The Marketing Effectiveness on Different Product Categories over the
Long Run
Leontine E. Veenje
Herestraat 97a, 9711LE, Groningen, (0)628248069, l.e.veenje@gmail.com Rijksuniversiteit Groningen, Department of Marketing, MSc Thesis Marketing Intelligence
First supervisor: M.J. Gijsenberg Second supervisor: K. Dehmamy
Abstract
How effective are advertising, distribution and price on the different kinds of classified product categories over the long run? We answer this question by analyzing the effects of the marketing instruments on dif-ferent products categories over 13 weeks. The product categories consists of low- or high-involvement products and hedonic or utilitarian products. We capture the cumulative generalized impulse response functions derived from a VARX model across 116 brands. Our results show that there are significant dif-ferences between the marketing effects and the different product categories. Price and distribution have relatively larger effect sizes on the different product categories than advertising. More specifically, price is found to be relatively more effective on low-involvement products compared to high-involvement products, whereas distribution appears to be more effective on high-involvement products than low-involvement products. In terms of hedonic and utilitarian products, we found that price has relatively more effect on hedonic products than on utilitarian products, whereas distribution has a greater effect on utilitarian products relative to hedonic products. All these product categories are combined in the Foote, Cone and Belding grid. This framework provides insights into the marketing effectiveness, and hence addresses the accountability, of the different marketing instruments on the combined product categories.
Keywords: Marketing instruments, product categories, involvement, hedonic, utilitarian products, time-series analyss
1. Introduction
The accountability of marketing is still a much debated topic. Every year companies struggle to allocate their marketing expenditures in the most effective way. The credibility of marketing within a company suffers from the lack of accountability (Rust et al. 2014). However, research shows that
Therefore, there is a growing interest in the ac-countability of marketing over the long run, taking these dynamics into account. To address the lack of accountability over the long run this research focuses on examining the effectiveness of three marketing instruments (i.e. advertising, distribu-tion and price) on different types of products over the long run. The different types of products con-sist of 1) high- or low-involvement products and 2) hedonic or utilitarian products. The product types will be further referred to as product catego-ries.
Examining the effects of the marketing instru-ments over time has become an established topic in marketing research. The effects of advertising and price on brand performance has been studied thoroughly. However, distribution is one of the few instruments that received less attention. Ata-man et al. (2010) found that the sales elasticity is 0.74 for distribution (for the short-term plus the long-term), which has a relatively larger effect than price and advertising. The authors attempt to answer one of the most critical questions in mar-keting research: which instruments are making brands the most successful? This research elabo-rates on previous research that examined the ele-ments of the marketing mix that make brands the most successful. We attempt this by addressing the following research question: How effective are
advertising, distribution and price on the different classified product categories over het long run?
The focus of this research is established on the different marketing dynamics. To the best of our knowledge there is a lack of research about
exam-ining the long-term effectiveness of marketing on different product categories. More specifically, to what extent differ the effects advertising, distribu-tion and price within a particular product category. Hence, this implies that we attempt to analyze the differences in effectiveness among the product categories on the long-term.
The major aim of this research is to propose a framework which provides new insights about the accountability (i.e. effects) of advertising, distribu-tion and price on different product categories. Measuring the marketing effectiveness within the four classified product categories is useful to de-termine the allocation of the marketing budget for different kinds of product categories over time.
This study is based on 57 categories of fast-moving-consumers-goods (FMCG) scanner data, which is sampled weekly over a 4-year period in The Netherlands. We use the 3 most successful brands (based on market share and volume of sales) and classify the different types of products in the four product categories. Besides this, we employ a 5% market share threshold for reliability purposes. We apply a VARX model that measures the impact of advertising, distribution and price over the long run (i.e. 13 weeks). We capture the cumulative differences in performance of all mar-keting instruments among the different product categories.
Sales Marketing instruments:
Advertising Distribution Price
Hedonic products vs. Utilitarian products Low-involvement products vs.
High-involvement products results into building an underlying framework that
gives guidance in this research. Second, we de-scribe the data set and the methodology. Third, we present the empirical results. Last, we discuss the results. the managerial implications. Limitations and formulate recommendations for future re-search.
2. Research framework
This research is focused on the effect of advertis-ing, distribution and price on sales within different product categories over time. Therefore, we first briefly discuss the effects of advertising, distribu-tion and price in general. Second, we address the marketing effects per product category and use this literature foundation to develop predictions. Figure 1 provides an overview of the conceptual framework that highlights the aspects of this re-search.
The conceptual framework attempts to provide marketing practitioners new insights to assess the long-term accountability of advertising, distribu-tion and price on: 1) low- or high- involvement products and 2) hedonic or utilitarian products.
Figure 1: Conceptual framework
2.1 Main effects of price, advertising and dis-tribution on sales
Advertising - Dekimpe and Hanssens
the effects of advertising. However, the focus of this research is to examine the effects on the dif-ferent product over the long run. Previous findings show that the impact of advertising consists of an immediate impact on sales (Leone and Schultz, 1980). In contrary, other research shows that ad-vertising consists of carry-over effects (Givon and Horsky, 1990). Furthermore, Dekimpe and Hanssens (1995) found that the effects of advertis-ing on sales are persistent within one year. Even though there is found a persistent advertising ef-fect within one year, advertising appears to con-sists of a wear-out effect . This implies that adver-tising has a decreasing effect over time (Risselada et al., 2014). Hence, overall we expect a positive advertising effect on sales which has a relatively lower effect since we capture the advertising ef-fects over the long run.
Distribution – Contrary to e.g.
advertis-ing, there are only a few studies devoted to inves-tigating the effects of distribution. One of the main reasons is the difficulty of obtaining distribution data. Tellis (1988) argues that distribution cover-age data is difficult to acquire since there is a lack of good measures. However, the research of Ata-man et al. (2010) was one of the first empirical studies that addressed the effects of distribution on brand sales over the long run. Distribution cover-age in this research is described as the extent to which distribution carries a brand. Research shows that distribution affects brand performance posi-tively. Availability of products in stores are main-ly important since it accelerates and eases the abil-ity of finding a particular product or brand
(Bronnenberg et al., 2000). Besides this, a broader distribution results in brand loyalty (Farley, 1964). It appears that brand loyalty depends on the ease of finding a particular brand (Carman, 1970). Hence, the more loyal the consumers get, the more likely they are to buy consistently the same brand. Market shares of FMCG tend to increase due to distribution (Farris et al., 1989). These findings suggest that distribution has a positive effect on sales over the long run.
Price - The effects of price is one of the
de-crease in sales. These empirical findings suggest that companies have to be careful by raising pric-es. In general, we expect that price will have a negative effect on sales.
Concluding, many studies examine the separate effects and determinants of the different marketing instruments. The study of Ataman et al. (2010) is seen as one of the first studies that includes all marketing instruments in one framework. They studied the effects of all marketing instruments simultaneously over a large number of categories over the long run. One of the main findings of their study is that product and distribution (place) have a relatively larger effect on brand sales than advertising (promotion) or discounting (price). We attempt to build a comparable framework in which the effects of advertising, distribution and price are examined simultaneously among different product categories over the long run. Consistent with Bronnenberg et al. (2006), we quantify the long run as one quarter of the business cycle (i.e. 13 weeks). Beyond these general effects of the marketing instruments, this conceptual framework will further elaborate on the marketing effects among the different product categories in order to develop predictions.
2.2 Low- versus high-involvement products
Involvement is studied in many different settings and contexts. Consequently, prior research shows there is not a single definition for the term in-volvement. Zaichkowsky (1986) presents a review paper that outlines previous studies and the history of involvement, concluding with the fact that the
definition of involvement fits in a certain domain. We define involvement as: the relevance of a product that matches a consumers’ needs and val-ues. Involvement consists of differences in stimuli: low- and high-involvement. There are two under-lying components in order to determine whether products are considered as low- or high-involved: 1) personal needs and 2) to which extent a product differentiates itself from other products (Zaichkowsky, 1986).
high-involvement products as products that emphasizes the motivation and ability of issue-relevant think-ing (i.e. through the central route) and 2) low-involvement products as products that are evaluat-ed basevaluat-ed on simple heuristics (i.e. processevaluat-ed via the peripheral route).
Marketing instruments can be employed in order to anticipate on the level of involvement (Korgaonkar & Moschis, 1982). Specifically, re-sults show that there are significant differences in consumer evaluations of low- and high-involvement products (Huettl and Gierl, 2012). Therefore, the promotional strategy should be di-vergent. One of the main propositions for low-involvement products is that consumers are not interested to process much information (Fennis, 2016; Korgoankar and Moschis, 1982), which is in line with the peripheral route of the ELM (Petty & Cacioppo, 1984). Therefore, short messages that highlight product benefits will be sufficient. In contrast to high-involvement products in which consumers feel the need to gather information. Therefore, intended actions should contain the information that a customer needs before purchas-ing a product. High-involved consumers are highly motivated to process all kinds of information, compare and evaluate products (Fennis, 2016). We further discuss the expectations of the marketing effects on the high- and low-involvement prod-ucts.
2.2.1 Effect of advertising on high- versus low-involvement products
Swinyard and Coney (1978) provided evidence that advertising can influence low-involvement behavior more easily than high-involvement be-havior. The authors state that low-involvement consumers benefit greatly from different kinds of advertising, whereas high-involvement consumers tend to protect themselves from influential promo-tions and search for objective information (Swinyard & Coney, 1978). Fennis (2016) states that advertising is mostly used to create awareness for the product or brand. Translated to the ELM advertising could have more impact on people that process information via the peripheral route than the central route, since creating awareness is often regarded to processing information via the periph-eral route (Fennis, 2016). Howard and Kerin (2006) also show that high-involved consumers are motivated to explore information more fully, whereas less involved consumers are likely to ap-ply simple heuristics. As such, the advertising ef-fectiveness is expected to be higher among low-involvement products.
2.2.2 Effect of distribution on high- versus low-involvement products
consumer is willing to follow the central route of persuasion, implying it concerns a high-involvement product. Low-high-involvement products are selected based on heuristics, so we opine that the probability that the importance of availability of a low-involvement product is less likely an is-sue. In other words, we expect that the distribution breadth is less important for low-involvement products. Low-involvement products involve low information processing (i.e. information pro-cessing via the peripheral route, Petty & Cacioppo), hence we expect that it is easier to switch products if a particular product is out of stock. In contrary, consumers that are looking for a high-involvement product (i.e. in the central route of processing information) are more moti-vated and able to find the particular product, hence the availability of that particular product is in all probability of greater importance.
2.2.3 Effect of price on high- versus low-involvement products
High-involvement products are more likely to be evaluated on the product itself rather than focusing on the price. Divine (1995) proposes a framework in which he shows how the variation of price and quality relates to involvement. His findings sug-gest that high-involvement consumers are more likely to use performance as determinant than price. In contrast to low-involvement consumers, that are more likely to use price as the decisive factor. Lichtenstein et al. (1988) found that there is a relationship between product involvement and price consciousness. It appears that high-involvement consumers prefer to use objective
quality cues rather than price. This is in line with the ELM that proposes that people following the central route are more likely to process more in-formation to obtain more product specific knowledge (Petty & Cacioppo, 1984). Hence, price should have a greater impact on low-involvement products relative to high-low-involvement products.
2.3 Hedonic versus utilitarian products
However, consumers tend to choose utilitarian products over the hedonic products when both products are presented abreast. This finding relates with the guilty feeling associated with buying he-donic goods instead of utilitarian goods (Khan and Dhar, 2006; Okada, 2005). Khan and Dhar (2006) define utilitarian products as basic needs whereas hedonic products are desired for fun. Therefore, the decision to buy hedonic goods is associated with guilt which results into a greater need for justification.
2.3.1 Effect of advertising on hedonic versus utilitarian products
Consumer responses and consumer behavior de-pend mainly on the benefits that a product, service or an idea contains (Fennis, 2016; Okada, 2005). Specifically, hedonic or utilitarian products offer different kinds of benefits, the first in the type of fun and enjoyment whereas the latter in a more functional manner (Mano & Oliver, 1993). There-fore, consumers tend to feel guilty when they de-cide to purchase hedonic products over utilitarian products, since the benefits are more difficult to justify. The main purpose of advertising is to in-form or persuade consumers concerning ideas, products or services (Tellis, 2004; Fennis, 2016). Hence, advertising could contribute to reduce guilt through persuading consumers in such a way that they can justify their behavior. Huhmann and Botherton (1997) state that persuasive information reduces the feeling of guilt, which implies that advertising is able to justify behavior. Therefore, it is plausible that the advertising effectiveness
within the hedonic product category is greater compared to the utilitarian product category.
2.3.2 Effect of distribution on hedonic versus utilitarian products
The broader the distribution coverage of a particu-lar product, the easier the consumers are able to find it in store (Ataman et al. 2010). Considering the fact that there has been little emphasis on dis-tribution among the product categories, we predi-cate upon the theory on utilitarian and hedonic products solely. In terms of purchasing, previous literature shows that consumers tend to pick utili-tarian products over hedonic products, since it is more easy to justify this decision (Okada, 2005). In particular, if both product types are presented close together, the consumers prefer the basic ne-cessities (i.e. utilitarian products) instead of prod-ucts that are hedonic in nature (Khan and Dhar, 2000; Okada, 2005). In all probability, utilitarian products are more often sold, since consumers tend to spend more easily money on utilitarian products. Therefore, we expect that distribution coverage, and hence availability of a product, is more important for the utilitarian product category compared to the hedonic product category.
2.3.3 Effect of price on hedonic versus utilitari-an products
show that spending money on utilitarian goods is easier to justify relative to spending money on hedonic products. Hence, higher prices for hedon-ic products makes it more diffhedon-icult to justify be-havior. Therefore, we expect that price is a more important determinant for hedonic products rela-tive to utilitarian products. This is in line with the research of Khan and Dhar (2010) that examined price discounts based on product bundles that are differentiated by the utilitarian and hedonic nature of the items. Results show that savings are more effective on hedonic products than utilitarian products. Framing the discount on hedonic prod-ucts serves as justification that reduces the guilty feeling consumers obtain by acquiring hedonic products, which is consistent with the research of Huettl and Gierl (2012). Hence, we assume that price has a greater impact on hedonic products than on utilitarian products.
3. Data and methodology 3.1 Data description
Data are available on 57 different categories in The Netherlands. These FMCG-categories correspond with IRI’s classification of the various product types and cover goods in the total Dutch market including the three largest su-permarket chains. The data consists of four-year weekly scanner data from 1994 through 1998. For each product category we select the three most successful brands, based on volume of sales and market shares. Besides this, the included brands are required to have at least a 5% market share and advertise at least once over the observation period.
Table 1: Overview product categories
The total sample includes 116 brands in 50 differ-ent kinds of FMCG-categories. These categories cover a broad range of food, beverages, household and personal care products. Our data set arises from two sources. First, the FMCG-data set from The Netherlands, which contains information about advertising expenditures, price ranges, the breadth of distribution and the volume of sales. Second, involvement and product type (i.e. hedon-ic/utilitarian) rankings of 131 different FMCG- categories were covered in a complementary da-taset. The hedonic/utilitarian variables are meas-ured as a dummy variable (0/1) whereas the level of involvement is measured on a scale from one to five. These data are solely used to complement the Dutch FMCG data in order to classify the product types into the low-, or high-involvement product category and the hedonic or utilitarian product category. The combination of these two datasets resulted in the opportunity of investigating the effects of advertising, distribution and price on the
Product cate-gory Number of categories Example of categories Number of brands Hedonic 21 Candy bars
Moisturizer 46
four classified product categories. Table 1 presents an overview of the classification of all 116 brands into the four product categories.
3.2 Methodology
To assess the effects of advertising, distribution and price on the four product categories in a dy-namic market, one should evaluate several issues (van Heerde et al., 2004). We focus especially on stationarity, and address missing data, collinearity and cross-sectional heterogeneity more briefly. Subsequently, these issues will be discussed. In order to derive the effects of marketing on the dif-ferent product categories over the long run we should capture the cumulative effects of all varia-bles over one quarter (i.e. 13 weeks). Comparable to the research of Dekimpe and Hanssens (1999) we operationalize all four variables in a VARX model at the brand level. VAR models are typical-ly used to examine time series allowing the inclu-sion of multiple variables. The structure of this VARX model is that every single variable is a linear function of their past lag(s) and the past lag(s) of other variables accounting for a determin-istic term. Specifically, we include advertising, distribution ,price and sales as endogenous varia-bles. All variables are expressed in log-transformation for scaling purposes and compara-bility (Deleersnyder et al., 2004).
In order to represent the data more decently (Zivot and Wang, 2006) we included a deterministic time trend as exogenous variable. The available data covers four-year of weekly FMCG data that con-sists of seasonal influences which are decomposed
by using the deterministic time trend that does not allow for any randomness. Modeling a VARX model is perfectly suitable for analyzing patterns (different effects) and dynamic behavior of time series data (Zivot and Wang, 2006), which is one of the main interests of this research.
One of the most important next steps is to deter-mine whether the data shows a unit root or can be considered stationary. Stationarity is highly im-portant since nonstationarity could result in unre-lated significant results. The stationary test is based on panel unit root tests, since individual unit root tests are limited in their power (Levin and Lin, Chu,1992; Baltagi, 2007). The Im Pesaran and ADF Fisher Chi square are used to examine the stationarity among heterogeneous groups. Hence, the stationarity of the log-transformed en-dogenous variables are determined by analyzing the Im Pesaran and Shin and ADF Fisher Chi-square panel unit-root tests. The W-stat tests the null hypothesis, implying there is unit root. All four series are highly significant at the 1% level (presented in Appendix A), which means that the unit root hypothesis is rejected. The four series are considered stationary, which means that the model can be specified in levels. The following VARX model is specified as:
With
lnSalesbp,t = Sales of brand b in product category p at time t
lnAdvertisingbp,t = Advertising of brand b in product category p at time t
lnDistributionbp,t = Distribution of brand b in product category p at time t
lnPricebp,t = Price of brand b in product category p at time t
δbp = Intercepts vector of brand b in product cate-gory p
Αbp,i = Coefficients matrix of brand b in product category p at lag i
τbp = Coefficient of the deterministic, exogenous trend variable for brand b in product category p εbp,t = Error vector for brand b in product category p at time t
The specified model shows that each single varia-ble is a linear function of the past lags of that vari-able and the past lags of other varivari-ables. Futhermore, we included the intercept and trend as exogenous deterministic components. The sub-script t-i refers to the optimal number of lags. The optimal number of lags are assessed based on the Schwarz criterion (BIC). The most appropriate number of lags are determined for each single brand. For 97% of all brands one lag fits the model the best.
Previous research indicated the key challenges of analyzing time series data (van Heerde et al., 2004). Besides determining stationarity of the
data, it is also highly important to examine the presence of missing data. To obtain reliable results we propose that the three most successful brands at least have an average market share of 5% over the observation period. Furthermore, all brands advertise at least once. In terms of missing data, we analyzed the data beforehand and discovered that 6.8% of all brands lack sales a certain period. However, the 5% market share threshold elimi-nates the brands that appeared to have missing sales data. Hence, the market share threshold solves this missing data issue.
Another issue that we address is (mul-ti)collinearity. In order to avoid near singular ma-trices we analyzed whether there is a chance of detecting (multi)colliniearity. However, the esti-mates are in 97% of the cases consisting of 1 lag. Hence, we expect that the chance that (mul-ti)collinearity emerges is small. Correlation tables provide insights about the detection of multicollinearity (Asteriou and Hall, 2016). We estimated the correlation coefficients of all brands to avoid any collinearity in the data. A correlation coefficient < 0.9 is considered as the limit, beyond that limit (multi)collinearity will in all probability occur (Asteriou and Hall, 2016). Consistent with our expectations, we find that the estimated corre-lation tables show that there are no significant correlation coefficients larger than 0.9. Hence, we can conclude that there is no sign of multicollinearity.
across all brands and across the subsets of brands can be evaluated by deriving the Added Z method (Rosenthal, 1991). The Added Z method is used to evaluate all four product categories in terms of significance. The effects of all brands across high-involvement, low-high-involvement, hedonic and utili-tarian products are examined based on the 95% confidence level.
Generalized Impulse Response Function
In order to assess the effectiveness of advertising, distribution and price on the different product cat-egories over the long run we derive the general-ized impulse response functions (GIRFs). The GIRF differs from the Impulse Response Function (IRF) in terms of the shock to the residual. Instead of a 1% shock to the residual, the GIRF accounts for a 1 standard deviation shock to the residual. One of the main critiques of IRFs is the sensitivity of ordering. Pesaran and Shin (1998) argue that the orthogonalized IRF is susceptible to changes in ordering. However, the GIRF is invariant to order-ing of the included variables in the VARX model. Besides taking the orthogonalized assumption into account by applying the GIRF, the tool is adequate to measure and describe the effects of marketing over the long run.
Table 2: Marketing effectiveness on product categories
*: p < 0.01
For the long-term effects we apply the generalized shocks and trace the cumulative effects of the marketing instruments on sales over 13 weeks. A Monte Carlo simulation with 100 repetitions is used to test the significance of the differences be-tween the product categories.
4. Results
Table 2 reports the results of the effects of adver-tising, distribution and price on the product cate-gories over 13 weeks. The weighted average is calculated since the key element of this research is to determine the effectiveness i.e. the importance of each marketing instrument per product cate-gory. Besides, we find it important to note that the interpretation of the results is been done more carefully, since we derived our results from GIRFs. This implies that our results cannot be interpreted as elasticities, but instead as a one standard deviation shock in advertising results in a 0.040 change in sales in the low-involvement product category. However, in this section we re-fer to the relative effects and the effect sizes of the different marketing instruments on the product categories.
Overall effects Low-involvement High-involvement Hedonic Utilitarian
Advertising 0.040* 0.056* 0.031* 0.041* 0.041*
Distribution 0.116* 0.109* 0.120* 0.076* 0.141*
The added Z method is used to evaluate the signif-icance across the product categories and the mar-keting instruments (Rosenthal, 1991). The added Z method determines whether there are significant differences between groups. In order to provide generalizations among the heterogeneous esti-mates at the brand level, we examined the overall effects by applying the added Z method for each effect. The Z-scores are derived from the one- sided p-value per estimate. All z-scores are summed up and divided by the square root of the amount of brands (Gijsenberg, 2014; Rosenthal, 1991). The results, in Appendix C, show that all variables across the different product categories are significant (< 0.01). This implies that our find-ings suggest that there are significant differences between groups.
Overall effects
The overall effects show that price and distribution are relatively close to each other. Price has the greatest effect size of b= -0.118 (43.0%), but dif-fers not much of distribution b= 0.116 (42.3%). The effect size of advertising is the smallest, b= 0.040 (14.7%).
Low-involvement vs. high-involvement prod-ucts
The impact of advertising is the greatest among low-involvement products (b= 0.056, p < 0.01) relative to high-involvement products. The main difference of advertising of low-involvement and high-involvement products is 0.025. Table 3 pro-vides an overview of the differences per product
Table 3: Main differences marketing effects
category. A positive sign indicates that e.g. adver-tising and price have larger effect sizes on low-involvement products relative to high-low-involvement products. In contrary, the effects of distribution on the involvement category, has greater effect for high-involvement products than low-involvement products, which is explained by the negative sign (difference = -0.011). In line with our expecta-tions, results show that price has a greater impact on low-involvement products relative to high-involvement products (b= -0.136, p < 0.01). Over-all, price is relatively more important for low-involvement products (45%) compared to advertis-ing (19%) and distribution (36%). In contrary, we find that the impact of distribution (47%) is great-er than advgreat-ertising (12%) and price (41%) for high-involvement products. Hence, there are sig-nificant differences between the low- and high-involvement product categories: the low-involvement product category encounters the greatest impact of price whereas the high-involvement product category shows that distribu-tion has relatively more effect.
Hedonic vs. utilitarian products
Contrary to our expectations, we find that adver-tising has the same effect for both hedonic and utilitarian products (difference: 0.00, p < 0.01). For both product categories results show that ad-vertising has the smallest impact relative to the other instruments within the product categories. In line with our expectations, we find that distribu-tion has the greatest impact among utilitarian products (b= 0.141, p < 0.01) relative to hedonic products (b= 0.076, p < 0.01). Results show that price encounters the greatest impact on hedonic products ( b= -0.154, p < 0.01). This indicates that price is more of an influence in terms of effective-ness within the hedonic product category (57%). Advertising (15%) and distribution (28%) are less effective for the hedonic product category. How-ever, results show that distribution gains influence in the utilitarian product category (51%), relative to advertising (15%) and price (34%). Hence, in terms of marketing effectiveness products that are classified in the hedonic category should focus more on price, whereas the products classified in the utilitarian category should focus more on dis-tribution.
Combining product categories
This research is extended by examining the effects of the combined product categories. Specifically, based on the Foote, Cone and Belding Grid (Vaughn, 1980) we combined the low-involvement, high-low-involvement, hedonic and utili-tarian product categories into one grid. The FCB-grid captures the level of involvement with the think/feel aspects (Fennis, 2016). Basically, the
levels of involvement consists of low-involvement and high-involvement, whereas the think/feel axis are associated with the utilitarian and hedonic need for products. Specifically, thinking is associ-ated with utilitarian motives, while feeling is asso-ciated with more hedonic, sensory-pleasure mo-tives (Putrevu & Lord, 1994). Hence, this research introduces a grid that captures the low-involvement/high-involvement with its hedonic/ utilitarian axis. The FCB-grid includes the follow-ing four quadrants:
1) low-involvement/hedonic 2) low-involvement/utilitarian 3) high-involvement/hedonic 4) high-involvement/utilitarian.
Consumers indicated to what extent products are considered as low-involvement or high-involvement products and whether they define products as hedonic or utilitarian products. We classified each brand into one of the four quad-rants, a detailed overview is presented in Appen-dix B.
Table 4: Joint effects of all product categories
*: p < 0.01
First, we applied the added Z method in order to determine whether there are significant differences between the four quadrants. All four quadrants are significant at the 1% level, hence we can conclude there are differences between the quadrants. Table 4 summarizes the advertising, distribution and price effects on the different product categories. Results indicate that advertising has the least im-pact (b= 0.055, p < 0.01) on low-involvement/hedonic products (i.e. quadrant 1). Distribution has a relative low impact as well (b= 0.063, p < 0.01). Price is the most effective in-strument among quadrant 1 (b= -0.183, p < 0.01). Second, we assessed the effects of the new cate-gorized quadrants by combining the GIRFs ac-cording to the classification in Appendix B.
Results show that advertising has the smallest ef-fect on quadrant 1 (i.e. low-involvement/hedonic products) relative to distribution and price. Price has the greatest effect size (b= -0.183, p < 0.01). In contrary to quadrant 2, where distribution is found to be more important (b= 0.151, p < 0.01). We find that advertising has less impact on the
combined product categories in quadrant 3, rela-tive to distribution and price. Finally, the distribu-tion breadth for quadrant 4 is found to be relative-ly more effective (b= 0.136, p < 0.01). It appears that for all four quadrants advertising seems the least effective. The overall effects show that price encounters the greatest effect on hedonic products, irrespectively the products are combined with high- or low-involvement. The same is applicable for the importance of distribution. Specifically, it seems that distribution is more effective for prod-ucts that are utilitarian in nature, unheeded wheth-er the utilitarian product is low- or high-involved. The results are based on a period of 13 weeks (one quarter). Companies have to allocate their market-ing budgets often in the beginnmarket-ing of the year. This extended framework helps companies in de-termining the allocation of their marketing ex-penditures over the long run.
5. Discussion
This paper has investigated the effectiveness of advertising, distribution and price on different product categories over the long run. The results are obtained from analyzing advertising, distribu-
Low-involvement
High-involvement Hedonic Utilitarian
tion, price and sales data that included 116 brands of 50 FMCG-categories over an observation peri-od of four years. We structure this discussion sec-tion around the general conclusions, managerial implications and provide an overview of limita-tions and hence suggeslimita-tions for future research.
5.1 Conclusions
In general, the overall effects of advertising, dis-tribution and price over the long run are in line with expectations. Specifically, advertising has a positive, long-term effect on sales (e.g. Dekimpe and Hanssens, 1995, Givon and Horsky, 1980;Sethuraman, 1991), similarly as distribution that encounters a positive effects over the long run (e.g. Ataman et al., 2010; Farris et al., 1989). In contrary, the effectiveness of price is found to be negative which implies that price affects sales negatively (Bijmolt, 2005; Tellis, 1988).
Across all product categories, we find that the overall effect size of distribution is relatively large over the long run which is consistent with the find-ings of Ataman et al. (2010). However, we did not expect that the effect size of distribution (42.3%) would be extremely close to the effects size of price (43.0%). This finding suggest that distribu-tion, next to price, has become one of the most expressive instruments. As expected, the effect size of advertising (14.7%) is smaller than the ef-fect sizes of price and distribution.
Low- versus high-involvement products
The advertising differences between low- and high-involvement products are in line with our
expectations. Advertising is found to be more ef-fective in the low-involvement product category relative to the high-involvement product category. The impact of advertising is the most influential when consumers find themselves in a peripheral route of persuasion, and hence are low-involved (Petty and Cacioppo, 1984). Specifically, simple heuristic decisions are affected by the use of ad-vertising. These findings support the findings that consumers in the central route (high-involved) are protecting themselves and attempt to avoid any persuasive advertisements (Swinyard and Coney, 1978). Hence, it is more effective to focus on ad-vertising on products in the low-involvement cate-gory.
brands, which implies that consumers have the motivation and ability to compare across products and brands. Therefore, the distribution coverage is more effective among high-involvement products in comparison to low-involvement products.
Price is one of the most studied elements in eco-nomics. Hanssens (2009) argues that price had become a more important determinant for FMCG. Price has a long-term negative effect size, which is sensible since price elasticities are always consid-ered negative. Our findings suggest that the price instrument is relatively more effective for low-involvement products than high-low-involvement products. Hence, these findings support previous research that states that low-involvement products are evaluated on price, whereas high-involvement products are evaluated on performance (Divine, 1995).
Combining these marketing effectiveness differ-ences among the product categories, we can state that products classified in the low-involvement category should focus on especially price and on advertising, whereas products classified in the high-involvement category should focus on distri-bution.
Hedonic versus utilitarian products
In contrast to our expectations, we find that adver-tising has exactly the same effect sizes for both hedonic and utilitarian products. We proposed that advertising could help to reduce the feeling of guilt, which is associated with purchasing hedonic products. However, our findings suggest that there
are no differences between the two product cate-gories.
Consistent with our expectations we find that dis-tribution has a relatively larger effects among utili-tarian products compared to hedonic products. This result supports our assumption, in which utilitarian products are preferred over hedonic products and hence are more often sold. Therefore, the availability of utilitarian products seems more important than for hedonic products.
Combining product categories into the FCB-GRID
Our research is extended by adding the Foote Cone and Belding grid in order to provide some in-depth insights. Specifically, this research at-tempts to examine the accountability through ana-lyzing the effects of advertising, distribution and price on different product categories over the long run. The results show that there are significant differences between the different product catego-ries and the marketing instruments. Hence, we are able to interpret those differences and analyze which instrument is more or less effective per product category. However, the different product categories (i.e. low- or high-involvement and he-donic or utilitarian) are not related in such a way that a brand belongs in one of those two catego-ries. In other words, a low-involvement product can be hedonic in nature. Therefore, we introduce the FCB-grid that combines all possible combina-tions in order to create a marketing strategy.
Consistent with the separate expectations of price on low-involvement and hedonic products, price is also found to be more effective when we combine the two product categories. Low-involvement/hedonic products are highly focused on evaluating price instead of performance and uses price as starting point to justify behavior. Advertising has the largest effect on low-involvement/utilitarian products compared to the other quadrants. Our findings show that advertis-ing has more effect on low-involvement/utilitarian products in comparison to low-involvement/hedonic products. However, based on
theory we expected that the effect size would be larger on the low-involvement/hedonic product category. Finally, distribution is successful operat-ing in low-involvement/utilitarian product catego-ries.
5.2 Managerial implications
Previous research shows that it is hard to address the accountability of marketing. Therefore, com-panies have a challenging task to allocate their marketing budget. This research attempts to create insights of the effects of different marketing in-struments on different product categories over the long run. Especially, the focus of this research is to propose a framework that takes dynamics into account. The effects of marketing are not calculat-ed as direct effects, which is one of the main rea-sons why companies find it hard to determine the accountability. Therefore, this research offers marketing practitioners a deeper understanding of the marketing effects over the long run. The mar-keting effects on different product categories over the long run are helpful in determining the alloca-tion of the marketing budget. The findings of this study should inform managers that there are dif-ferent effects among difdif-ferent product categories categories over time.
that there are significant differences between the marketing effectiveness and the product categories in the FCB-grid.
More specifically, price has a relatively larger effect on low-involvement/hedonic products than on low-involvement/utilitarian products. In contra-ry, distribution is relatively less effective among low-involvement/hedonic products, just as
adver-tising in comparison to
low-involvement/utilitarian products. These findings imply that products classified into the low-involvement/hedonic category should particularly focus on the price instrument and focus less on advertising and price. The relative distribution effects are found to be larger on the low-involvement/utilitarian product category, whereas the effects sizes of advertising and price are small-er. Hence, findings indicate that companies should focus more on distribution among low-involvement/hedonic products.
The high-involvement/hedonic category encoun-ters relatively more price effects than the high-involvement/utilitarian product category. Howev-er, high-involvement/utilitarian products have a greater effect size on distribution. The advertising effects sizes do not differ much within the both categories. Companies that market high-involvement/hedonic products should focus more on price relative to advertising and distribution, whereas high-involvement/utilitarian products should focus particularly on distribution and less on advertising and price. In table 5 we present a framework that consists of the four combined
Table 5: The major focus on instruments per category
product categories and the marketing instruments advertising, distribution and price. The star indi-cates on which instrument a particular product category should focus on.
Our research shows that it is important to use clas-sification of products in order to determine an ef-fective marketing strategy. Allocating marketing budgets will become more justifiable (i.e. increas-es accountability) over the long run by classifying products into product categories and focus on the corresponding, effective marketing instrument(s).
5.3 Limitations and Future Research
This study provides an overview of the effects of the different marketing instruments on the classi-fied product categories over the long run. How-ever, there are limitations that offer opportunities for future research. First, the data consists of FMCG scanner data, which can be too restrictive. Future research could expand this research by ex-amining the effects of marketing on different product categories in other setting settings, e.g. for durables or services (Deleersnyder, 2004). Second, our research is based on a dataset from the
lands. Hence, it remains unclear whether our re-search can be generalized across other countries. Third, the data was limited on advertising, distri-bution and price. The inclusion of product infor-mation could provide a more complete analysis of the effectiveness of marketing. Moreover, this would provide insights into the effectiveness of the traditional marketing mix instruments (i.e. the 4P classification; McCarthy, 1960). Fourth, this research focuses on the top three brands based on the volume of sales within each product category. Therefore, future work could examine whether these findings also hold by including the effects of smaller brands. Fifth, we applied a VARX model and derived the generalized impulse responses. In recent years the vector autoregressive models have become more popular in empirical research, and thus has been more criticized. The aim of a VAR model is to estimate multivariate time series, by estimating the system of equations simultaneously. Each specified equation in a system is based on the identification of determining the endogenous and exogenous variables. However, Sims (1980) states that exogenous variables are often included by default instead of including them as a result of good reasoning (and hence based on findings) to believe that the variables are strictly exogenous. Besides this, Sims argues that there are doubts about interpreting shocks. The results of a VAR are modelled by an impulse, thus a shock. How-ever, researchers show in different settings that it is hard to conclude what a shock precisely meas-ures (Sims, 1992; Bernanke and Mihov, 1996) . Our findings are based on the generalized impulse response function, which implies that there is a
one standard deviation shock. According to Sims (1980;1992) it is questionable if these shocks truly measure the relevant part, since it is difficult to quantify the shock. Finally, we applied the GIRFs that measures a standard deviation shock in the response variable. However, we are therefore un-able to interpret the estimates as elasticities. Fu-ture research could expand our work by examining the effects based on a 1% shock, which means that the estimates can be interpreted as elasticities.
References
Asteriou, D. & Hall, S.G. (2016), Applied econo-metrics. Palgrave Macmillan.
Ataman, M.B., Van Heerde, H.J., & Mela, C.F. (2010), The long-term effect of marketing strategy on brand sales. Journal of Marketing Research, 47(5), 866-882.
Baltagi, B.H. (2007), Panel Unit Root Tests and Spatial Dependence. Journal of Applied Econometrics, 22(2), 339-360.
Bernanke, B.S & Mihov, I. (1996), What Does the Bundesbank Target? Economic Review.
Bijmolt, T. H.A., van Heerde, H.J. & Pieters, R.G.M. (2005), New Empirical Generalizations on the De-terminants of Price Elasticity. Journal of Marketing Re-search, 42 (2), 141–56.
Bronnenberg, B. J., Vijan, M. &Vanhonacker, W.R. (2000), The Emergence of Market Structure in New
RepeatPurchase Categories: The Interplay of Market Share and Retailer Distribution, Journal of Marketing Research, 37, 16–31.
Bronnenberg, B. J., Mela, C.F. &Boulding, W. (2006), The Periodicity of Pricing. Journal of Marketing Research, 43 (3), 477–93.
Carman, J.M. (1970), Correlates of Brand Loyalty: Some Positive Results. Journal of Marketing Research, 7(1), 67-76.
Dekimpe, M.G. &.Hanssens, D.M. (1995), The Per-sistence of Marketing Effects on Sales. Marketing Science,
14(1), 1-21.
Dekimpe, M.G. & Hanssens, D.M. (1999), Long-run effects of price promotions in scanner markets. Journal of econometrics, 89 (1), 269-291.
Dekimpe,, M.G. and Hanssens, D.M. (1999), Sus-tained Spending and Persistent Response: A New Look at Long-Term MarketingProfitability. Journal of Marketing Research, 36( 4), 397-412.
Deleersnyder, B., Geyskens, I., Gielens, K. & De-kimpe, M.G. (2002), How Cannibalistic Is the Internet Channel? A Study of the Newspaper Industry in the United Kingdom and The Netherlands. International Journal of Research in Marketing, 19 (4), 337–383.
Deleersnyder, B, Dekimpe, M.G., Savary, M. & Parker, P.M. (2004), Decomposing the Sales Promotion Bump with Store Data. Quantitative Marketing, 2(4), 347-383.
Dhar, R. & Wertenbroch, K. (2000), Consumer Choice Between Hedonic and Utilitarian Goods. Journal of Marketing Research, 60-71.
Divine, R.L.(1995), The Influence of Price on the Relationship between Involvement and Consideration Set Size. Marketing Letters, 6(4), 309-319.
Farley, R. (1964), A Communications Theory of Urban Growth. American Journal of Sociology, 69 (6), 669-670.
Farris, P.W., Olver, J. & De Kluyver, C. (1989),The Relationship Between Distribution and Market Share. Mar-keting Science, 8 (2), 107–28.
Fennis, B. & Stroebe, W. (2016), The psychology of Advertising. Routledge.
Gielens, G.I.K. and Dekimpe, M.G. (2002), The Market Valuation of Internet Channel Additions. Journal of Marketing, 66 (2), 102–19.
Gijsenberg, M.J. (2014), Going for gold: Investigat-ing the (non)sense of increased advertisInvestigat-ing around major sports events. International Journal of Research in Market-ing, 31, 2-15.
Givon, M. & Horsky,D. (1990), Untangling the Ef-fects of Purchase Reinforcement and Advertising Carryover. Marketing Science, 9(2), 171-187.
Hanssens, D.M. (2009), Empirical generalizations about marketing impact : what we have learned from aca-demic research. Marketing Science Institute.
Harvey, M.G., Lusch, R.F. & Cavarkapa, B. (1996), A Marketing Mix for the 21st Century. Journal of Market-ing, 4, 1-15.
Heerde, H.J., Leeflang, P.S.H.& Wittink, D.R. (2004), Decomposing the Sales Promotion Bump with Store Data. Market Science, 23(3), 317-334.
Heerde van, H. J., Gijsenberg, M. J., Dekimpe, M. G., & Steenkamp, J. B. E. (2013), Price and Advertising Effectiveness Over the Business Cycle. Journal of Market-ing Research, 50(2), 177-193.
Howard, D.J. & Kerin, R.A. (2006), Broadening the Scope of Reference Price Advertising Research: A Field Study of Consumer Shopping Involvement. Journal of Mar-keting, 70( 4), 185-204.
Hsee, C.K., Yu, F., Zhang, J. & Zhang, Y. (2003), Medium Maximization. Journal of Consumer Research, 30(1), 1-14.
Huettl, V. & Gierl, H. (2012), Visual art in advertis-ing: The effects of utilitarian vs. hedonic product positioning and price information. Marketing Letters, 23(3), 893-904.
Huhmann B.A.& Botherton T.P., 1997. A content analysis of guilt appeals in popular magazine advertise-ments. Journal of Advertising, 26, 35-46.
Khan, U. & Dhar, R. (2000), Consumer Choice be-tween Hedonic and Utilitarian Goods. Journal of Marketing, 43(2), 259-266.
Khan, U. & Dhar, R. (2006), Licensing Effect in Consumer Choice. Journal of Marketing, 37(1), 60-71.
Khan, U. & Dhar, R. (2010), Price-Framing Effects on the Purchase of Hedonic and Utilitarian Bundles. Journal of Marketing Research, 47(6), 1090-1099.
Leone, R.P and Schultz, R.L. (1980), A study of marketing generalizations. Journal of Marketing, 44(1), 10-18.
Leone, R. P. (1995), Generalizing What Is Known About Temporal Aggregation and Advertising Carryover. Marketing Science, 14 (3), 141–150.
Levin , A. and Lin, C. (1992), Unit Root Tests in Panel Data: Asymptotic and Finite Sample Properties. Uni-versity of Cambridge.
Lichtenstein, D.R., Bloch, P.H. & Black, W.C. (1988), Correlates of Price Acceptability. Journal of Con-sumer Research, 15(2), 243-252.
Mano, H. & Oliver, R.L. (1993), Assessing the Di-mensionality and Structure of the Consumption Experience: Evaluation, Feeling, and Satisfaction. Journal of Consumer Research, 20(3), 451-466.
Mazumdar, Tridib, S.P. Raj, and Indrajit Sinha (2005), Reference Price Research: Review and Propositions. Journal of Marketing, 69 (4), 84–102.
Okada, E.M. (2005), Justification Effects on Con-sumer Choice of Hedonic and Utilitarian Goods. Journal of Marketing Research, 42(1), 43-53.
Pesaran, H.H. & Shin, Y. (1998), Generalized Im-pulse Response Analysis in Linear Multivariate Models. Economic Letters, 58(1), 17-29.
Petty, R.E. & Cacioppo, J.T. (1984), The Elabora-tion Likelihood Model of Persuasion. Advances in Consumer Research, 11, 673-675.
Prelec, D. & Loewenstein, G. (1991), Negative Time Preference. The American Economic Review, 81(2), 347-352.
Putrevu, S. & Lord, K.R.(1994), Comparative and Noncomparative Advertising: Attitudinal Effects under Cognitive and Affective Involvement Conditions. Journal of Advertising, 23 (2), 77-91
Risselada, H., Verhoef, P. C., & Bijmolt, T. H. A. (2014). Dynamic effects of social influence and direct mar-keting on the adoption of high-technology products. Journal of Marketing, 78(2), 52-68.
Rosenthal, R. (1991), Meta-Analytic Procedures for Social Research. Newbury Park: Sage.
Rust, R.T, Ambler, T. Carpenter, G.S., Kumar, V. & Srivastava, R. K. 2004. Measuring Marketing Productivi-ty: Current Knowledge and Future Directions. Journal of Marketing, 68, 76-89.
Sethuraman, R., Srinivasan, S. & Doyle , K.(1999), Asymmetric and Neighborhood Cross-Price Effects: Some Empirical Generalizations. Marketing Science 18 (1), 23–41.
Sethuraman, R. & Tellis, G.J. (1991), An Analysis of theTradeoff Between Advertising and Pricing. Journal of Marketing Research, 31 (2), 160–74.
Sims, C.A.(1980), Macroeconomics and Reality. Econometrica, 48(1), 1-48.
Sims, C.A. (1992), Interpreting the Macroeconomic Time Series Facts: The Effect of Monetary Policy. Europe-an Economic Review, 36(5), 975-1000
Srinivasan, S., & Hanssens, D. M. (2009), Market-ing and Firm Value: Metrics, Methods, FindMarket-ings, and Future Directions. Journal of marketing research, 46(3), 293-312.
Strahilevitz, M. & Myers, J.G. (1998), Donations to Charity as Purchase Incentives: How Well They Work May Depend on What You Are Trying to Sell. Journal of Con-sumer Research, 24 (4), 434-446.
Swinyard, W.R. & Coney, K.A. (1978), Promotion-al Effects on a High-versus Low-Involvement Electorate. Journal of Consumer Research, 5(1), 41-48.
Tellis, G.J. 1988. The Price Elasticity of Selective Demand: A Meta-Analysis of Econometric Models of Sales. Journal of Marketing Research, Vol. 25, No. 4 (Nov., 1988), pp. 331-341.
Tellis, G. J. (2003), Effective Advertising: How, When and Why Advertising Works. Thousand Oaks, Calif.: Sage Publications
Torres, I.M. & Briggs, E. (2007), Identification Ef-fects on Advertising Response: The Moderating Role of Involvement. Journal of Advertising, 36( 3), 97-108
Vakratsas,Demetrios,and Tim Ambler (1999), How Advertising Works: What Do We Really Know? Journal of Marketing, 6 (1), 26–43.
Waterschoot, W. & van den Bulte, C. (1992) The 4P Classification of the Marketing Mix Revisited. Journal of Marketing, 56, 83-93.
Zaichkowsky, J.L. (1986), Conceptualizing In-volvement. Journal of Advertising, 15(2), 4-14.
Appendix A – Results of Panel Unit Root Tests
Im Pesaran and Shin Test
ADF-Fisher Chi-square Test
Appendix B – FCB-grid classification
Hedonic products Utilitarian products
Category Description Brands Category Description Brands
Low-Involvement Category 32 Category 90 Category 118 Category 133 Category 134 Category 161 Category 180 Category 196 Category 207 Category 388 Category 390 Category 419 Soup Sauces Chocolate Candybar Chocolate bars Gum Gingerbread Sweet snack Peanuts Fries Mashed pota-to Hair products 16/17/18 41/42 53/54 61/62/63 64/65 67/68 71 73/74/75 85/86/87 113 116 130/131 Category 34 Category 49 Category 54 Category 72 Category 260 Category 482 Category 485 Category 487 Category 492 Category 493 Category 533 Category 535 Category 536 Instant soup Ketchup Salad dressing Sugar Frozen meal Toilet paper Tissues Main wash Dishwashing liquid Diswasher detergent Coke Fanta Icetea 19/20 25/27 31/32 37 109/111 146/147 149/150 151/152/153 154/155/156 157/158 160/161/162 163/164/165 166/167/168 High-Involvement Category 24 Category 26 Category 52 Category 122 Category 201 Category 413 Category415 Category 430 Category 542 Rice Spaghetti Mustard Cereal Crisps Shampoo Conditioner Moisturizer Apple juice 12 13/15 28/29/30 58/59 79/70/81 124/125/126 127/128/129 136/137/138 170 Category 0 Category 5 Category 48 Category 66 Category 94 Category 97 Category 102 Category 218 Category 222 Category 227 Category 231 Category 241 Category 246 Category 422 Category 437 Category 442 Milk Olive oil Gravy Cat food Ground coffee Instant coffee Tea
Appendix C- Results and calculation of Added Z method
Deriving the significance level of the added Z method creates insights concerning differences between groups. The T-statistic of every variable at the brand level is calculated by the estimate/standard devia-tion. The p-value is used to calculate the z-scores. The following formula is applied:
Besides determining whether groups differ from each other, we derived also the overall effects to exam-ine if there are also significant effects overall.
Overall Low-involvement High-involvement Hedonic Utilitarian
Advertising 5E-16 1E-10 8E-07 4.6E-08 4E-09
Distribution 0.0004 7E-32 3E-85 4E-92 1E-100
Price 1E-06 5E-64 1E-60 1E-19 2E-43
Added Z scores for the combined product categories based on the FCB-grid.
