, 1995). This study comprises two variations of newness: (1) new variants and (2) new products, where the degree of newness is higher for new products. When the product is new to the firm and the customer, there is a general lack of knowledge about the product (Bonner, 2010). For extremely new products, so a high degree of newness, customers have problems providing relevant feedback and the organization of the new product may find it difficult to understand how the new product will be consumed by the customers (Narver, et al. 2004). New products are often developed to gain product advantage, however, this benefit of product advantage through product innovativeness might be lowered because of the negative relationship with customer familiarity. That means that higher product innovativeness is likely to ask customers to adapt their habit of consuming (Calantone, Chan, & Cui, 2006). Furthermore, an essential determinant of the adoption of new products is information about the innovation. Uncertainty about the new product will slow the rate of adoption (Hall, 2004). The more original the product, the less information is available and the higher the degree of uncertainty.
Hypothesis 4: The degree of newness negatively influences the number of adopters.
Hypothesis 5: Products with a low degree of newness are adopted sooner than products with a high degree of newness.
Hypothesis 6: Products with a low degree of newness are more affected by internal influences than products with a high degree of newness.
2.4 The business cycle
Lamey, et al. (2012) found that PL share increases in times of economic downturn. In flourishing economic times, many customers keep buying PLs because they become familiar with the brand and the quality of the product. During an economic downturn, many NB manufacturers start to increase their prices to make up for the profits they lose. For instance, Unilever and Proctor & Gamble pushed up their prices when times were tough, and shoppers are choosy (Financial Times, 2012). However, such a strategy results in even fewer profits meaning that it works counterproductive (Deleersnyder et al. 2009, Lamey et al. 2012, Steenkamp and Fang 2011). Customers are creating a higher level of price awareness and are searching for lower prices in economic slowdowns (Estelami, Lehmann, and Holden, 2001). As already mentioned, consumers become more price-sensitive in times of economic slowdown; therefore it can be expected that people are more risk averse and prevent buying new product introductions. Consumers are preferably looking for price deals (Quelch, 2008). That results in the hypothesis that during times of contraction, or economic slowdown, there are fewer adopters of new product launches. Furthermore, consumers create a higher level of price awareness when the economy is diminishing, resulting in consumers being less innovative and interested in new product introductions. It is expected that there are fewer innovators during contraction, as well as consumers being influenced internally.
Hypothesis 7: The number of adopters is lower for products introduced during contraction than during expansion.
Hypothesis 8: Products introduced during contraction are adopted later than products introduced during expansion.
2.5 The interaction effect between the product category and the business cycle
Then, the conceptual model illustrates an interaction effect between the product category and the business cycle. During times of contraction, consumers become more price-sensitive as mentioned before by for instance Estelami, et al. (2001). Consumers are looking for price deals (Quelch, 2008). As a result, consumers would rather buy food products in the supermarket than at the butchery or other specialty stores, because of the lower prices at supermarkets. The “Centraal Bureau van de Statistiek” (CBS), the Dutch company (that collects data about, e.g. economic growth and the population), showed results about the price increases at the greengrocery, the bakery, and the butchery. The prices of these specialty stores increased a lot from 2006 until 2011, were the prices at the supermarkets were considerably less (CBS, 2012). Hence, it can be expected that the number of adopters during contraction is affected by the product category.
Hypothesis 10: The product category and the business cycle have an interaction effect on the number of adopters.
Hypothesis 11: The product category and the business cycle have an interaction effect on the timing of adoption.
Hypothesis 12: The product category and the business cycle have an interaction effect on the degree of internal influences.
3. METHODOLOGY
In this chapter, the data analysis will be explained. First, the data collection is explained, followed by a data description. Then, the research method of this thesis is elaborated upon by looking at the model classification, the Bass model, and the linear regression models. Also, data cleaning is performed and discussed in this chapter.
3.1 Data collection
For this thesis, different datasets are used that are collected by the marketing research firm GfK, a company that provides external data. The datasets are provided in table 2. The first dataset contains information of over 660,000 low-involvement products, aggregated on product level. It includes the following variables: “Barcode attributes”, “Barcode description”, “Private label”, “Brand”, “Category name”, “Measurement unit”, “pa From Date”, and “pa To Date”. The next eight datasets consist of panel data that is measured in the period 2008-2015, where each dataset contains information about product purchases within one year. The dataset is on individual level and measures daily sales per individual per product. Per year, around 15,000,000 observations are provided, including the following variables (in each dataset the same): “Panellist”, “Data of purchase”, “Banner name”, “Barcode”, “Total unit sales”, “Total value sales”, “Total volume sales”, and “Quarter”. The tenth and final dataset contains information about product innovations in the period 2008-2015 in the Netherlands. In total, 9,079 products are included in this dataset, which are products that are launched in the period of measurement (including 2007). The dataset is aggregated on product level and includes the following variables: “Country name”, “Category name”, “Barcode”, “First purchase”, “Last purchase”, and “Product type”. The datasets and variables are described in more detail in the next section.
Dataset Information 1 Barcode attributes 2:9 Purchases 2008:2015
3.2 Data description 3.2.1 Sample size Product categories
Since this thesis concerns new product introductions, only the new products that are launched between 2007 and 2015 are relevant to use. These products are divided into multiple product categories. In total, the dataset contains 38 different product categories, including non-food as well as food products. Some categories overlap with other categories, for instance dry cat food, wet cat food, dry dog food, and wet dog food are combined into one category ‘pet food’. That results in a new dataset containing 9,079 products within 15 “overlapping” product categories, as can be seen in Appendix 1.
The dataset including the innovations is used for further analysis because new product introductions help to answer the research question. Some relevant variables of the other datasets were added to the dataset. Variables that are not important are deleted to make the final dataset more structured.
Newness
Business Cycle
To find out whether the business cycle does influence the dependent variables “number of adopters” and “timing of adoption”, the GDP per year in 2008-2015 of the Netherlands will be used. As described by “Centraal Bureau van de Statistiek” (CBS), the gross domestic product (GDP) can be used to show the economic situation of a country. GDP is also used in this thesis to indicate whether the economy is flourishing or not. The data provided by the CBS about the GDP in the Netherlands can be found in Appendix 3A and Appendix 3B. The values of the table of Appendix 3A are corrected for inflation, where the price level of 2010 is taken as the reference group (CBS, 2017). The years 2008, 2010, 2011, 2014 and 2015 could be seen as times of expansion, whereas 2009, 2012 and 2013 are times of contraction. For the analysis, dummies are used to indicate whether the product was launched during contraction or expansion.
3.2.2 Aggregation
The initial datasets were either aggregate level data or individual level data. To be able to test the hypotheses provided before, the new dataset is aggregated on the product level meaning that each unique barcode number is provided once. Furthermore, some new variables are added to see the quarterly purchases of the product over the 8-year period. The new variables are explained in the next section.
3.2.3 New variables
The initial dataset already included the following variables:
1. A variable “week” indicating the week the product is purchased.
2. Dummy variables (for every week) that indicate whether the purchase is a first purchase or a repeat purchase.
To find the parameters of the Bass model, some extra variables need to be added to the initial dataset:
1. Every 13 weeks of the variables “first purchase” are combined to get quarterly data of the first purchases.
2. A variable indicating the cumulative amount of first purchases per quarter (Cumsales).
4. A variable indicating the cumulative amount of first purchases from the previous quarter squared (Cumsalessqrt).
3.2.4 New dataset
The initial dataset contains the purchases in rows; however, in order to estimate the Bass model, these values need to be in columns. Therefore, a new dataset is created containing the variables “P“, “Sales”, “Cumsales”, “Cumsaleslag”, “Cumsalessqrt”, and “ProdID”. The variable “P” indicates the period in which the purchase took place. In total, there were 29 periods in which purchases took place. However, the last period was shorter (less than 13 weeks). Therefore, period 29 is not used for further estimation. Period 0 indicates the fourth quarter of 2008 and period 28 indicates the third quarter of 2015. The variable “Sales” is the amount of first purchases per period. The variable “Cumsales” shows the cumulative amount of first purchases per period. The variable “Cumsaleslag” indicates the cumulative amount of first purchases of the previous period, where “Cumsalessqrt” is the squared value of “Cumsaleslag”, and the final variable “ProdID” indicates the barcode of the product to which the previously mentioned variables belong. This dataset contains all the variables that are needed to calculate the parameters of the Bass model per product, which will be explained in the next section.
3.3 Research method
models are at the macro level (Leeflang et al. 2017). For this thesis, the diffusion models are more useful because the hypotheses are not related to individual-level data.
3.3.1 Model classification
Models can have different purposes. Based on Leeflang, Wittink, Wedel, and Naert (2000), models can be divided into descriptive, predictive, and normative models. Descriptive models describe the decision processes of managers or consumers. Predictive models try to forecast or predict events or outcomes. Normative models are used to get optimal courses of action (Leeflang, et al. 2000). The traditional diffusion models, and in this case the Bass model can be used for forecasting and evaluating the hypotheses relating to influences of an innovation, which is a descriptive purpose.
3.3.2 Bass model
The Bass model mentions that part of the first customers is influenced only by mass-media communication (external influence) and the other just by word-of-mouth communication (internal influence). Bass divides customers into these two groups, the first group being the “Innovators” and the second group the “Imitators”. The Bass model in absolute terms (number of adopters) has the following equation (Leeflang, et al. 2017):
! ! =!" !!" = ! ! − ! ! +!!! ! ! − ! ! (1) Where
!(!) = The number of initial adopters at time !; !(!) = The cumulative number of adopters at time !; ! = The number of ultimate adopters (market potential); ! = The coefficient of innovation;
! = The coefficient of imitation.
buyers, referred to ! as the “coefficient of imitation” (Mahajan, Muller and Bass, 1990). These adopters are influenced internally, so through interpersonal communication between prior adopters and potential adopters within the social system (Mahajan and Peterson, 1985). The Bass model in relative terms (the fraction of potential adopters who adopted the product at time t: F(t)) has the following equation (the parameters are the same as the previous equation) (Leeflang, et al. 2017):
! ! = !" + ! − ! ! ! −!!(!(!))! (2)
Or simplified (Leeflang, et al. 2017):
! ! = !!+ !!! ! + !! (!(!))! (3)
The parameters can be estimated using the ordinary least squares (OLS) method, a time-invariant estimation procedure. The simplified Bass model can be translated to make parameter estimation possible and looks as follows:
! ! = !!+ !!! !" − 1 + !!!! !" − 1 + ! ! (4)
Where
!!= Sales of the period
!!! !" − 1 = Cumulative sales up to the previous period
!!!! !" − 1 = Cumulative sales up to the previous period squared
Then, the parameters can be estimated by filling in the following formulas (Mahajan and Peterson, 1985): ! =−!!− !! !− 4! !!! 2!! ! =!! ! ! = −! ∗ !!
After finding the parameters of the products, these values were added to the variables !, ! and !.
3.3.3 Linear regression model
The next step of this thesis is to find out whether there are differences in the number of adopters, timing of adoption, and internal influences effects for food and non-food products, new variants and new products, and during contraction or expansion. Also, the interaction effect between the product category and the business cycle is measured by the regression models. To find these differences, linear regression models are performed. Linear regression model 1 tests whether the independent variables product category, degree of newness, and the stage of the business cycle have significantly different values for the dependent variable ! (the number of adopters). Linear regression model 2 tests whether these independent variables have significantly different values for the dependent variable ! (the timing of adoption). Linear regression model 3 tests whether these independent variables have significantly different values for the dependent variable ! (the internal influence). The models look as follow:
! = !!+ !!!!+ !!!!+ !!!!+ !!!!!! + ! Where
Y = (1) The number of adopters, (2) the timing of adoption, (3) and the internal influence rate !!= The intercept
!!!!= The product category (food or non-food)
!!!! = The degree of newness (new variant or new product)
!!!! = The business cycle (contraction or expansion)
!!!!!! = The interaction effect of product category and business cycle 3.4 Data cleaning
Bass model predictions. Another reason is that the Bass model can be used to predict future sales (Mahajan, 1990), which is only possible when the products are still active and available in supermarkets. When continuing the analysis with products that are still active, these results can be used to predict future sales. Based on these arguments, the products that are not active anymore are deleted from the dataset. That results in 5,622 products that are used for further estimation.
3.4.1 Missings
Besides, the Bass diffusion model is initially used for durable products (Bass, 1969). In this thesis, the Bass diffusion model is used to analyse non-durable goods. Consequently, it is important to see whether the model performs well. It is assumed that the parameters ! and ! of the Bass model have a value between 0 and 1 (Bass, 1969). Literature also mentioned that these values must be positive in order for the model to make sense (Bass 1969; Jiang, Bass 2006). Therefore, the products that are outside the range of 0 and 1 are deleted from the dataset. Also, the parameter estimation of ! resulted sometimes in a missing value, because of the reason that !3 was a negative value. The products containing missing values are also deleted from the dataset because they cannot be used to perform the comparison analysis. Some reasons why the Bass model cannot be estimated correctly for all products is because (1) there are less than four periods of initial purchases or (2) there are less than 25 purchases in total. Both cases indicate a negative value for either ! or ! or a missing value; therefore these are not taken into account for the estimation of the Bass model. That results in a new dataset of 3,204 observations (or products).
3.4.2 Outliers
Furthermore, the product category Soft Drinks contains an outlier. For this, the same holds that the outlier can manipulate the outcomes and therefore this product is also deleted from the dataset. Finally, the product category Dental Care and Coffee show an obvious outlier and also one product of Dental Care and one product of Coffee are deleted from the dataset. Furthermore, the variable “pa from date” has some outliers. The variable indicates when the product was introduced and by looking into the data it became clear that three products were launched in 1970, which is not defined as a new product and deleted from the dataset. Also, three products were introduced in 2016, even though the first purchase already took place years before. Since this is very unrealistic, these three products were also deleted from the dataset. Finally, 64 products were introduced on 01-01-2007, where the first purchase took place years later. It is expected that the introduction date is not well documented, and therefore these 64 products are also deleted from the dataset. That results in the final dataset containing 3,130 products that are used for further analysis. The variables of the final dataset are described in table 3.
Barcode Special barcode for each product Category name The name of the category
First purchase Date of the first purchase Last purchase Date of the last purchase
Product type Whether the product is a new variant or a new product (dummy) Brand Name of the brand of the product
Pa from date Date when the product was introduced
Pa to date Date of removal of the product or “active” if the product is still available M Value of the parameter ! of the product
P Value of the parameter ! of the product Q Value of the parameter ! of the product
Product category Whether the product is a Food product or a Non-food product (dummy) Date Same as “Pa from date”, but only the indicating the year
4. RESULTS
This chapter starts with some descriptive statistics, followed by the results of the Bass model parameter estimations. Furthermore, the differences regarding the product category, the newness of the product, and the business cycle are discussed. The chapter ends with a validation test.
4.1 Descriptive statistics
First of all, some descriptive statistics are provided. In total, there are 15 product categories of which 10 belong to the food product category (mineral water, coffee, ice cream, yoghurt, cereal, soft drinks, biscuits, juices, beer, and chocolate) and 5 to the non-food product category (dental care, household cleaners, pet food, hair care, and laundry detergent). Moreover, 1,600 (51%) products fall within the food category whereas 1,530 (49%) products fall within the non-food category. Furthermore, the dataset consists of 280 (8.9%) new products and 2,850 (91.1%) new variants. The amount of products observed per year is visualized in the graph below. Because the number of products per year in 2007 and 2008 are too small, these are combined with 2009. Finally, the amount of products that are launched during contraction is 1,486 (47.5%), where 1,644 (52.5%) are launched during times of expansion.
Graph 1: Amount of new products per year
Additionally, the following table shows the observations and the average values of the parameters of the Bass model per variable:
Variable Dummy Observations ! ! ! Product
category
Food 1600 305.8004 0.02248856 0.2743683
Non-food 1530 183.0854 0.01872049 0.3089747
Product type New variant 2850 247.4118 0.01990715 0.2924601 New product 280 229.5627 0.02817385 0.2793197 Business
cycle
Contraction 1486 260.2542 0.02026715 0.2543268 Expansion 1644 232.7638 0.02098970 0.3246904 Table 4: Descriptive statistics
4.2 Bass model estimation
The average values of the parameters of the Bass model per product category can be found in table 5. This chart makes it more coherent what the values of the parameters are, however, the actual numbers are taken into account for the hypotheses testing instead of the average values. At first glance, the non-food products seem to have a lower parameter of !, in other words, a smaller number of adopters within the estimation period. Before continuing the analysis, an ANOVA test is performed to find evidence whether there is a difference between food and non-food products in the number of adopters. In appendix 5, the results of the ANOVA test are provided. Since the p-value is lower than the significance factor of 5% (p-value < 0.05), the finding above is confirmed, and the parameter of ! is significantly different between food and non-food products. Later in this section, the difference will be discussed in more detail. Besides, it is remarkable that the Bass model can be used for non-durable goods as well, even though most of the studies of diffusion models are focusing on durable goods.
Juices 170 261.7164 0.02854356 0.2928858 Beer 67 216.4843 0.02625218 0.2944063 Chocolate 209 369.0517 0.0145227 0.1998061 Dental Care 368 183.1014 0.01536589 0.2929294 Household cleaners 324 241.6513 0.01525706 0.3236308 Pet Food 160 144.0947 0.02360813 0.2547318 Hair Care 422 127.2041 0.02393311 0.2888366 Laundry Detergent 256 225.4263 0.01627871 0.380589 Total 3,130 - - -
Table 5: Estimation of the parameters per product category
In figures 2 and 3 below, the Bass model is illustrated over a period of 28 quarters, starting in the fourth quarter of 2008 (T=0) until the third quarter in 2015 (T=28). The average values of the Bass model parameters are used, that are shown in the table explained before. The green line shows the food product categories Biscuits and Ice cream, and the blue line indicates the non-food product categories Dental care and Laundry detergent. As you can see, the green line of both figures shows a higher number of adopters (or market potential). Also, it could be seen that products belonging to the category food are adopted sooner because the green line is steeper at the beginning of the product lifecycle. In the following parts, the overall results of the differences in the number of adopters, the timing of adoption, and the degree of internal influence will be provided.
Figure 2: Average values of the Bass model parameters
4.3 Linear regression models
In order to find out whether there are significant differences in the Bass parameters for the independent variables, three linear regression models are performed. The outcomes of the regression models that are used for the hypotheses testing are shown in the tables below and in Appendix 6. It is important to mention that all models are significant (p-value < 0.001) and can be used for further analysis.
! Estimate Std. Error t-value Pr(>|t|)
(Intercept) 253.07 24.67 10.258 < 2e-16 ***
Product category – Non-food -95.92 18.42 -5.209 2.02e-07 *** Product type – New Variant 28.33 23.38 1.212 0.22568 Business Cycle - Contraction 58.36 18.69 3.123 0.00181 ** Product category – Non-food and
Business Cycle - Contraction
-58.83 26.71 -2.203 0.02767 *
*** p-value < 0.001, ** p-value < 0.01, * p-value < 0.05 Multiple R-squared: 0.02975, Adjusted R-squared: 0.02851 F-statistic: 23.96 on 4 and 3125 DF, p-value: < 2.2e-16 Table 6: Regression output model 1
! Estimate Std. Error t-value Pr(>|t|) (Intercept) 0.031293 0.002307 13.564 < 2e-16 *** Product category – Non-food -0.006168 0.001722 -3.582 0.000347 *** Product type – New Variant -0.008019 0.002187 -3.667 0.000249 *** Business Cycle - Contraction -0.003427 0.001748 -1.961 0.049992 * Product category – Non-food and
Business Cycle - Contraction 0.005504 0.002497 2.204 0.027605 * *** p-value < 0.001, ** p-value < 0.01, * p-value < 0.05
! Estimate Std. Error t-value Pr(>|t|) (Intercept) 0.312685 0.015724 19.886 < 2e-16 *** Product category – Non-food 0.008654 0.011738 0.7373 0.461016 Product type – New Variant 0.008571 0.014903 0.575 0.565228 Business Cycle - Contraction -0.098428 0.011911 -8.264 < 2e-16 **** Product category – Non-food:
Business Cycle - Contraction
0.056383 0.017022 3.312 0.000935 *** *** p-value < 0.001, ** p-value < 0.01, * p-value < 0.05
Multiple R-squared: 0.0302, Adjusted R-squared: 0.02896 F-statistic: 24.33 on 4 and 3125 DF, p-value: < 2.2e-16 Table 8: Regression output model 3
4.4 The product category 4.4.1 Hypothesis 1
For hypothesis 1 “the number of adopters of food products is higher than for non-food products”, the outcomes of table 6 are taken into account. As already described before, the parameter ! is significantly different for food products compared to non-food products. In this case, the food products are the reference category. Table 6 shows that the effect of non-food products on the number of ultimate adopters is -95.92. As hypothesized, non-food products have a higher market potential of 95.92 compared to non-food products. Consequently, hypothesis 1 is supported; the number of adopters of food products is higher than for non-food products.
4.4.2 Hypothesis 2
4.4.3 Hypothesis 3
The results of hypothesis 3 “food products are more affected by internal influences than non-food products” are presented in table 8 above. The p-value of the effect of the product category on the parameter ! is higher than the significance factor of 5% (p-value: 0.461016). That means that the coefficient of imitation does not differ between the food- and non-food categories and further analysis about the estimates is not allowed. Consequently, hypothesis 3 is not supported, since there is no evidence that food products are more affected by internal influences than non-food products.
4.5 The degree of newness 4.5.1 Hypothesis 4
The dataset consists of products that are very new (new products) or moderately new (new variants). To test hypothesis 4 “the degree of newness does have a negative influence on the number of adopters”, the results of table 6 are taken into account. The p-value of the parameter ! for product type is higher than the significance level of 5% (p-value: 0.22568), which means that the values are not significantly different. As a consequence, the number of adopters does not significantly differ between very new and moderately new products, which makes further analysis not allowed. Hypothesis 4 is not supported, since there is no evidence that the degree of newness does have a negative influence on the number of adopters.
4.5.2 Hypothesis 5
For hypothesis 5 “products with a low degree of newness are adopted sooner than products with a high degree of newness” the outcomes can be found in table 7. The effect of the product type on parameter ! has a significant p-value (p-value < 0.001), therefore the coefficient of innovation differs per degree of newness. The products that are considered as “new products” are the reference category. The effect of new variants on parameter ! is -0.008, meaning that the fraction of adopters being innovators is 0.008 lower for products with a lower degree of newness. Since the coefficient of innovation is higher for products with a high degree of newness, hypothesis 5 is not supported. There is an opposite effect; products with a high degree of newness are adopted sooner than less innovative products.
4.5.3 Hypothesis 6
made clear that the p-value of this effect is higher than the significance level of 5% (p-value: 0.565228), meaning that the coefficient of imitation does not significantly differ per degree of innovativeness. Consequently, hypothesis 6 is not supported; there is no evidence found that products with a low degree of newness are more affected by internal influences than products with a high degree of newness.
4.6 The business cycle 4.6.1 Hypothesis 7
To test whether hypothesis 7 “the number of adopters is lower for products introduced during contraction than during expansion” can be supported; the results of table 6 are discussed. The p-value of the business cycle on parameter ! is lower than the significance level of 1% (p-value < 0.01), meaning that the number of adopters does significantly differ between products introduced during contraction and expansion. Here, the reference category is expansion. Contraction has a positive effect (estimate: 58.36) on the number of adopters, hence products that are launched during contraction have a higher market size of 58.36 than during expansion. Hypothesis 7 is not supported, because there is an opposite effect. Even though these results are found, it might be biased because some products are introduced in 2011 and some in 2015. That means that the estimation period of products launched in 2015 is considerably lower, which has an effect on the estimation of the number of adopters.
4.6.2 Hypothesis 8
For hypothesis 8 “products introduced during contraction are adopted later than products introduced during expansion”, the results of table 7 are discussed. The p-value of the business cycle on parameter ! is lower than the significance level of 5% (p-value < 0.05), meaning that the coefficient of innovation does significantly differ between economic times of contraction and expansion. Contraction has a negative effect (estimate: -0.003) on the number of innovators, for products introduced during contraction the fraction of all adopters being innovators is 0.003 lower than for products introduced during expansion. Hypothesis 8 is supported, because products introduced during contraction are adopted later than products introduced during expansion.
4.6.3 Hypothesis 9
hypothesis testing, the results of table 8 are taken into account. The p-value is lower than the significance level of 0.1% (p-value < 0.001), meaning that the coefficient of imitation significantly differs between contraction and expansion. Contraction has a negative effect (estimate: -0.984) on the number of imitators, so for products introduced during contraction the fraction of all adopters being imitators is 0.0984 lower than during expansion. Hypothesis 9 is supported, because products introduced during contraction are less affected by internal influences than products introduced during expansion
4.7 The interaction effect between the product category and the business cycle 4.7.1 Hypothesis 10
The result of hypotheses 10 “the product category and the business cycle have an interaction effect on the number of adopters.” is shown in table 6. The interaction effect of the product category and the business cycle, on the number of adopters, has a lower p-value than the significance level of 5% (p-value < 0.05). Therefore, the estimate can be interpreted. The reference category of the interaction effect is food products and expansion. The interaction effect has an estimate of -58.83. That means the following: the effect of contraction on the number of adopters is (58.36 + -58.83) 0.47 lower when the products belong to the non-food category. Therefore, hypothesis 10 is supported, the product category and the business cycle have an interaction effect on the number of adopters, however, the effect is extremely small. 4.7.2 Hypothesis 11
Then, hypothesis 11 “the product category and the business cycle have an interaction effect on the timing of adoption.” is tested, of which the outcomes are presented in table 7. The interaction effect of the product category and the business cycle, on the timing of adoption, has a lower p-value than the significance level of 5% (p-value < 0.05). Therefore, the estimate can be interpreted. The interaction effect has an estimate of 0.006. That means the following: the effect of contraction on the number of adopters is (-0.003 + 0.006) 0.003 higher when the products belong to the non-food category. Therefore, hypothesis 11 is supported, the product category and the business cycle have an interaction effect on the timing of adoption.
4.7.3 Hypothesis 12
Therefore, the estimate can be interpreted. The interaction effect has an estimate of 0.056. That means the following: the effect of contraction on the number of adopters is (-0.098 + 0.056) 0.042 lower when the products belong to the non-food category. Therefore, hypothesis 12 is supported, the product category and the business cycle have an interaction effect on the degree of internal influences.
Finally, a summary of the hypotheses conclusions can be found in table 9.
Hypothesis Supported Remarks
H1 The number of adopters for new food products is higher than for new non-food products.
Yes H2 Food products are adopted sooner than non-food products Yes H3 Food products are more affected by internal influences than
non-food products.
No Opposite
effect H4 The degree of newness does have a negative influence on the
number of adopters
No evidence H5 Products with a low degree of newness are adopted sooner than
products with a high degree of newness
No Opposite
effect H6 Products with a low degree of newness are more affected by
internal influences than products with a high degree of newness
No evidence H7 The number of adopters is lower for products introduced during
contraction than during expansion
No Opposite
effect H8 Products introduced during contraction are adopted later than
products introduced during expansion
Yes H9 Products introduced during contraction are less affected by
internal influences than products introduced during expansion
Yes H10 The product category and the business cycle have an interaction
effect on the number of adopters.
Yes H11 The product category and the business cycle have an interaction
effect on the timing of adoption.
Yes H12 The product category and the business cycle have an interaction
effect on the degree of internal influences.
4.8 Validation
4.8.1 Multicollinearity
First of all, multicollinearity means that the predictors are correlated with each other that can lead to unreliable parameter estimates (Leeflang, et al. 2015). For descriptive modelling, it is essential that the parameter estimates are reliable; hence multicollinearity could lead to biased results. To check for multicollinearity, the variance inflation factor (VIF) is used, which should be lower than 5. The VIF scores can be found in table 10. Since the VIF values are < 5, there is no issue of multicollinearity.
Variable VIF score
Product category 1.907277
Product type 1.002194
Business cycle 1.959823
Product category and business cycle 2.892310 Table 10: VIF scores
4.8.2 R-squared
Besides, the squared values are interpreted and can be found in the table below. The R-squared values are low, which can be explained by the fact that there are much more variables needed for the model to explain the variation in the response variable. There are more variables that can predict the results, however that does not affect the validation of the parameter estimates. Therefore, the low R-squared is not considered a problem of validation. Regression Multiple R-squared
Model 1 0.02997 Model 2 0.008781 Model 3 0.03277 Table 11: R-squared scores 4.8.3 Other assumptions
which is not the interest of this thesis. As a result, a Durbin Watson test is not useful to perform.
5. CONCLUSION
The last chapter of this thesis begins with a discussion of the results, followed by managerial implications. Furthermore, limitations and suggestions for further research are provided. 5.1 Discussion
Recently, there is an increase in the studies about new product introductions of fast moving consumer goods. The market share of private labels has been increased a lot in the expense of national brands. Some studies argue that national brands should invest in new product development to fight against the private labels. Hence, marketing managers of national brands are interested in studies about the process of new product introductions. This study contributes to that because it shows how new product introductions of fast moving consumer goods perform and explains whether the product category, the degree of newness and the economic situation show a different process concerning the diffusion process of product innovations. Remarkably, most of the diffusion studies are related to durable goods. However, this study shows that diffusion models can also be implemented to non-durable goods. The following research question was proposed: Are there differences in the diffusion process of food and non-food product categories within the FMCG industry, and does the newness of the product and the business cycle affect the diffusion?
imitators are measured to be higher during expansion. Based on the results of this thesis, it can be concluded that products that are launched during times of expansion are adopted sooner than products that are launched during times of contraction. Also, products introduced during times of expansion are more affected by internal influences, meaning that more consumers are influenced by word-of-mouth communication. These findings are is in line with the findings of e.g. Quelch (2008) and Lamey, et al. (2012) because they say that consumers become less risk averse in times of economic expansion.
Is there an interaction effect between the product category and the business cycle on the number of adopters, the timing of adoption, and the degree of internal influence? The results showed that, as already expected, there is an interaction effect between the product category and the business cycle on the number of adopters, the timing of adoption, and the degree of internal influence. That is because all interaction effects were significant. First of all, the positive effect of contraction on the number of adopters is lower when products are non-food products, however, the effect is really small. Moreover, the negative effect of contraction on the timing of adoption is higher when products are non-food products, and the negative effect of contraction on the degree of internal influence is lower when products are non-food products. Overall, there is an interaction effect between product category and the business cycle on the number of adopters, the timing of adoption, and the degree of internal influence.
5.2 Managerial implications
This study is a contribution to the new product development process of fast moving consumer goods and shows marketing managers how various products have been adopted by consumers. In the next part, some marketing implications are provided that can help marketing managers to improve the launch of new products.
Kalish and Sen (1986) explain that if early adopters of a product positively influence the late adopters, a low introductory price is an effective pricing strategy. Non-food products have a high imitation factor, which means that many consumers wait a bit before adopting a new product and are influenced by internal factors such as word-of-mouth communication within their social system. Henceforth, managers can lower the prices of new non-food products to boost sales and make consumers adopt the product sooner.
against private labels. For food products, it is crucial to use other factors in launching new products, thinking of shelf places and the use of displays, especially in the Netherlands. That is because the Dutch consumer, in comparison to all European consumers (44%), buys new products less often (25%) when visiting a supermarket (Nielsen Company, 2015).
Furthermore, products that have a high degree of newness are more successful in the number of innovators adopting the product compared to a low degree of newness. This means that highly innovative products are adopted sooner than products that are rather similar to older versions of the product. Therefore, managers should think about this when investing in new product development. As already explained, the curiosity factor is an essential factor for food products; hence consumers are interested in adopting innovative fast moving consumer goods. Finally, managers can take the business cycle into account when investing in new product development. As literature suggests and also found in this thesis, consumers adopt new products sooner when the product was introduced during times of economic growth. Consequently, when managers see an economic growth, they might consider launching a new product.
5.3 Limitations and suggestions for further research
There are some limitations regarding this thesis that need to be acknowledged. First, while the Bass model has been used a lot in marketing research, it has several shortcomings. The model was already implemented in 1969, and since then there have been a lot of studies that reformulated the model. Marketing influences (advertising and price promotion), and price are not included in the model. The dataset of this thesis did not include variables about marketing instruments and price; however further research could improve this study by accounting for those variables. That would probably also increase the R-squared of the regression models.
Third, the results do not necessarily represent all markets worldwide. The data only contains information about the Dutch fast moving consumer goods market. Consequently, the findings are not generalizable to countries outside the Netherlands. Further research could improve this issue by including data from other countries to the initial dataset to make it more generalizable.
Fourth, the negative influence of expansion on the number of adopters is quite remarkable. As explained, this result might be biased because products that are launched in 2009 are compared to products that are introduced in 2015, which makes it quite obvious that the number of adopters is higher for products introduced earlier in the estimation period. Therefore, the results of the number of adopters per year have a high change to be biased. Further research could improve these results by comparing the products that are launched in the same year.
6. REFERENCES
Ashokkumar, S., & Gopal, S. (2009). Diffusion of Innovation in Private Labels in Food Products. ICFAI Journal Of Brand Management. 6(1). 35-56.
Baker, A. M., Donthu, N., & Kumar, V. (2016). Investigating How Word-of-Mouth
Conversations About Brands Influence Purchase and Retransmission Intentions. Journal Of
Marketing Research (JMR), 53(2), 225-239.
Baltas, G., Doyle, P., & Dyson, P. (1997). A model of consumer choice for national vs. private label brands. Journal of the Operational Research Society. 48(10). 988–995.
Bass, F. (1969). A New Product Growth Model for Consumer Durables. Management Science. 15(1). 215-27.
Beard, C., & Easingwood, C. (1996). New product launch: Marketing action and launch tactics for high-technology products. Management. 25(2). 87–103.
Bhattacharya, H., & Innes, R. (2016). Concentration, Product Variety, and Entry-for-merger: evidence from new product introductions in the U.S. Food Industry. American Journal Of
Agricultural Economics, 98(5), 1360-1376.
Bonner, J. (2010). Customer interactivity and new product performance: Moderating effects of product newness and product embeddedness. Industrial Marketing Management. 39. 485-492.
Calantone, R. J., Chan, K., & Cui, A. S. (2006). Decomposing Product Innovativeness and Its Effects on New Product Success. Journal Of Product Innovation Management, 23(5), 408-421.
Carson, S. J., Jewell, R. D., & Joiner, C. (2007). Prototypicality advantages for pioneers over me-too brands: the role of evolving product designs. Journal Of The Academy Of Marketing
Science. 35(2). 172-183.
Castellion, G., & Markham, S. K. (2013). Perspective: New product failure rates: Influence of argumentum ad populum and self-interest why do products fail? Journal of Product
Centraal Bureau van de Statistiek (2017). Opbouw binnenlands product (bbp); nationale rekeningen. Retrieved 3 December 2017, from
http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=82262NED&D1=128-142&D2=38-47&HDR=T&STB=G1&VW=T
Chatterjee, I., Küpper, J., Mariager, C., Moore, P., & Reis, S. (2010). Trends that will Shape the Consumer Goods Industry. McKinsey & Company. Retrieved 3 November 2017, from www.mckinsey.com/clientservice/consumerpackagedgoods
Dean, T., Griffith, D. A., & Calantone, R. J. (2016). New Product Creativity: Understanding Contract Specificity in New Product Introductions. Journal Of Marketing, 80(2), 39-58. doi:10.1509/jm.14.0333
Deleersnyder, B., Steenkamp, J-B., & Leeflang, P. (2009). The Role of National Culture in Advertising’s Sensitivity to Business Cycles: An Investigation Across Continents. Journal of Marketing Research. 46(10). 623-36.
Estelami, H., Lehmann, D., & Holden, A. (2001). Macro-Economic Determinants of Consumer Price Knowledge: A Meta-Analysis of Four Decades of Research. International Journal of Research in Marketing. 18(12). 341-355
Gielens, K. (2012). New Products: The Antidote to Private Label Growth?. Journal Of
Marketing Research (JMR), 49(3), 408-423.
Gielens, K., & Steenkamp, J. E. (2007). Drivers of consumer acceptance of new packaged goods: An investigation across products and countries. International Journal Of Research In
Marketing, 24(2), 97-111.
Hall, B. (2004). Innovation and diffusions. NBER Working Paper 10212. Cambridge.
van Heerde, H, Gijsenberg, M, Dekimpe, M., & Steenkamp, J-B. (2013). Price and Advertising Effectiveness over the Business Cycle. Journal of Marketing Research. 50(2). 177-193.
