DETERMINANTS OF A
SUCCESSFUL PRICE PROMOTION
Taking the perspective of the manufacturer
Determinants of a successful price promotion
Taking the perspective of the manufacturer
June, 2018
Master thesis
Msc. Marketing Management and Intelligence
University of Groningen
Faculty of Business and Economics
Department of Marketing
Marieke Szymanski | S2346672
Prinsenstraat 9a
9711CL Groningen
+31622113035
m.szymanski@student.rug.nl
Management Summary
Price promotions are active at large scales but are a risk for the health of brands, since reference prices are decreasing due to the frequency of promotions initiated by the retailers. The power of retailers over manufacturers, makes it difficult for manufacturers to stay away from price promotions. This research aims at discovering determinants for a successful price promotion, taking the perspective of the manufacturer.
Potential effects relating to promotional variables, price framing, competition and weather are discussed and tested on three brands within the same category. Research is performed using multiplicative regression which resulted in interesting results.
Findings show the influence of promotional variables, price framing and temperature on the success of a price promotion. These effects however, vary per brand which could be explained by the loyalty of the different buyer segments. This is something to bear in mind for future research next to the relative small size of data on which this thesis is based.
Table of Content
Management Summary 2
1. Introduction 5
1.1 Practical Relevance 5
1.2 Research Question 7
1.3 Structure of this Thesis 7
2. Theoretical Framework 8
2.1 Promotional variables 10
2.1.1. Display and Feature Advertising 10 2.1.2. Effects of price promotions 11
2.2 Types of Price Promotions 12
2.3 Competition 14
2.3.1. Asymmetric Competition 14 2.3.2. Competition and timing 14
2.3.3. Brand switching 15
2.4 The Influence of Weather 16
2.4.1 Influence of Temperature 17 2.5 Control Variables 18 2.6 Conceptual model 19 3. Research design 20 3.1 Data 20 3.1.1. Dummy variables 23 3.2 Research method 23 3.3 Model Specification 24 4. Results 25 4.1 Descriptive statistics 25
4.2 Assumptions for testing a linear model 26
4.3 Model estimation 29 Brand 1 29 Brand 2 31 Brand 3 32 5. Conclusion 33 5.1 Discussion on Brand 1 33 5.2 Discussion on Brand 2 34 5.3 Discussion on Brand 3 35 5.4 Overall discussion 36 5.5 Managerial implications 38
1. Introduction
Many manufacturers of fast-moving consumer goods invest heavily in price promotions (Walters and Mackenzie, 1988, Blattberg, Briesch and Fox, 1995). The result is that consumers are confronted with price promotions very often via TV commercials, in-store communication, billboards, leaflets, online banners, etc. Taking this into account one could realize that the amounts of money that goes around in price promotions are huge and at some point in time return to the company. In fact, these expenses return to companies mostly in the form of up-lifts in sales in the short run (Blattberg and Wisniewski, 1989; Moriarty, 1985). However, many researches have been showing that the frequency of price promotions leads to a decreasing consumer’s reference price (Lattin and Bucklin, 1989; Kalwani, Yim, Rinne and Sugita, 1990; Mayhew and Winer, 1992), which implies negative consequences for companies.
Next to this, Mela, Gupta and Lehmann (1997) showed that long-term effects of price promotions are mostly absent. Taking this together with the negative consequences of short-term effects, a logical question to ask would then be: ‘Why are we still promoting with large amounts via price?’. There could be many answers to this question. One could for example be that price promotions will generate trial which in turn could lead to penetration. Nevertheless, we could now argue that price promotion is not the means to the goal of building brand equity but in fact, as pointed out by Mela et al. (1997), destroys brand equity.
Finding a way to abandon price promotions and replace them with marketing activities that build on brand equity would be a possible solution for this problem. That would probably come in the form of a revolution since the whole market needs to change, for something like this to work. For that reason, it may be wise to find out in what way, price promotions can be improved in order to account (at least somehow) for the negative consequences as pointed out before.
1.1 Practical Relevance
In the Netherlands there are 3 big retail parties. Together they have a market share of 83.7% (Te Pas, 2018)which makes the market quite oligopolistic and therefore rather challenging for manufacturers from time to time. Next to this, price promotions are quite tempting for retailers since research has revealed the existence of a deal prone segment (Montgomery, 1971 and Webster, 1965). This segment is attracted to the store due to promotions and would not be visiting the store without an active price promotion. Price promotions, mostly in the form of loss leader promotions (advertised offer for sale of a product priced at or below retailer cost, see Davidson, Sweeney and Stampfl, 1984 and Mason and Meyer, 1985), function therefore as traffic generators. This traffic is very interesting for retailers because most of the time these shoppers are not only buying the product that is on promotion but complement the promoted products with related or even unrelated products. In this way, promotions can be a very interesting ‘tool’ for increasing traffic and so sales (Walters and Rinne, 1986). Therefore, and contrary to manufacturers, retailers appreciate the expansive power of price promotions (Nijs, Dekimpe, Steenkamp and Hanssens, 2001). This potential benefit in combination with the retailer’s strong power (83.7% market share) makes it hard for manufacturers to stay away from price promotions because that could lead to losing the retailer’s support with all its consequences.
So, retailers have a more positive approach towards price promotions and on top of that, they also have the ability to continuously track their effectiveness and improve them whenever necessary. This ability emanates from the large amounts of daily information on product sales derived from checkout scanners installed by the retailers. Usage of this data can provide retailers with a strong power over manufacturers as found by Kumar and Leone (1988), since the manufacturer’s access to data is more limited.
1.2 Research Question
In order to figure out how manufacturers could get most returns on price promotions it is wise to first understand the sales bump that arises from a price promotion (Van Heerde, Leeflang and Wittink, 2004). The sales bump explains in what form the investment of the price promotion returns to the company. Optimizing this sales bump could be realized whenever the drivers of the sales bump are clarified so that managers could take these into account when designing a price promotion. This seems to be rather challenging for managers, next to pointing out why certain promotions are more successful than others. The aim of this research is therefore to figure out the drivers of a sales bump in order to be able to optimize price promotions using the following research question:
‘What are the determinants of a successful price promotion for manufacturers of fast-moving consumer goods?’
In order to answer this question several multiple regression models will be estimated with the use of GfK data which is complemented with Nielsen data in order to make the promotions measurable. Findings reveal that for different brands within the same category, the determinants for the success of a price promotion are not the same. This could be ascribed to the difference in shopper segments, i.e. loyalty. However, there are also some variables found to be significant influential for all brands for example several promotional variables as well as the type of retailer.
1.3 Structure of this Thesis
2. Theoretical Framework
In this section different views, perspectives and arguments found in literature related to this topic of ‘the effectiveness of price promotions’ will be presented and discussed. Out of these insights, hypotheses will be formed and finally, graphically shown in the conceptual model.
Price promotions in this thesis are referred to as: ‘temporal price reductions offered to the consumer’ (Blattberg et al., 1995). The research of Blattberg et al. provides a clear overview on the topic since many empirical generalizations concerning price promotions are discussed and presented. For this reason, it functions here as a leading article by gathering different views, perspectives etc. on price promotions.
The first empirical generalization and also the most straightforward and logical one Blattberg et al. (1995) mention is: ‘Temporary retail price reductions substantially increase sales’. These increases in sales are caused by several effects as pointed out by Van Heerde et al. (2004). Increases in sales due to price promotions (sales bumps) can consist of three parts which are cross-brand effects (same product is bought but from another brand due to the promotion), cross-period effects (same product is bought but at another point in time due to the promotion) and category expansion effects (occurs when a product from a certain category is bought due to the price promotion while normally this (type of) product would not be bought). These effects will further be discussed in section 2.1.2.
Figure 1: Difference of responses to price promotions among categories
Chandon, Wissink and Laurent (2000) also performed research in this field of shoppers’ responsiveness to price promotions. The authors are rather critical against the general assumption of monetary savings being the only benefit of sales promotions that motivates people to buy the promoted product. Therefore, they created a framework in which multiple consumer benefits of price promotions are pointed out by performing in-depth interviews and critically analysing them. According to Chandon et al., promotional benefits are either hedonic or utilitarian and for managers in order to be most effective, the authors state that they should try to match the type of promotion to the type of product being promoted.
2.1 Promotional variables
2.1.1. Display and Feature Advertising
Another topic that many research is devoted to, is the influence of promotional variables such as display and feature support, advertisements and the place where the price promotion is stalled in the supermarket (e.g. end-of-the-aisle vs. on a regular shelf). The effect of display and feature advertising was mentioned in an empirical generalization by Blattberg et al. (1995) as ‘Display and feature advertising have a strong effect on item sales’ since the effect was found in many studies such as Walters and Rinne (1986), Bemmaor and Mouchoux (1991) and Kumar and Leone (1988). This effect can be explained by the dual-process theory which implies that human-beings can process information with two different modes: System 1 and System 2 (Kahneman and Frederick, 2002; Stanovich, 1999). The system 1 mode can be described as experiential, implicit and heuristic, whereas the system 2 mode can be described as rational, explicit and systematic (Evans, 2008). Display and feature advertising trigger the system 1 mode of processing information because they turn into cues that could finally make the shopper persuade of buying the product on promotion (Hoyer and Maclnnis, 2013). This purchase would mostly be an impulse buy. System 2 could prevent the shopper from impulse buying but due to the trigger that arises from display and feature advertising, the system 1 approach rules out system 2.
The individual influence of either displays or feature ads, however, seem to be difficult to point out (Blattberg et al., 1995; Zhang, 2006). This is because displays, and feature ads are often used in combination, and therefore the probability for multicollinearity to appear when doing research in this area, is quite high. Zhang (2006) summarized in her article that many other researchers pointed out interaction effects between displays, feature ads and price cuts but that these effects are quite different among the researches. For example, Gupta (1988) found a negative interaction between display and price cut whereas Papatla and Krishnamurthi (1996) found a positive interaction between the same variables.
consideration sets is greater than that of displays forming consideration sets, while the probability of using displays as price-cut proxies is greater than that of using feature ads.
This can be explained by price-cut proxies being described as mechanisms with a mere presence of a promotional signal which makes shoppers believe that the product offered at this place is on special offer (Inman, McAlister and Hoyer, 1990). Products offered in a feature advertisement, on the other side, will relatively faster end up in the consideration set because it has been made prominent to shoppers. This is picked up by the shoppers since they often rely on heuristics when forming their consideration set (Hauser and Wernerfelt, 1990; Robberts and Lattin, 1991). All in all, this would imply that shoppers process products on displays more as products that are on special offer, whereas products in feature ads will faster end up in the consideration set because here shoppers rely on heuristics in general (not specific on price) when processing. So, the difference between display and feature advertising is the emphasis on price marketing instrument.
This explanation greatly matches with another finding of Zhang (2006) that products on displays are more inherent to price cuts as products on feature ads, since products placed on second placement are almost always in combination with a price cut whereas feature advertising could also function as just brand building and not communicating anything about price.
2.1.2. Effects of price promotions
Putting this insight into the perspective of the manufacturer one would argue that category expansion effects would be the most desired since with these effects, additional sales could be derived that also could have positive impact on the long term (Van Heerde et al., 2004). Since these effects are largest at price promotion when both display and feature advertising being active, we formulate the following hypothesis:
H1a: The more promotional variables active at a price promotion (both display and feature), the more successful a price promotion will be.
However, as discussed previously, this may not be the case for all categories since Grover and Srinivasan (1992) found in their analysis that for different consumer segments, different promotional variables are influential. This can be explained by the loyalty of a customer. Grover and Srinivasan (1992) studied the differential response of consumer segments towards price promotions within the coffee category and identified these consumer segments on the basis of brand choice probabilities. For some segments the use of feature advertising had a significant impact on sales during a price promotion whereas for other segments coupons, or both coupons and feature advertising, were found to be significantly influential. Since the segments respond differently towards promotional variables, it is likely that this also holds for different brands within the same category. Therefore, hypothesis H1b is formulated as:
H1b: The influence of promotional variables active at a price promotion, is not the same across brands.
2.2 Types of Price Promotions
First of all, it might be useful to bear in mind that consumers on average spend a fraction of a second on looking at the package and point-of-sale material (POSM) (Lennon, 2013). Next to this, grocery shopping is highly habitual, as 69 percent of the shoppers buy the same brand as bought previously when they go shopping in the same category (Lennon, 2013). Therefore, it is of great importance for manufacturers to signal something from their packages or POSM that attracts the attention of the shopper, in order to let them buy your product. Specifically, temporal price reductions should be as convincing as possible in order to be most effective. Krishna, Briesch, Lehmann and Yuan (2002) showed that price reductions in terms of percentage are perceived larger than price reductions in terms of absolute numbers, while offering the exact same amount of discount. DelVecchio, Krishnan and Smith (2007) confirm this finding by asking shoppers about their general sense against percentage-off and cents-off promotions. However, when shoppers take the time and take a closer look at the product and start calculating the real price discount, price reductions in terms of absolute numbers are processed more easily and in turn, price reductions in terms of percentage will cause difficulty that could result in uncertainty about the resulting price. As people tend to prefer more precise outcomes over less precise ones (Camerer and Weber, 1992 and Winkler, 1991), price promotions with price reductions in absolute terms will be most effective. Taking into account that different types of price reductions could be beneficial in different ways, hypothesis 2a is formulated as:
H2a: The type of price reduction influences the success of a price promotion.
Situations in which percentage-off promotions are valued more favourable will be more common since shoppers only spend a fraction of a second looking at POSM and therefore almost never start calculating and comparing the price advantage of the promotion. Only in situations in which shoppers really take the time to find out the price advantage, cents-off promotions will be more effective. Next to this, shoppers perceive percentage-off price promotions larger than cents-off price promotions (Briesch et al., 2002 and DelVecchio et al., 2007) as discussed previously. From this perspective hypothesis 2b is formulated as:
2.3 Competition
2.3.1. Asymmetric Competition
This section will be devoted to the effects that competitors have or could have on the success of a price promotion, in perspective of the manufacturer. The article of Blattberg et al. (1995) is again taken here as a starting point. One of their empirical generalizations deals with an effect of competition: ‘Cross-promotional effects are asymmetric and promoting higher quality brands impacts weaker brands (and private label products) disproportionately.’ (Blattberg et al. 1995). A reason for this asymmetry could be the differences in brand equity: brands with larger brand equity benefit more from a price promotion, as brands with smaller brand equity. Sivakumar and Raj (1997) found that these asymmetric responses to price promotions exist on the basis of the quality, i.e. promotions of high-quality products are to a lesser extent negatively affected by promotions of low-quality products. Interestingly, this relation holds regardless of the dominant player in the market. So, when a private label is dominant within a certain category, the asymmetric effect will still be apparent (Sivakumar and Raj, 1997).
2.3.2. Competition and timing
Kumar and Leone (1989) identified the more general and underlying relation which can be described as: The price promotion of Brand A will lead to a decrease in sales of Brand B during this promotion. However, when the price promotions of Brand A and Brand B are taking place at the same time, this effect may not be apparent. In this situation, the advantage that Brand A had over Brand B is disappeared according to game theory (Von Neumann, Morgenstern, Kuhn and Rubinstein, 1944), therefore hypothesis 3 is formulated as:
H3: A price promotion of a direct competitor taking place during your price promotion will negatively affect the success of this price promotion.
2.3.3. Brand switching
Interestingly brand substitution (switching) is found to have the largest impact on a price promotion, i.e. on average 75% of the increased demand due to a price promotion comes from brand switching (Bell et al., 1999, Chiang, 1991, Chintagunta, 1993, Gupta, 1993). Gupta (1993) even pointed out with his research, using data from the coffee category, the percentage attributable to brand switching as part of the whole ‘sales bump’, to be 84% whereas purchase acceleration only accounted for 14% and stockpiling for 2%.
Van Heerde et al. (2004) refuted these relative high percentages for brand switching and claimed that only 33% of the sales promotion bump is due to brand switching instead of 75%. They review the previous articles in the field which clearly shows that the average percentage of brand switching among the different researches is about 75%. However, these percentages greatly vary among different categories (see Figure 1). Research by Chintagunta (1993) shows a brand switching percentage of 40% using data from the category yoghurt whereas Bell et al. (1999) found a brand switching percentage of 94% for the category margarine. Hence, it is crucial to take into account that the sales promotion bump can very much vary among categories.
Turning back to the difference between 33 and 75% attributable to brand switching, there should be some reason for these diverse findings on the same topic. The reason can be found in the way the numbers and percentage are derived. Previous researchers have been working with gross numbers, i.e. focused is on the gross change in sales for the non-promoted brands, when category volume is held constant. Van Heerde et al. (2003) on the other side, have been working with the net numbers and so they accounted for increasing category volume. Since the reason for the difference in outcomes at the different studies lies at the way these numbers are derived, one cannot argue that there is one best way. The optimal method or which percentage to work with should therefore depend on the situation.
that because Pepsi is on promotion, people will buy 4 instead of 2 bottles (purchase acceleration) and so the next week when Coca-Cola is on promotion these consumers are not about to buy Coca-Cola due to their inventory of Pepsi.
2.4 The Influence of Weather
The previously discussed influences and effects are to a certain extent all under control of the stakeholders which are connected to a certain price promotion. This control lies mostly at the retailer as they have a lot of power over price promotions. There can be situations in which ‘external factors’ are influential to the effectiveness of price promotions. The weather is such a factor, out of anyone’s control, which could be influential in many cases and possibly in this case too. Therefore, these possible effects will be further discussed at this section.
Steele (1951) wrote an article on this topic considering the effect of weather on the total sales of a department store. According to her, weather could affect retail sales in four different ways. First, the weather could make going to the store uncomfortable. Second, it could produce situations in which it physically prevents one from going to the store (e.g. heavy snowfall). Third, the weather may have psychological effects on consumers, which could affect their shopping habits. Fourth, some products are more desirable during a certain weather type (e.g. ice cream in summer).
2.4.1 Influence of Temperature
Cheema and Patrick (2012) did extensive research on the ‘Influence of Warm Versus Cool Temperatures on Consumer Choice’ in which they found several interesting effects. They found the effect of temperature on complex choices, which works in a way that the warmer it gets, the less likely it is that people will make difficult gambles. This can be explained by the underlying process of resource depletion. People can process information in two ways: the system 1 and system 2 approach. Since the system 2 approach is more effortful as system 1, one would need more resources when processing information with system 2 (Evans, 2008). So, when the temperature increases, one would process more information with the use of System 1 (Cheema & Patrick, 2012, Pocheptsova, 2009). This effect is closely related to the effect pointing out that higher temperatures result in a decrease in performance on complex tasks (Cheema & Patrick, 2012).
Some other findings in this field of temperature pointing out the existence of the same effect in a different setting: the higher the temperature, the less likely it is one would buy something innovative (Cheema & Patrick, 2012), the more difficulties one has with concentrating and so the worse performance will be (Witterseh, Wyon & Clausen, 2004), the lower office productivity will be (Van de Vliert & Van Yperen, 1996). And finally, Bandyopadhyaya (1978) pointed out that people in tropical climates have less inclination to work due to the high temperatures.
Taking these findings into consideration it is rather likely that temperature could have an effect on the effectiveness of a price promotion. This could be the case because if the temperature rises, shoppers are more likely to process information using the System 1 approach. If this System 1 approach is than in turn fed with cues of the certain price promotion, it is likely that temperature positively influences the success of a price promotion. Therefore, hypothesis 4 is formulated as:
2.5 Control Variables
Now we have discussed several potential determinants of a successful price promotion which will be of main interest in this thesis. It is not said however, that only these factors are the determinants for a successful price promotion. For that reason, several control variables will be added to the conceptual model and these variables will also be included in further analysis. In this way potential effects of these variables are taken into account when explaining the success of a price promotion, but these effects will not be emphasized.
2.6 Conceptual model
Below the different hypothesis including their directions are graphically shown in a conceptual model. The independent variables connected to the hypotheses, as indicated between brackets, ranging from 1 to 4 are showing a direct effect on the dependent variable ‘Success of a price promotion’. Success of a price promotion will be measured using the sales of the price promotions. Therefore, in this research it is assumed that successful price promotions account for relatively more sales than less successful price promotions. Hypothesis 1b is not graphically shown in the conceptual model since this effect will not be part of this initial model. In order to test hypothesis 1b, the same model will be tested for different brands in order to compare the results for the brands. This will be done using multiplicative regression to be performed in R. Details about the research design will be discussed at the next chapter.
3. Research design
In this chapter the way in which the previously presented hypotheses will be tested, are discussed. First of all, there will be elaborated on the data, its nature and where it roughly consists of. Thereafter, the research method will be introduced explaining how the analysis will look like and finally, the estimation model will be presented.
3.1 Data
The idea behind this thesis finds it origin at a Dutch A-brand producer and that is also the place where the data comes from. The data is not generated by the firm itself but is derived from GfK and Nielsen. The GfK data forms the basis for the dataset since this data specifically focuses on promotions and is therefore already quite complete. A price promotion can only end up in the GfK dataset whenever this promotion is communicated in a certain way, because that is what GfK specifically measures. Think of a promotion that has been advertised by the retailer via a tv-spot or a leaflet for example.
Although the GfK dataset is quite complete already, some fundamental variables are missing. These missing variables are the sales of the promotions and are retrieved from a large dataset from AC Nielsen. The data ranges from week 1, 2015 until week 4, 2018 and only the promotions that have been active at the three largest retailers Albert Heijn, Jumbo and Plus (Te Pas, 2018) are included. Also, the promotions from the 10 largest and most close competitors are incorporated. Bringing this data together results in a dataset with 985 promotions.
Table 1 gives an overview of the variables included in the datasets.
Variable Description Measured
Date Week where the promotion was
active
In Weeks
Product Specific name of the product Product names Product type Names used to identify different
layers in the category
Types of product in the category
Brand Brand of the product that is on promotion
Brand names
Packaging Kind of package where the product is packed in
Type of package
Retailer Retailer where the promotion was active
AH/Plus/Jumbo
Leaflet_Feature Indication of whether the promotion was advertised via leaflet and feature
Dummy: No leaflet and feature (0), leaflet and feature (1)
Leaflet_Display Indication of whether the promotion was advertised via leaflet and display
Dummy: No leaflet and display (0), leaflet and display (1)
Leaflet_FD Indication whether the promotion was advertised via leaflet, feature and display
Dummy: No leaflet, feature and display (0), leaflet, feature and display (1) TV_Feature Indication whether the promotion
was advertised via tv and feature
Dummy: No tv and feature (0), tv and feature (1) TV_FD Indication whether the promotion
was advertised via tv, feature and display
Dummy: No tv, feature and display (0), tv, feature and display (1)
TV.Leaflet_Feature Indication whether the promotion was advertised via tv, leaflet and feature
Dummy: No tv, leaflet and feature (0), tv, leaflet and feature (1)
TV.Leaflet_Display Indication whether the promotion was advertised via tv, leaflet and display
Dummy: No tv, leaflet and display (0), tv, leaflet and display (1)
TV.Leaflet_FD Indication whether the promotion was advertised via tv, leaflet, feature and display
Dummy: No tv, leaflet, feature and display (0), tv, leaflet, feature and display (1) Offer
percentage-off
Indication of whether the promotion is framed with percentage-off
Dummy: No percentage-off (0), Percentage-off (1) Offer based on
quantity
Indication of whether the promotion is framed based on quantity
Promo volume Sales of promoted article during the week of the promotion
In liters
Reduction rate Percentage of reduction on the normal sales price
Percentage
Total volume Total amount of volume during the year
In liters
Average volume Average amount of volume per week (based on the sales of the year)
In liters
Uplift in sales Promo volume divided by average volume
Percentage
Normal sales price Price charged for the product when not on promotion
In Euro’s
Temperature Average temperature during the week of the promotion
Degrees Celsius
Competition Indication whether a competitor is on sale during the same week at the same retailer
Dummy: No other promotion (0), Other active promotion (1)
Own competition Indication whether your brand is on sale during the same week at another retailer
Dummy: No other own promotion (0), Other active own promotion (1)
3.1.1. Dummy variables
Table one shows a great number of dummy variables and this section will elaborate on the nature of these dummies. In the dataset of GfK, promotional variables were covered with the indication of either, leaflet, tv-spot or both leaflet and tv-spot. As touched upon before, the communication of a promotion via a one of these medias is a condition for a promotion to end up in the dataset.
The Nielsen dataset also contained promotional variables about whether the promotion was display or feature advertised. Since a promotion in all three datasets always is display, feature or display and feature advertised these set of promotional variables is combined with the ones from the GfK dataset into one dummy variable. Table 2 gives an overview of the possible sets of promotional variables. The base is Leaflet and Display.
GfK
Nielsen
Feature Display Feature & Display
Leaflet Leaflet_Feature Leaflet_Display Leaflet_FD
TV TV_Feature TV_Display TV_FD
Leaflet & TV Leaflet.TV_Feature Leaflet.TV_Display Leaflet.TV_FD
Table 2: Overview of dummy promotional variables
3.2 Research method
3.3 Model Specification
Below the multiplicative model that will be estimated, is presented (Equation 1). This model is derived from the SCAN*PRO model (Wittink, Addona, Hawkes and Porter, 2011) which aims to quantify the effects of promotional activities initiated by the retailers. In a multiplicative model all variables are assumed to interact and therefore it is hard to interpret the different regressors. In order to better be able to interpret the regressors, the numerical variables in the model are linearized through log-transformation.
𝑃𝑆𝑗𝑡 = 𝛼𝛽1𝐿𝐹𝑗𝑡𝛽
2𝐿𝐷𝑗𝑡𝛽3𝐿𝐹𝐷𝑗𝑡𝛽4𝑇𝑉𝐹𝑗𝑡𝛽5𝑇𝑉𝐹𝐷𝑗𝑡𝛽6𝑇𝑉.𝐿𝐹𝑗𝑡𝛽7𝑇𝑉.𝐿𝐷𝑗𝑡𝛽8𝑇𝑉.𝐿𝐹𝐷𝑗𝑡
𝛽9𝑇𝑜𝑂𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑗𝑡𝛽
10𝑇𝑜𝑂𝑞𝑢𝑎𝑛𝑡𝑗𝑡𝛽11𝐶𝑡𝛽12𝑂𝐶𝑡𝛽13𝑇𝑒𝑚𝑝𝑡𝑅𝑅𝑗𝑡 𝛽14𝛽15𝑅𝑒𝑡𝑡𝜀𝑡
Equation 1: Multiplicative model
Where,
𝑃𝑆𝑗𝑡 = Promo Sales of brand j in week t
𝐿𝐹𝑗𝑡 = Leaflet and Feature advertising of brand j in week t 𝐿𝐷𝑗𝑡 = Leaflet and Display advertising of brand j in week t
𝐿𝐹𝐷𝑗𝑡 = Leaflet, Feature and display advertising of brand j in week t
𝑇𝑉𝐹𝑗𝑡 = TV and Feature advertising of brand j in week t
𝑇𝑉𝐹𝐷𝑗𝑡 = TV, Feature and Leaflet advertising of brand j in week t 𝑇𝑉. 𝐿𝐹𝑗𝑡 = TV, Leaflet and Feature advertising of brand j in week t 𝑇𝑉. 𝐿𝐷𝑗𝑡 = TV, Leaflet and Display advertising of brand j in week t
𝑇𝑉. 𝐿𝐹𝐷𝑗𝑡 = TV, Leaflet, Feature and Display advertising of brand j in week t
𝑇𝑜𝑂𝑝𝑒𝑟𝑒𝑛𝑡𝑗𝑡 = Type of Offer percent’s off, of brand j in week t 𝑇𝑜𝑂𝑞𝑢𝑎𝑛𝑡𝑗𝑡 = Type of Offer quantity based, of brand j in week t 𝐶𝑡 = Competition in week t
𝑂𝐶𝑡 = Own Competition in week t 𝑇𝑒𝑚𝑝𝑡 = Temperature in week t
𝑅𝑅𝑗𝑡 = Reduction rate of brand j in week t 𝑅𝑒𝑡𝑡 = Retailer of the active promotion in week t
4. Results
Now that the model has been specified, the model can be estimated in order to find out the determinants of a successful price promotion. In this chapter these determinants will be presented, but prior to this we will discuss some descriptive statistics in order to create a setting in which the results are easier to interpret. Next to this we will discuss the assumptions of the general linear model. Since the multiplicative model will be tested using OLS, these assumptions need to be satisfied. (Leeflang et al., 2015).
4.1 Descriptive statistics
The data contains promotions of three retailers which are: Albert Heijn, Jumbo and Plus. For each of the retailers a graph is presented below, in order to highlight and explain the differences among the retailers. The graphs show the promotion volumes per week of Brand 1 next to the average volumes to easily check the size of the promotion volume relative to the average volume.
Table 3: Descriptives retailers (Distrifood.nl)
Figure 2-4: Volumes AH, Plus and Jumbo
4.2 Assumptions for testing a linear model
In order to start with testing the model, six assumptions need to be satisfied (Leeflang et al., 2015). We will go through these assumptions step by step. The first one is about the nonzero expectation which is about unbiasedness in the model and can be caused by misspecification. To check whether the predictor variables suffer from bias, a plot of the residuals is created (Figure 5, Appendix 1) which does not show a systematic pattern in the residual values. To ensure this finding the RESET-test (Ramsey, 1969) is performed and showed significance with a value of 4.0623 and a p-value < 0.05. This indicates some form of misspecification in the model which will be taken into account when discussing the results and conclusions. For Brand 2 the situation concerning the nonzero expectation is almost the same as for Brand 1 (see Figure 8, Appendix 2, RESET: 19.23, p-value < 0.05) and for Brand 3 all conditions for satisfying this assumption are met (see Figure 11, Appendix 3, RESET: 0,1347, p-value > 0.05).
The second assumption states that the error term should be homoscedastic. In order to test this assumption, the Breusch-Pagan test is performed. The Breusch-Pagan shows no significant heteroskedasticity with a BP-value of 0.1464 and p-value of 0.1443, therefore this assumption is satisfied (Leeflang et al., 2015). For Brand 2 and 3 this assumption can also be satisfied since the p-values at the BP-test were both larger than 0.05.
The third assumption concerns autocorrelation which is about covariances between residuals. Autocorrelation can be detected with the Durbin-Watson test (Durbin and Watson, 1951) and is not detected within the model (DW=1.88134, p-value=0.486). The same holds for Brand 2 (DW=1.997258, p-value= 0.902) and Brand 3 (DW=1.970045, p-value=0.716).
Next to this, the disturbances should be normally distributed, and this can be checked with the use of a histogram and qq-plot (Appendix 1, Figure 6 and 7). The plots are quite favourable against this assumption. However, in order to be sure whether this assumption is satisfied a Shapiro-Wilk test for non-normality is performed (Shapiro and Wilk, 1965). The test pointed at the existence of non-normality (SW= 0.79812, p-value <0.05) which can be the result of outliers in the residuals. Since the sample size is not quite large and the fact that these extreme observations could contain relevant information, decided is to not dive deeper into this potential problem. Another remedy for this issue could be to log-transform the criterion variable (Leeflang et al., 2015). Since this log-transformation is already included in the model, we will not transform the model in order to completely meet the conditions for this assumption. The outcomes of the analysis will therefore be interpreted with care. The same situation holds for Brand 2 since the histogram and qq-plot (Appendix 2, Figure 9 and 10) look quite positive against this assumption however the Shapiro-Wilk test shows significance (SW= 0.71767, p-value< 0.05). Considering this result, the outcomes of the analysis concerning this brand will also be interpreted with care. For Brand 3 this assumption about normality can be satisfied looking at the histogram, qq-plot (Appendix 3, Figure 12 and 13) and positive result of the Shapiro-Wilk test (SW= 0.98765, p-value > 0.05).
4.3 Model estimation
Brand 1
Multiplicative linear regression was used to estimate the model as specified at chapter 3.3 for Brand 1. Table 4 shows the results of the analysis. The model explains 76.9% of the variance (adjusted 𝑅2 = 0.7691, 𝑅2 = 0.8108) which is considered rather high. The model has a p-value
< 2.2e-16 and is therefore significant. The variables Leaflet, Feature and Display, Percent’s off and Retailer Plus are found to be significant at a 0.1% significance level where Leaflet, Feature and Display has a p-value of 0.000263, Percent’s off 8.48e-05 and Retailer Plus of 2.37e-07. Next to this, the variables Leaflet and Feature (p-value 0.004033) and TV, Leaflet and Feature (p-value 0.002348) are found to be significant at a 1% significance level. Lastly the variables Leaflet and Feature (p-value 0.0269069), Reduction rate (p-value 0.048011) and Retailer Jumbo (p-value 0.016494) are found to be significant at a 5% significance level. All other variables do not show any significance.
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1
Table 4: Summarized output from model estimation (Brand 1)
Coefficients Estimate Std. Error t value Pr (>|t|)
Intercept 9.14001 1.50930 6.056 5.54e-08 ***
TV and Feature 0.81935 0.52382 1.564 0.12210
TV, Feature and Display 1.09056 0.48012 2.271 0.026069 *
Leaflet and Feature 1.52198 0.51252 2.970 0.004033 **
Leaflet and Display -0.90241 0.47910 -1.884 0.063611 .
Leaflet, Feature and Display 1.79286 0.46734 3.836 0.000263 ***
TV, Leaflet and Feature 2.18924 0.69443 3.153 0.002348 **
TV, Leaflet and Display 0.93823 0.64774 1.448 0.151765
TV, Leaflet, Feature and Display 1.93080 0.46376 4.163 8.48e-05 ***
Percent’s off -0.17489 0.47306 -0.370 0.712669 Quantity based -0.43826 0.28394 -1.543 0.127040 Competition -0.12398 0.13630 -0.961 0.366051 Own competition -0.05526 0.17791 -0.311 0.756998 Temperature -0.04633 0.12219 -0.379 0.705649 Reduction rate 0.86035 0.42781 2.011 0.048011 * Retailer Jumbo -0.55290 0.22526 -2.454 0.016494 *
Brand 2
Table 5 shows output of the analysis concerning for Brand 2. The model explains 72.09% of the variance (adjusted 𝑅2 = 0.7209, 𝑅2 = 0.7636) which is considered rather high but lower
than Brand 1. The model has a p-value < 2.2e-16 and is therefore significant. The variables TV and Feature and Retailer Plus are found to be significant at a 0.1% significance level where TV and Feature has a p-value of 0.000752 and Retailer Plus of 2.25e-15. The variables Leaflet and Display (p-value 0.005813), Leaflet, Feature and Display (p-value 0.009924), TV, Leaflet, Feature and Display (p-value 0.002525), Quantity based (0.004893), Temperature (p-value 0.009833) and Retailer Jumbo (p-value 0.001422) are found to be significant at a 1% significance level. Lastly the variables TV, Leaflet and Feature and TV, Leaflet and Display are found to be significant at a 5% significance level where TV, Leaflet and Feature has a p-value of 0.026436 and TV, Leaflet and Display of 0.014523. The other variables do not show any significance.
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1
Table 5: Summarized output from model estimation (Brand 2)
Coefficients Estimate Std. Error t value Pr (>|t|)
Intercept 9.26640 1.89722 4.884 4.98e-06 ***
TV and Feature 2.38040 0.68016 3.500 0.000752 ***
Leaflet and Feature 0.27897 0.67243 0.415 0.679310
Leaflet and Display 2.47373 0.87368 2.831 0.005813 **
Leaflet, Feature and Display 1.81302 0.68697 2.639 0.009924 **
TV, Leaflet and Feature 2.06172 0.91224 2.260 0.026436 *
TV, Leaflet and Display 2.16964 0.86909 2.496 0.014523 *
TV, Leaflet, Feature and Display 2.12659 0.68239 3.116 0.002515 **
Percent’s off -0.25956 0.62434 -0.416 0.678675 Quantity based -1.05180 0.36376 -2.892 0.004893 ** Competition -0.25022 0.15461 -1.618 0.109388 Own competition -0.02771 0.06519 -0.425 0.671935 Temperature 0.29335 0.11101 2.643 0.009833 ** Reduction rate 0.71630 0.50045 1.431 0.156093 Retailer Jumbo -0.82138 0.24884 -3.301 0.001422 **
Brand 3
Table 6 shows the output of the from the model estimation of Brand 3. The model explains 97.55% of the variance (adjusted 𝑅2 = 0.9755, 𝑅2 = 0.98) which is considered really high
because almost all variance of the dependent variable Promo Sales is explained by the variables included in the model and therefore the explanatory power is really high. The p-value of the model is < 2.2e-16 and therefore the model is significant. Next to this, almost all variables are significant. First of all, the variables TV, Feature and Display, Leaflet, Feature and Display, TV, Leaflet, Feature and Display, Quantity based, Retailer Jumbo and Retailer Plus are found to be significant at a 0.1% significance level. TV, Feature and Display, Leaflet, Feature and Display, Tv, Leaflet, Feature and Display and Retailer Plus all have a p-value smaller than 2e-16. Quantity based has a p-value of 0.000262 and Retailer Jumbo has a p-value of 04.02e-07. The variables TV, Leaflet and Feature (p-value 0.001812) and Temperature (p-value 0.004699) are significant at a 1% significance level. Lastly, the variables Percent’s off (p-value 0.015321) and Own Competition (p-value 0.019847) are significant at 5% significance level.
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1
Table 6: Summarized output from model estimation (Brand 3)
Coefficients Estimate Std. Error t value Pr (>|t|)
Intercept 9.20227 0.60491 15.213 < 2e-16 ***
TV, Feature and Display 3.86801 0.26325 14.107 < 2e-16 ***
Leaflet, Feature and Display 3.71358 0.25585 15.118 < 2e-16 ***
TV, Leaflet and Feature -0.82219 0.25029 -3.285 0.001812 **
TV, Leaflet, Feature and Display 3.88844 0.25611 15.182 < 2e-16 ***
Percent’s off -0.51568 0.20579 -2.506 0.015321 * Quantity based -0.79044 0.20204 -3.912 0.000262 *** Competition -0.08609 0.05218 -1.650 0.104920 Own competition -0.14963 0.06229 -2.402 0.019847 * Temperature 0.12842 0.04351 2.952 0.004699 ** Reduction rate 0.25548 0.14840 1.722 0.090992 .
Retailer Jumbo -0.56050 0.09695 -5.781 04.02e-07 ***
5. Conclusion
In this chapter, the results of the previous chapter will be discussed. The discussion will start with the output of the analysis on Brand 1 and thereafter, the results of the analyses on Brand 2 and 3 will be discussed. On this basis the hypotheses will either be accepted or rejected. Finally, the managerial implications will be discussed as well as the limitations and suggestions for future research.
5.1 Discussion on Brand 1
Considering the results of Brand 1, there has only been significance found for several sets of promotional variables (except for the control variables). What can be said about the variables being significant in comparison to the ones showing insignificance, is that there is no clear system in significance. Whenever a certain variable, TV for example, is part of the set of promotional variables, this single variable cannot function as a success factor for this set of promotional variables to be significant. Next to this, almost all sets of promotional variables that contain more (more than 2) variables are found to be significant. One exception is the combination TV, Leaflet and Display which is not found to be of significant different influence on the success of a price promotion compared to the base (Leaflet and Display). However, the set of promotional variables that contains all variables (TV, Leaflet, Feature and Display) did show significance which makes accepting hypothesis 1a very plausible.
5.2 Discussion on Brand 2
The output of the analysis on Brand 2 shows significance on a lot of different variables as compared to the analysis on Brand 1. First, nearly all sets of promotional variables show positive significance which means that price promotions, promoted with these sets of variables substantially and positively differ from promotions promoted with the base set (Leaflet and Display). Since this is the case in almost every scenario, accepting hypothesis 1a becomes less plausible as it was for Brand 1. Next to this, the situation concerning display and feature advertising as discussed in the previous section, seems not to be apparent for Brand 2. Therefore, the study of Van Heerde et al. (2004) finds less support, taking also the results of Brand 2 into consideration.
Interesting for Brand 2 is the fact that other variables show significance compared to the output of Brand 1. Considering the dummy variable on price framing: Quantity based price reduction is found to be negatively significant compared to the base, Cents-off. This result would make us accept hypothesis 2a and also matches the theory of people preferring more precise outcomes over less precise ones (Camerer and Weber, 1992 and Winkler, 1991). However, hypothesis 2b cannot be accepted since the variable Percent’s off did not show significance.
This finding does not necessarily mean that the type of price framing is the reason for being unable to accept hypothesis 2b. It could be the case that price framing is moderated by another variable. After checking whether this could be the issue in our dataset, we could say that the reduction rate could be the possible moderator. Price frame options other than Cents-off, are mostly active in combination with a relatively lower reduction rate. Additional research is therefore needed to be able to decide upon acceptance or rejection of hypothesis 2b.
5.3 Discussion on Brand 3
The output of the analysis on Brand 3 shows significance on nearly all variables except for the variable Competition. The included sets of promotional variables are all found to be significantly different from the base. One set of variables (TV, Leaflet and Feature) however, shows a negative correlation which is not in line with earlier results in literature and hence, not with our hypotheses. Careful second analysis of the data indicates that this negative estimation value is connected to a single data point which is a price promotion with a relative low reduction rate. Therefore, this negativity might seem odd but is rather logical. Other sets of promotional variables show positive significance but since there are no sets of promotional variables included with only two variables (only sets of three and four variables) we do not have enough evidence to accept hypothesis 1a for Brand 3.
Variables indicating the type of price promotion show negative significance compared to the base, cent’s off. This indicates that price promotions framed using percent’s off or quantity based are less effective as price promotions framed using cent’s off. On this basis hypothesis 2a can be accepted as was the case for Brand 2.
Considering hypothesis 2b, the situation as pointed out for Brand 2 (influence of a potential moderator) is found to be apparent here as well.
The analysis on Brand 3 is the first indicating significance regarding the effects of competition and shows negative significance regarding the variable Own Competition. This means that whenever Brand 3 is on promotion at a certain retailer, the success of this price promotion decreases whenever this brand is on promotion at another retailer.
The last variable Temperature also showed positive significance which means that the case for Brand 2 concerning Temperature also applies for Brand 3.
5.4 Overall discussion
Having discussed the separate effects for the different brands, this section will unite these effects by turning back to the hypotheses and answering the research question. Table 7 gives an overview of the hypotheses and an indication of whether these hypotheses are supported.
The first hypothesis deals with the promotional variables and whether more active promotional variables lead to a more successful price promotion. Generally, we can conclude that promotional variables do have an impact on the success of a price promotion since many sets of promotional variables showed significance at all three brands. However, we cannot conclude that whenever more promotional variables are active during a price promotion that this price promotion will be more successful than whenever less promotional variables are active at this promotion. This is because almost all sets of promotional variables showed significance for Brand 2 and Brand 3. Interestingly, Brand 1 showed less significance on the smaller sets of promotional variables and using these significance levels, hypothesis 1 can partly be accepted.
Next to this, we can conclude that Brand 1 greatly differs on the number of variables being significant (concerning promotional variables but also concerning the other included variables) and this can be explained by the loyalty of the shoppers for this Brand. Brand 1 has the strongest segment of loyal buyers in the market which should make them less vulnerable for promotional effects. This is also shown in this analysis. The effect of loyalty on the success of a price promotion matches the idea that brands having more loyal shoppers, need less advertising than brands with less loyal shoppers, when striving for optimal advertising and trade promotional policies (Agrawal, 1996).
The second hypothesis is about the influence of price framing on the success of a price promotion. For Brand 2 and 3, support is found to accept hypothesis 2a which corresponds to findings in other related studies (Camerer and Weber, 1992 and Winkler, 1991). Since this support was not found for Brand 1, we partly accept hypothesis 2a.
For the third hypothesis, which is about the effect of competition on the success of your price promotion, none of the related variables showed significance. Therefore, hypothesis 3 is rejected and the research of Kumar and Leone (1989), stating that a price promotion of Brand A, will lead to a decrease in sales of Brand B during this promotion, is not supported with this research.
The last hypothesis, relates to the influence of Temperature on the success of a price promotion. This hypothesis can partly be accepted since support is found for this hypothesis at Brand 2 and 3. The reason for finding no support for this hypothesis with Brand 1, could be the large number of loyal buyers for this brand. However, there could also be other reasons for this difference, as for example an unequal division of promotions with ‘good’ and ‘bad’ weather among the brands. Therefore, we recommend performing future research on this topic.
Hypothesis Supported
H1 The more promotional variables active at a price promotion (both display and feature), the more successful a price promotion will be.
Partly
H2a The type of price reduction influences the success of a price promotion.
Partly
H2b The percent’s-off of price reduction type is the most influential on price promotion’s success.
No
H3 A price promotion of a direct competitor taking place during your price promotion will negatively affect the success of this price promotion.
No
H4 Temperature has a positive influence on the success of a price promotion.
Partly
Table 7: Summary of hypotheses
5.5 Managerial implications
This research revealed many insights that could be useful to marketing practitioners and at this section these insights are discussed.
The general conclusion stating that the influence of the determinants on a successful price promotion is moderated by brand, might be the most useful implication for managers. First of all, this potential effect was not hypothesized, but this does not mean that it is of lesser importance than the other results. Taking this conclusion in mind, managers should make sure that they understand their brand, their products and their buyers to the fullest, in order to make the right decisions concerning price promotions. We, for example, observed that buyers of Brand 1 are less responsive to promotional variables, than buyers of Brand 3 are. A price promotion may therefore be more useful for Brand 3 than for Brand 1.
Next to this, it may be wise for managers to critically assess the promotional tools they could use for price promotions and also assess how these might be influential to their successful price promotion, since their influence greatly varies across brands. Since no support was found for accepting hypothesis three, another managerial implication would be to rather not focus on actions of competitors when designing your own price promotion, because the research did not yield any results indicating competition as a threat for the success of your price promotion.
5.6 Limitations and suggestions for future research
This research revealed many interesting and useful insights, but there are also a number of limitations that are mentioned below.
First, the research is performed using a relative small amount of data. The analysis is performed with the use of three brands, more specifically three sku’s. from within one category. Next to this, we pointed out that the responses to price promotions vary greatly among categories (Van Heerde et al. 2004), which makes the take-outs from this research less generalizable among other categories. In order to make this research more generalizable and increase its power, additional research should be performed with products from different brands and different categories. Also, only the price promotions from the three largest retailers within the Netherlands are taken into account within this research, and again to increase the validity and generalizability of the results, this number should be increased at future research.
Second, the variation of several variables is rather minimal which decreases validity of the results concerning these variables. The price framing options for example, suffered from a lack of variety. This should be something to bear in mind when performing future research in order to be able to properly draw conclusions.
6. References
Agrawal, D. (1996). Effect of Brand Loyalty on Advertising and Trade Promotions: A Game Theoretic Analysis with Empirical Evidence. Marketing Science, 15(1), 86-108.
Bandyopadhyaya, J. (1978). Climate as an Obstacle to Development in the Tropics.
International Social Science Journal, 30, 339-352.
Bell, D., Chiang, J., & Padmanabhan, V. (1999). The Decomposition of Promotional Response: An Empirical Generalization. Marketing Science, 18(4), 504-526.
Bemmaor, A., & Mouchoux, D. (1991). Measuring the Short-Term Effect of In-Store Promotion and Retail Advertising on Brand Sales: A Factorial Experiment. Journal of
Marketing Research,28(2), 202-214.
Blattberg, R., Briesch, R., & Fox, E. (1995). How Promotions Work. Marketing
Science, 14(3), G122-G132.
Blattberg, R., Buesing, T., Peacock, P., & Sen, S. (1978). Identifying the Deal Prone Segment. Journal of Marketing Research, 15(3), 369-377.
Blattberg, R., & Wisniewski, K. (1989). Price-Induced Patterns of Competition. Marketing
Science, 8(4), 291-309.
Camerer, C. & Weber, M. J. (1992) Recent developments in modeling preferences: Uncertainty and ambiguity. Journal of Risk and Uncertainty, 5(4), 325-370
Chandon, P., Wansink, B., & Laurent, G. (2000). A Benefit Congruency Framework of Sales Promotion Effectiveness. Journal of Marketing, 64(4), 65-81.
Cheema, A., & Patrick, V. (2012). Influence of Warm Versus Cool Temperatures on Consumer Choice: A Resource Depletion Account. Journal of Marketing Research, 49(6), 984-995.
Chiang, J. (1991). A Simultaneous Approach to the Whether, What and How Much to Buy Questions. Marketing Science, 10(4), 297-315.
Chintagunta, P. (1993). Investigating Purchase Incidence, Brand Choice and Purchase Quantity Decisions of Households. Marketing Science, 12(2), 184-208.
Davidson, W., Sweeney, D., & Stampfl, R., (1984). Retailing Management, New York: Wiley.
DelVecchio, D., Krishnan, H., & Smith, D. (2007). Cents or Percent? The Effects of
Evans, J. St. B. T. (2008), Dual-Processing Accounts of Reasoning, Judgement, and Social Cognition. Annual Review of Psychology, 59, 255-278
Grover, R., & Srinivasan, V. (1992). Evaluating the Multiple Effects of Retail Promotions on Brand Loyal and Brand Switching Segments. Journal of Marketing Research, 29(1), 76-89. Gupta, S. (1988). Impact of Sales Promotions on When, What, and How Much to
Buy. Journal of Marketing Research, 25(4), 342-355.
Hauser, J., & Wernerfelt, B. (1990). An Evaluation Cost Model of Consideration Sets. Journal of Consumer Research, 16(4), 393-408.
Van Heerde, H., Gupta, S., & Wittink, D. (2003). Is 75% of the Sales Promotion Bump Due to Brand Switching? No, Only 33% Is. Journal of Marketing Research, 40(4), 481-491. Van Heerde, H., Leeflang, P., & Wittink, D. (2004). Decomposing the Sales Promotion Bump with Store Data. Marketing Science,23(3), 317-334.
Inman, J., McAlister, L., & Hoyer, W. (1990). Promotion Signal: Proxy for a Price Cut? Journal
of Consumer Research, 17(1), 74-81.
Hoyer, W. D. & Maclnnis, D. J. (2013). Consumer Behavior, International Edition, Winfield, KS: South-Western College Pub.
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263-291.
Kahneman, D., & Frederick, S. (2002). Representativeness revisited: attribute substitution in
intuitive judgement. In Heuristics and Biases: The Psychology of intuitive Judgement, ed. T.
Gilovich, D. Griffin, D. Kahneman, 49-81, Cambridge, UK: Cambridge Univ. Press Kalwani, M., Yim, C., Rinne, H., & Sugita, Y. (1990). A Price Expectations Model of Customer Brand Choice. Journal of Marketing Research, 27(3), 251-262.
King, D. (1999), Chocs Away: Boxed Chocolate Mints Industry. Industry Overview, 1022(5) King, C., & Narayandas, D. (2000). Coca-cola’s new vending machine (A): pricing to capture value, or not? Harvard Business School Case # 9-500-068.
Krishna, A., Briesch, R., Lehmann, D.R., & Yuan, H. (2002). A meta-analysis of the impact of price presentation on perceived savings. Journal of Retailing, 78(2), 101-118
Kumar, V., & Leone, R. P. (1988). Measuring the effect of retail store promotions on brand and store substitution. Journal of Marketing Research, 178-185.
Leeflang, P., Wieringa, J., Bijmolt, T., & Pauwels, K. (2015). Modeling markets: Analyzing marketing phenomena and improving marketing decision making (International series in quantitative marketing). New York: Springer.
Lemmon, B. (2013) Finding faster growth – Happy shoppers spend more. Opinion leader. Retrieved from: www.tnsglobal.com
Mason, M. B., & Mayer, M. (1984). Modern Retailing, Plano TX: Business Publications. Mayhew, G., & Winer, R. (1992). An Empirical Analysis of Internal and External Reference Prices Using Scanner Data. Journal of Consumer Research, 19(1), 62-70.
Mela, C., Gupta, S., & Lehmann, D. (1997). The Long-Term Impact of Promotion and Advertising on Consumer Brand Choice. Journal of Marketing Research, 34(2), 248-261. Montgomery, D. (1971). Consumer Characteristics Associated with Dealing: An Empirical Example. Journal of Marketing Research, 8(1), 118-120.
Moriarty, M. M. (1985). Retail Promotional Effects on Intra-and Interbrand Sales Performance. Journal of Retailing, 61(3), 27-47.
Von Neumann, J., Morgenstern, O., Kuhn, H., & Rubinstein, A. (1944). Theory of Games and Economic Behavior (60th Anniversary Commemorative Edition). Princeton University Press. Nijs, V., Dekimpe, M., Steenkamp, J. B. E. M. & Hanssens, D. (2001). The
Category-Demand Effects of Price Promotions. Marketing Science, 20(1), 1-22.
Papatla, P., & Krishnamurthi, L. (1996). Measuring the Dynamic Effects of Promotions on Brand Choice. Journal of Marketing Research, 33(1), 20-35.
Pauwels, K., Hanssens, D., & Siddarth, S. (2002). The Long-Term Effects of Price Promotions on Category Incidence, Brand Choice, and Purchase Quantity. Journal of
Marketing Research, 39(4), 421-439.
Pocheptsova, A., Amir, O., Dhar, R., & Baumeister, R. (2009). Deciding without Resources: Resource Depletion and Choice in Context. Journal of Marketing Research, 46(3), 344-355. Radas, S., & Shugan, S. (1998). Seasonal Marketing and Timing New Product
Introductions. Journal of Marketing Research, 35(3), 296-315.
Ramsey, J. (1969). Tests for Specification Errors in Classical Linear Least-Squares Regression Analysis. Journal of the Royal Statistical Society. Series B
(Methodological), 31(2), 350-371.
Roberts, J., & Lattin, J. (1991). Development and Testing of a Model of Consideration Set Composition. Journal of Marketing Research, 28(4), 429-440.
Sivakumar, K. (1996). Tradeoff between Frequency and Depth of Price Promotions: Implications for High- and Low-Priced Brands. Journal of Marketing Theory and Practice, 4(1), 1-8.
Sivakumar, K., & Raj, S. (1997). Quality Tier Competition: How Price Change Influences Brand Choice and Category Choice. Journal of Marketing, 61(3), 71-84.
Stanovich, K.E. (1999). Who is Rational? Studies of Individual Differences in Reasoning. Mahwah, NJ: Elrbaum.
Steele, A. (1951). Weather's Effect on the Sales of a Department Store. Journal of Marketing, 15(4), 436-443.
Te Pas, H. (2018, January 26). IRI: Jumbo wint, AH verliest marktaandeel. Retrieved from http://www.distrifood.nl
Van De Vliert, E., & Van Yperen, N. (1996). Why Cross-National Differences in Role Overload? Don't Overlook Ambient Temperature! The Academy of Management
Journal, 39(4), 986-1004.
Vernon, R. (1966). International Investment and International Trade in the Product Cycle. The Quarterly Journal of Economics,80(2), 190-207.
Webster, F. (1965). The "Deal-Prone" Consumer. Journal of Marketing Research, 2(2), 186-189.
Walters, R., & MacKenzie, S. (1988). A Structural Equations Analysis of the Impact of Price Promotions on Store Performance. Journal of Marketing Research, 25(1), 51-63.
Walters, R. G., & Rinne, H. J. (1986). An Empirical Investigation into the Impact of Price Promotions on Retail Store Performance. Journal Of Retailing, 62(3), 237.
Winkler, R. L. (1991). Ambiguity, Probability, Preference and Decision-Analysis. Journal of
Risk and Uncertainty, 4(3), 285-297
Witterseh, T., Wyon, D. P. & Clausen, G. (2004), The effects of moderate heat stress and open-plan office noise distraction on SBS symptoms and on the performance of office work. Indoor Air, 14, 30–40.
Wittink, D.R., Addona, M.J., Hawkes, W.J., Porter, J.C. (2011). SCAN*PRO: The estimation, validation and use of promotional effects based on scanner data. In: J.E. Wieringa, J.C. Hoekstra, P.C. Verhoef (eds.) Liber AMicorum in Honor of Peter S. H. Leeflang. University of Groningen: Groningen
7. Appendices
7.1 Appendix 1: Plots and table for assumptions Brand 1
Figure 5: Plot of residuals against time
Table 4: VIF-values
7.2 Appendix 2: Plots and table for assumptions Brand 2
Figure 8: Plot of residuals against time
Figure 9: Histogram of residuals
Figure 10: Normal qq-plot
7.3 Appendix 3: Plots and table for assumptions Brand 3
Figure 11: Plot of residuals against time
