THE IMPACT OF DELISTING ECONOMY PRIVATE LABELS ON CATEGORY SALES AND SUBSTITUTION PATTERNS

(1)

THE IMPACT OF DELISTING

ECONOMY PRIVATE LABELS ON

CATEGORY SALES AND

SUBSTITUTION PATTERNS

Stephen Ophof

S2919192

Master’s Thesis

MSc Marketing Management

Faculty of Economics and Business

University of Groningen

(2)

THE IMPACT OF DELISTING

ECONOMY PRIVATE LABELS ON

CATEGORY SALES AND

SUBSTITUTION PATTERNS

Master’s Thesis

MSc Marketing Management Faculty of Economics and Business

University of Groningen

Sportlaan 157

7691BK Bergentheim, The Netherlands +31 6 23567276

S.Ophof@student.rug.nl

(3)

MANAGEMENT SUMMARY

The competition field of supermarkets has dramatically changed throughout Europe. Hard discounters, like ALDI and Lidl, obtained market share from the existing supermarkets. At the moment, Lidl is the third supermarket chain of The Netherlands with a share of 11%. Thanks to small assortments and high efficiency, these hard discounters can sell products at bottom prices, compared with the traditional supermarkets. As a result, big parties like Albert Heijn, Tesco and Carrefour suffer from the enormous expansion throughout Europe. Cheap locations were filled with hard discount outlets. Traditional retailers had to develop defense strategies to keep their customer base loyal to them and prevent their customers from visiting hard discounter’s stores. A much-followed price-based strategy is the introduction of an economy private label (EPL), a product line positioned below the private label, like Tesco’s Everyday Value or Albert Heijn’s AH BASIC. The effectiveness of EPLs is questioned: former research showed that the introduction of EPLs can alienate certain shopper segments. It may also jeopardize the traditional retailer’s perceived quality. And more important, despite the introduction of EPLs, the market share of both ALDI and Lidl increased in all markets where they are active. As EPLs’ effectiveness is doubted, traditional retailers can consider delisting their budget label. Delisting items may reduce supply chain costs, increase search efficiency and increase assortment satisfaction. However, delisting this assortment seems to be a prisoner’s dilemma as no retailer wants to take the first step.

In this study, we conduct a field experiment in an average-sized supermarket where we delist EPL items, with a high share in volume and turnover, from four product categories. This field experiment was guided by the following main question: “What impact does delisting economy private labels have on traditional retailer’s category sales and substitution patterns?”. The four categories are chosen based on the market share of EPL items in this category and product type (classified as hedonic or utilitarian). These categories were: mozzarella cheese, melba toast, scrub sponges and aluminium foil. To exclude possible endogenous effects, sales data before, during and after the experiment are measured at three control stores.

(4)

items and/or the item with the smallest price distance. We also find significant differences in sales or substitution regarding product type (hedonic versus utilitarian). It seems that if a product is classified as hedonic, people tend to substitute easier than when a product is labeled as utilitarian. Finally, the larger the price distance between the delisted item and the remaining substitutes, the less consumers will tend to buy a substitute. After the experiment, it seems that the delisted items do not revert immediately to their initial market share. In the most cases, the products where people switched to during the experimental phase still have a higher sales share than before the experiment. Due to time constraints, long-term data is unfortunately not present, but we expect that unit and turnover shares will restore to their initial values on the long-term.

Our research shows that surviving without economy private labels is possible. For consumers, it can lead to decreased choice complexity and increased search efficiency. For retailers, cost savings will occur in the supply chain and reference price and willingness-to-pay will go up. Instead of allocating more facings to the residual products, retailers could decide to introduce more premium private label items to increase category turnover and quality perception. As hard discounters started selling national brands, this means that a (premium) private label is “the” trump card to build a unique proposition on assortment level. However, long term effects are unknown. Volume sales drop in one category during the experiment. It might be that consumers visit competitors to buy the delisted item.

Regarding our study, we have to acknowledge some limitations regarding generalizability and validity as this experiment was conducted in only one store in a rural area of The Netherlands and due to time constraints, only in four product categories in a short period of time. Economic circumstances could have influenced the outcomes. Because EPL items are price-sensitive items, the outcomes may have been different when the macro-economy was in times of recession. Directions for further research could to be conduct a similar experiment on a broader scale. Another meaningful experiment could be to investigate what happens with category sales and substitution patterns if EPL items are replaced with premium private label items.

(5)

MANAGEMENTSAMENVATTING (NL)

Het concurrentieveld in de supermarktwereld is drastisch veranderd in de Europese landen. Harddiscounters, zoals ALDI en Lidl, hebben marktaandeel afgesnoept van de bestaande partijen. Lidl is op het moment van schrijven de derde supermarktketen van Nederland (11% marktaandeel). Dankzij kleine assortimenten en een hoog efficiencyniveau kunnen harddiscounters toch genoeg marge maken met lage verkoopprijzen. Het gevolg was dat grote partijen als Albert Heijn, Tesco en Carrefour last kregen van de enorme opkomst van hard discount in Europa. Nieuwe, goedkopere supermarktlocaties werden opgevuld door de komst van harddiscounters. Men moest verdedegingssstrategieën bedenken om haar klantenkring loyaal te houden en te voorkomen dat vaste klanten harddiscounters gaan bezoeken. Een veel gevolgde prijsstrategie was de introductie van een budgethuismerk, een productlijn gepositioneerd onder het huismerk, zoals Everyday Value bij Tesco of AH BASIC bij Albert Heijn. Het budgethuismerk was geen geweldige oplossing, aangezien meerdere onderzoeken reeds aantoonden dat juist de introductie van budgethuismerken ervoor zorgde dat bepaalde klantensegmenten minder loyaal werden aan de traditionele supermarkt. Daarnaast kunnen budgethuismerken de kwaliteitsperceptie van een retailer negatief beïnvloeden. Als resultaat zien we dat ALDI en Lidl, ondanks de introductie van budgethuismerken, in alle markten significant marktaandeel hebben gewonnen, hoewel deze groei voor ALDI de laatste jaren is afgenomen. Omdat de effectiviteit van budgethuismerken betwijfeld wordt, zouden de traditionele servicesupermarkten kunnen overwegen om het budgethuismerk te saneren. Het saneren van de budgetlijn zou een enorme kosten- en ruimtebesparing kunnen opleveren in de logistieke keten en de zoekefficiëntie en assortimentstevredenheid van de consument kunnen verbeteren. Hoewel er reeds over gesproken wordt blijkt het saneren van het budgethuismerk een prisoner’s dilemma te zijn, aangezien geen enkele retailer de eerste stap durft te nemen.

In dit onderzoek is een veldexperiment uitgevoerd in een supermarkt met een gemiddeld verkoopvloeroppervlak (1200 m2_{) waar het budgethuismerk geschrapt is in vier}

(6)

(7)

PREFACE

This master’s thesis is the final project of my master’s degree Marketing Management at the University of Groningen. The last five months can be described as an intensive time, but also a time I enjoyed and I learned a lot. Doing a field experiment was something that was already in my head for a very long time. After discussing this research idea with my supervisor, I only had to get permission from a store franchisee. It took a bit of time, but finally, all lights turned green and I was able to conduct this field experiment.

I am grateful to all people that helped me during the process of writing this master’s thesis. First of all, prof. dr. Laurens Sloot, who transferred a lot of retail expertise to me and he had the honor to supervise me ;-). Particularly, I appreciate his excellent guidance and support during this process. Secondly, I would like to thank the franchisee of the retail stores where I was able to conduct my field experiment. Without him, doing this field study was not even possible. Finally, it would like to thank the second supervisor, dr. Wander Jager, for his feedback.

I hope that you enjoy reading this piece of work and if there are any questions, feel free to contact me.

Stephen Ophof

MSc Marketing Management University of Groningen

(8)

TABLE OF CONTENTS

1. INTRODUCTION ... 10

1.1BACKGROUND PROBLEM... 10

1.2PROBLEM STATEMENT ... 11

1.3RESEARCH QUESTION ... 13

1.4PRACTICAL AND ACADEMIC RELEVANCE ... 14

1.5STRUCTURE OF THE THESIS ... 14

2. THEORY AND BACKGROUND ... 15

2.1THEORETICAL BACKGROUND ON ITEM DELISTING ... 15

2.2HEDONIC AND UTILITARIAN PRODUCT CATEGORIES ... 17

2.3DELISTING EPLS’ IMPACT ON SUBSTITUTION PATTERNS AND CATEGORY SALES ... 18

3. CONCEPTUAL FRAMEWORK AND HYPOTHESES ... 20

3.1DELISTING EPLS’ IMPACT ON CATEGORY SALES ... 20

3.2DELISTING EPLS’ IMPACT ON SUBSTITUTION PATTERNS ... 21

3.3THE MODERATING ROLE OF PRODUCT TYPE ... 22

3.4THE MODERATING ROLE OF SALES PRICE DISTANCE ... 23

3.5CONTROL VARIABLES ... 23

4. METHODOLOGY ... 24

4.1PRE-TEST SURVEY... 24

4.2DELISTING EXPERIMENT ... 26

4.3SHELF TRANSFORMATION ... 28

4.4PRICE DISTANCE MEASUREMENT ... 29

5. RESULTS ... 30

5.1INTRODUCTION ... 30

5.2DESCRIPTIVE DATA ANALYSIS ... 31

5.2.1 Experiment group 1: mozzarella ... 32

5.2.2 Experiment group 2: melba toast ... 33

5.2.3 Experiment group 3: scrub sponges ... 34

5.2.4 Experiment group 4: aluminium foil ... 36

5.2.5 Turnover effects on the long-term ... 37

5.3OPERATIONALIZATION OF VARIABLES ... 37

5.4HYPOTHESIS TESTING ... 38

5.4.1 Effect of delisting on category unit sales... 39

5.4.2 Effect of delisting on category turnover ... 40

(9)

5.4.4 Substitution patterns with respect to product type ... 42

5.4.5 Substitution patterns with respect to price distance ... 43

5.5HYPOTHESIS OVERVIEW... 44

6. CONCLUSIONS AND IMPLICATIONS ... 45

6.1CONCLUSIONS AND DISCUSSION ... 45

6.2IMPLICATIONS... 46

6.3LIMITATIONS AND FURTHER RESEARCH DIRECTIONS ... 47

REFERENCES ... 49

APPENDICES ... 54

APPENDIXA:SETUP OF PRE-TEST SURVEY ... 54

APPENDIXB:EPLPRODUCT ITEMS ... 55

(10)

1. INTRODUCTION

This section will first introduce the topic by describing the background problem. After that, the problem statement will address the issues that raised. This will consequently lead to the main research question and subquestions to be answered and the theoretical and practical relevance of this study. We end this section with a view on this thesis’ structure.

1.1 Background problem

A long time ago, I talked with my grandparents about the grocery stores in their neighborhood. They still remember the introduction of the first supermarket in town, which was a very innovative store that time. They had self-service and shopping carts. There only was one supermarket in their neighborhood. And the shelves were packed with only a few hundred products in total. A shortcoming or blessing?

The last few decades, the competition field of supermarkets dramatically changed throughout Europe. The rise of hard discounters, such as ALDI and Lidl, damaged the good, profitable position of traditional food retailers. Thanks to their limited SKU numbers and private label focus, hard discounters have an extremely efficient supply chain, low staffing levels and low marketing costs. This efficiency allows them to sell products at bottom prices, compared with traditional food retailers (Steenkamp & Kumar, 2009). This strategy is being rewarded and appreciated by consumers and trade journals. ALDI provides “good value for money” and has influence on overall pricing policy trends in food (Harvard Business School, 2015). Competitor Lidl was rewarded several times by marketing research institute GfK for being the best retailer in vegetables and fruit (Levensmiddelenkrant, 2011). The Dutch consumers’ association Consumentenbond rewarded several Lidl products because they offer the best quality for the lowest price (Levensmiddelenkrant, 2013). Traditional grocery retailers, like Albert Heijn, Tesco, Delhaize and Carrefour suffer from the enormous expansion of hard discount throughout Europe. In some countries, the “better value-for-money” proposition let hard discounters capture up to 35% market share (Cleeren et al., 2010).

(11)

market with more than 1,600 stores, mainly in the easternmost states (ALDI, 2017). Lidl has opened the first 100 stores in dozens of cities along the East Coast and will open more in the coming year (Forbes, 2017). The German newcomer is about to bring a shockwave to the US food retail market, as it also did in the United Kingdom. Sales dropped significantly at the traditional British “big four”: Tesco, Morrisons, Sainsbury’s and Asda (Business Insider, 2016).

The outlets of hard discounters are minimally decorated and carry a relatively limited assortment comparing to traditional retailers: 1,500 SKUs versus 20,000 SKUs for an average Dutch supermarket. On store level, the staffing level and property rent are relatively low. On more operational level, the supply chain is extremely efficient: most of the stores’ products are displayed in their shipping cartons on pallets to make restocking quick and easy (Business Insider, 2016; Steenkamp and Kumar, 2012). However, ALDI and Lidl tend to improve their store design with a bigger assortment and more store comfort, but still with bottom prices. In Belgium, the “Lidl of the Future” has opened in December 2016, which has a bigger assortment, wider aisles and an improved atmosphere. The store developed from a hard discount to quality discount store with a greater emphasis on fresh items and locally sourced products (Knack, 2016; Brickmeetsclick, 2017).

1.2 Problem statement

(12)

The initial purpose of these discussed EPLs is to limit sales losses in categories where the hard discounters’ attractiveness is high and it may also discourage customers from visiting those hard discounters. In practice, EPLs can be found in most of the dry grocery and non-food product categories, carried alongside a standard and/or premium private label and national brands.

From an academic perspective, not much is known about whether EPLs are actually effective. Former research in the field of private labels showed that the introduction of EPLs can alienate certain shopper segments (Vroegrijk, Gijsbrechts & Campo, 2016). EPLs may also increase consumers’ price sensitivity and drive them toward hard discounters rather than binding them (Hansen & Singh, 2008). Carrying too many EPLs may also jeopardize the traditional retailer’s perceived quality (Ailawadi, Pauwels & Steenkamp, 2008). According to the CCO of Dutch grocery retailer PLUS, delisting the budget line is an important dilemma and delisting may help because carrying EPLs has no effect on hard discounters’ growth (Distrifood, 2015). At the Dutch private label congress, it was claimed that, on average, economy private labels have a lower quality than hard discounters’ products and therefore it can affect the premium store image of the traditional grocery retailers (TGRs) (Joppen, 2014). A recent market publication by IRI showed that, after enjoying continuous year-on-year growth for many years, private label market share by value and unit sales has fallen in Europe (IRI, 2016). The value of the European market dropped with 0.6% and market share measured by unit sales dropped with 0.5%. However, this fall is not consistent across all types of private labels: the premium private labels actually increased their market share. IRI questions the long-term viability of the economy private label.

(13)

Introducing a premium private label is advisable because this can shift the focus of a traditional retailer from price competition to quality differentiation. (Joppen, 2014; McEwen, 2014). This improves the unique selling proposition of the traditional retailer and differentiates them more from hard discounters. Consumers who belong to the Millennials group or Generation Z continue to dominate retailer’s minds. Millennials are attracted by premium private label brands, because these products offer “great quality at a great price” (Forsey, 2016).

In sum, delisting economy private labels can possibly realize:

• Reductions in supply chain costs, which includes transport costs, store handling costs and labor costs;

• Increased search efficiency, assortment satisfaction and retailer image;

• Increased reference price and higher willingness-to-pay because of delisting the cheapest alternative;

• Increased fit between assortment, customers’ needs and retailer positioning. When replaced with premium private label items, the first two bullet points do not apply.

1.3 Research question

First, we need to know whether economy private labels are important for a TGR’s customer base. It would be interesting to investigate the impact of delisting economy private labels at TGRs. Therefore, we introduce the following main research question:

“What impact does delisting economy private labels have on traditional retailer’s category sales and substitution patterns?”

This main research question will be guided by the following subquestions:

1) What is the difference of impact between delisting hedonic and utilitarian product categories in a retailer’s assortment on category turnover and volume sales?

2) What type of substitution patterns occur and to what extent do they occur?

(14)

1.4 Practical and academic relevance

This study has both practical and academic relevance. First, this study is relevant for retail companies that do not follow a hard discount strategy but offer an economy private label. Reducing the number of offered items is a common way to save operational costs. Because the effectiveness of offering EPL items is questioned, delisting them may increase quality perceptions, decrease logistic or supply chain costs, reduce price sensitivity and create a better fit with the traditional retailer’s market position. It also has implications for shelf space allocation. From an academic perspective, this study is, as far as we know, the first field experiment in which we try to investigate the impact of delisting economy private label items in-store. Former field experiments were based on other product brands. Thus, this study can be seen as a starting point and foundation for future field experiments concerning economy private labels.

1.5 Structure of the thesis

(15)

2. THEORY AND BACKGROUND

To get a better understanding of delisting, this section addresses existing knowledge and theories on delisting and assortment reductions. First, there will be elaborated on brand delisting. Second, a distinction will be made between delisting hedonic and utilitarian products. And last, substitution patterns and category sales will be explained.

2.1 Theoretical background on item delisting

The general assumption in human life is “more = better”. From a psychological view, we all seek for variety and life satisfaction (Shon, 2008). Therefore, it sounds reasonable that a big assortment satisfies the needs and wants of more customers and therefore more assortment is better (Oppewal & Koelemijer, 2005). It has been claimed that the opposite may be true (Broniarczyk & Hoyer, 2005). Concerning assortment in retail environments, it seems there is an optimal point and beyond that point, assortments can dissatisfy people. Reducing the size of an assortment would decrease search complexity, and that might induce nonbuyers in this category to start purchasing products. This may lead to positive sales effects. Food retailers increased their floor space and their assortment the last years. Next to national brands, most retailers nowadays have a three-tiered private label portfolio strategy, which contains an economy private label, standard private label and premium private label. Delisting SKUs could lead to consumer complaints, and some studies found a short-term negative sales effect (e.g. Sloot et al., 2008). However, the results of that study, which took place in the detergent category, showed that delisting positively affected search efficiency and assortment satisfaction without lowering the perceived assortment variety. A smaller assortment can also lead to higher store satisfaction: during an experiment with an in-store display with a variety of jam offerings, the authors found that a more limited offering lead to a higher conversion effect (Iyengar & Lepper, 2000). More choice seems to produce “paralysis” rather than liberation, according to “The Paradox of Choice” (Schwartz, 2004). However, it is clear that there will be a certain threshold and after that threshold, assortment reductions will unavoidably lead to consumer complaints. Nowadays, retailers often introduce their own organic label, which increases choice complexity even more. For instance, a great amount of product categories at Dutch market leader Albert Heijn have five product lines:

(16)

2) mid-quality products for an acceptable price, above EPL but below national brands, belonging to the standard private label,

3) high quality products for a price lower than national brands, belonging to the premium private label (“AH Excellent”),

4) much advertised brand products, available at different traditional grocery retailers, called the national brands (e.g. Coca-Cola, Pringles)

5) premium sustainability private label. For food products: “AH Biologisch”, for household products: “AH Eco” (Albert Heijn, 2017).

A more important result of delisting EPLs is to fill in the space gap of delisted items by launching more innovative products that better reflect the retailer’s image. The retailer can also get higher margins on these items comparing with EPL items, for example by extending the premium private label, as mentioned in the introduction section. Another important result is a changed, better balance in the retailer’s overall marketing mix. The low margins of economy private labels are being captured by, on average, large margins on the standard private label products. As a result, these types of products are offered for a too high price (Sloot, 2017). Delisting EPL items will possibly diminish the retailer’s need to capture high margins on standard private labels and then they can offer a more competitive private label assortment which can compete more with products offered at hard discounters.

(17)

whereas long-term studies refer to real assortment reduction situations (e.g. Sloot & Verhoef, 2008). With this study, we fit between these two, as we want to let consumers perceive that EPL items are removed permanently. But due to time constraints, we can only conduct this experiment on a short term. In Table 1, there is an overview of different types of studies regarding consumer perceptions on short and long term.

Table 1: Overview of some studies on assortment unavailability

Length of unavailability Consumer perception Studies

Short-term Temporarily out-of-stock Verbeke et al. (1998), Campo et al. (2000), Campo et al. (2003), Sloot et al. (2005)

Long-term Permanent delisting Drèze et al. (1994), Sloot et al. (2006), Sloot and Verhoef (2008),

Short-term Permanent delisting This study

2.2 Hedonic and utilitarian product categories

(18)

classify them according to consumer perceptions. Because both types of products can elicit different types of consumer behavior, we make the distinction of hedonic versus utilitarian products in this research. In past research, distinction between hedonic versus utilitarian products have been made in the field of price promotions, consumption and purchase intentions (e.g. Kivetz & Zheng, 2017; Kim & Kim, 2016; Kim, 2016), but this distinction has never been linked to assortment unavailability before.

2.3 Delisting EPLs’ impact on substitution patterns and category sales Delisting of the EPL items in different product

categories will both influence substitution patterns and category sales. A substitution pattern or effect happens when consumers switch some of their consumption out of goods that are now relatively more expensive and buy more other goods instead (The Economist, 2017). Such an effect can occur within the same product category or between product categories. For retail category managers, it is important to know whether and which substitution patterns occur within the same product category or between product categories. When a consumer decides to buy a substitute in another category, this will negatively affect category performance and that is what a category manager always should avoid. In this research, we focus on the

within-category substitution patterns. Within-category substitution patterns will be switching brands (e.g. choosing private label instead of economy private label). We therefore expect that a customer who, in the past, chose an EPL item to buy, will now switch to buy a standard private label item, due to a relative small price distance compared with national brands (e.g. example in Figure 1). Switching to standard private labels is desirable for a retailer. In the first place, these products have higher margins than economy private labels. Second, with standard private labels, a retailer is able to build better customer-based brand equity, which is defined as the “differential effect that brand knowledge has on the customer response to the marketing of that brand” (Keller, 1993). Standard private labels can help in claiming a clear position of the retailer as a brand in the target market.

(19)

Category sales can be defined both as the turnover or as the amount of sold SKUs of a certain product category in a store. In this study, we will investigate impact on both turnover as sales in SKUs. Regarding category sales, there are no former studies that considered delisting EPLs, but only other product types and brands. These studies have considered the impact of assortment reductions on category sales. For example, a 10% item reduction in eight product categories lead to positive sales effects (Drèze, Hoch & Purk, 1994). On the long-term, a 25% item reduction in the detergent category lead to two percent less sales, but a higher perceived search efficiency and a decrease in actual search time (Sloot et al., 2006). Delisting EPL items is a new step in research about assortment reductions in the retail landscape.

(20)

3. CONCEPTUAL FRAMEWORK AND HYPOTHESES

Based on the theory and background, the conceptual framework for this research is depicted below (Figure 2).

Figure 2: Conceptual framework

To get a better understanding of how the previously reviewed variables relate and influence each other, we now address relational and directional aspects of the construct, which will eventually result in the formulation of hypotheses.

3.1 Delisting EPLs’ impact on category sales

Delisting items from product categories could influence category sales. As mentioned in the section above, one can choose to buy another item within-category or between-category. Both will affect category sales. Research on reducing different types of items predominantly showed positive effects on category sales. Delisting EPL items will probably lead to a switch to private label items, as explained in section 2.3. We therefore expect that, also in this case, delisting EPL items will have a small negative effect on category sales in units (because not all EPL item purchases will be replaced with substitute products) but, at the other hand, a positive effect for category sales in turnover. Concerning category sales, this brings us to the formulation of the following hypotheses:

H1: Delisting EPL items in a product category will negatively influence category sales

in units.

H2: Delisting EPL items in a product category will positively influence category

(21)

3.2 Delisting EPLs’ impact on substitution patterns

While multiple researchers (e.g. Corstjens & Corstjens, 1995; Verbeke, Farris & Thurik, 1998) developed different classifications of reactions to assortment reductions situations, Sloot, Verhoef and Franses (2005) make the most extensive distinction by summarizing the six most frequent and important responses as shown in Figure 3 below. To make this clearer, the responses by Sloot, Verhoef and Franses are divided into two categories, namely ‘substitution’ and ‘non-substitution’, because it was found that consumers first decide between these two options and after that make a more detailed choice within that chosen category (Sloot et al., 2005). Because of increasing online sales, ‘channel switch’ could be a new, fourth option in the non-substitution category. For non-food products, we already see the trend of channel switch, for example when buying clothes, also called the showrooming effect (visit the store, but buy online). We expect that this trend can blow over to the fast mover consumer goods (FMCG) industry in the future, but this will take some years, as online FMCG sales are nowadays only around 3% of total FMCG sales and order fulfillment costs are still too high due to relatively small margins on food products compared to non-food products (ABN AMRO, 2016). Therefore, we expect no channel switch possibilities that affect this study.

Figure 3: Possible responses to unavailability of a product item

Substitution

(22)

3. Category switch: the consumer buys a substitute product from another product category;

Non-substitution

4. Store switch: the consumer switches store (or buys online) to purchase the item; 5. Postponement: the consumer postpones the intended purchase until the next time he visits the same store;

6. Cancellation: the consumer cancels his intended purchase completely. (Sloot et al., 2005). In this situation with EPL items, item switch is impossible, because in general, there exist no another varieties or sizes of the economy private label. This means, when a consumer decides to substitute, only option two and three exist, namely brand switch and category switch. Brand switch will be the most chosen option as the willingness to satisfy needs and wants can be high for fast mover consumer goods. In a non-substitution situation, store switch is also impossible, because the economy private label is only available at that particular store. There are similar products available, but there is no prove that these products are fully identical. A study conducted by ECR Europe (2003), which examined European assortment unavailability situations, found that brand switch is the most popular response followed by store switch and purchase postponement. In practice, brand switch here means that the largest group of consumers who had the intention to buy an EPL item, will now switch to a standard private label item.

H3: Delisting economy private label items in a product category will provoke more

substitution towards standard private label items than to other brands.

3.3 The moderating role of product type

(23)

drop of sales after delisting EPL items because people will postpone or switch stores rather than choosing a more expensive alternative in the same store. Therefore, we propose that:

H4: Delisting EPL items which are classified as utilitarian will encounter a negative

effect on category unit sales, whereas economy private label items which are classified as hedonic will encounter no negative effect on category unit sales.

H5: Delisting EPL items which are classified as utilitarian will encounter a negative

effect on category turnover, whereas economy private label items which are classified as hedonic will encounter a positive effect on category turnover.

3.4 The moderating role of sales price distance

Another moderating variable could be the sales price distance between the economy private label and standard private label. It sounds reasonable that the bigger the sales price distance, the smaller the substitution effect will be. When the sales price distance is too big, people may choose to buy the concerning product at a competitor, both other traditional grocery retailers and hard discounters.

This brings us to the following hypothesis, which is formulated as:

H6: The larger the sales price distance between the delisted economy private label

items and the sales price of the remaining substitutes, the less consumers will tend to buy a substitute.

3.5 Control variables

To exclude possible endogenous effects, sales data before, during and after the experiment are measured at three control stores. These stores are all in a rural area, carry the same brands and are owned by the same store franchisee, but have a different weekly turnover (Table 2).

Table 2: Turnover index (Experiment store=100)

Store Turnover

(24)

4. METHODOLOGY

This section will describe and explain the method for data collection, boundaries and plan for data analysis. First, we focus on the pre-test survey which is conducted in the store. After that, the methodology for the delisting experiment will be described. This section is the foundation for the results and implications which we will discuss in the sections after this.

4.1 Pre-test survey

In the theory section, we already showed that the dimension (hedonic versus utilitarian) to which a product belongs is more or less a perception of customers, rather than an established fact. We therefore should argue to what dimension the product really tends. Do Vale and Duarte (2013) developed a scale to test product categories on those two dimensions. This scale will be used for a pre-test survey, in which we ask respondents to assess different product categories and we will reproduce this on store level.

Method

To have representative survey results, store visitors of the store where we conduct the delisting experiment will be asked to fill in the pre-test survey. In total, 111 participants (male: n=24, female: n=87) in the age range of 24 to 76 years, responded to the survey in the store. All respondents had the Dutch nationality. We introduced the topic by telling respondents that the questions belong to a thesis project concerning in-store consumer behavior and that their data provision will remain confidential and anonymous. After that, we asked the participants to classify the product categories using a seven-point semantic differential scale.

Measures

(25)

label products and these products should have a significant sales share in the product category (at least 40% share in category volume). These conditions lead to a small set of chosen product categories. Regarding food products, we chose the following: mozzarella, melba toast, unflavored non-sparkling water and apple sauce. Regarding non-food products, we chose scrub sponges and aluminium foil (Table 3).

Table 3: Total sales of calendar week 2 till 14 of 2017 in the experiment store.

Product group* EPL turnover (EUR) EPL unit sales Total turnover (EUR) Total unit sales EPL turnover share (%) EPL volume share (%) Mozzarella 438.30 974 1346.06 1574 32.56% 61.88% Melba toast 301.04 424 660.28 886 45.59% 47.86% Unflavored water 550.88 438 1082.81 690 50.88% 63.48% Apple sauce 360.47 737 894.49 1233 40.30% 59.77% Scrub sponges 154.05 185 298.37 354 51.63% 52.26% Aluminium foil 368.36 265 595.66 375 61.84% 70.67%

* All the corresponding items (test items and control items) can be found in appendix B.

An overview of EPL items and visualization in the corresponding product categories can be found in Table 5 (p. 28) and in appendix A and B. The experimental items are the items which undergo the delisting procedure. The control items are the items we expect consumers will possibly buy instead of the delisted items. In the table above, we notice that EPL turnover share is pretty consistent with the volume share, the biggest ratio difference between EPL turnover and volume share is for mozzarella (respectively 32% versus 62%). This can be explained by the huge price distance between the EPL item and SPL item in this category. Outcomes

(26)

Table 4: Classification of product categories

Classification 1 _{Confidence interval}2

M 3 _SD _LB _UB Mozzarella 5.30 0.92 5.12 5.45 Melba toast 6.07 0.76 5.93 6.22 Unflavored water 1.74 0.89 1.57 1.90 Apple sauce 5.17 1.04 5.00 5.38 Scrub sponges 1.62 0.69 1.50 1.74 Aluminium foil 1.71 0.81 1.56 1.86

1_{1=Very utilitarian, 7=Very hedonic.}

2_{Confidence intervals are measured at 95% with 1000 bootstrapping samples.}

3_{A mean between [1-3.5] is classified as utilitarian, between [3.51-4.50] as neutral and between [4.51-7] as}

hedonic.

As can be observed in Table 4 above, we can classify the product categories mozzarella (M=5.30; SD=0.92), melba toast (M=6.07; SD=0.76) and apple sauce (M=5.17; SD=1.04) as being hedonic product categories, as they all have a mean above 4.50. On the contrary, the product categories unflavored water (M=1.74; SD=0.89), scrub sponges (M=1.62; SD=0.69), and aluminium foil (M=1.71; SD=0.81) can be classified as being utilitarian product categories, according to our sample, as they all have a mean below 3.51. Detailed statistical results of this pre-test survey can be found in Appendix C.

4.2 Delisting experiment

To collect data about the impact of delisting economy private labels, we conducted a field experiment in a real-life shopping environment.

Method

One average-sized brick-and-mortar store of a Dutch nationwide chain (1200 m2 _{sales area),}

(27)

Measures

Furthermore, to be able to measure category sales and substitution patterns, we chose EPL items which had a minimal revenue share of 40% in the product category. The delisted EPL items also had to be possibly replaced by customers with more expensive alternatives (e.g. private label brand and/or national brands). To control for category promotion effects and weather effects, we use sales intelligence data from three other stores, which will be placed in the so-called control group. These other stores differ in floor area, turnover and market area, but have the same assortment, are all in a rural area and are all part of the same supermarket chain.

In practice, delisting means that the chosen control items will disappear completely from the shelves. The space gap that arises will be filled fairly with the residual products that are placed in the left and right position of the delisted items. There also was a possibility to leave the space gap and place a message in front of the shelves with a reason of unavailability, but this will inevitably lead to postponement effects, and this is not the purpose of this experiment, as we want to measure category sales and substitution patterns. Even though the experiment takes place in a short-term period, we expect to find interesting and relevant effects. We use real-time sales intelligence data from a software system, implemented by the retail chain from the mentioned four stores.

(28)

Table 5: Delisted and residual items for the delisting experiment

Product category Delisted item Residual items

Mozzarella

(hedonic product category)

AH BASIC Mozzarella 125 grams Sales price €0.45 Price per kg €3.60 Index = 100* 1) AH Mozzarella 125 grams, Sales price €1.19, Price

per kg €9.52, Index = 264 2) Galbani mozzarella maxi 200 grams, Sales price €1.99, Price

per kg €9.95, Index = 276 3) Galbani mozzarella 125 grams, Sales price €1.39 Price per kg €11.12, Index = 308

4) AH Biologisch mozzarella 125 grams, Sales price €1.79, Price

per kg €14.32, Index = 397 Melba toast

(hedonic product category)

AH BASIC Melba toast 200 grams Sales price €0.71 Price per kg €3.55

Index = 100

1) Van der Meulen melbatoast naturel

100 grams, Sales price €0.69, Price per kg €6.90, Index = 194 2) Buitoni melba toast naturel 100 grams, Sales price €0.79, Price

per kg €7.90, Index = 222 3) AH Biologisch melba toast 100 grams, Sales price €0.96, Price

per kg €9.60, Index = 270

Aluminium foil

(utilitarian product category)

AH BASIC Aluminium foil 30 meters

Sales price €1.39 Price per meter €0.05

Index = 100

1) AH Aluminium foil 20 meters, Sales price €1.99, Price

per meter €0.10, Index = 200 2) Toppits aluminium foil 10 meters, Sales price €2.59, Price

per meter €0.26, Index = 520 Scrub sponges

(utilitarian product category)

AH BASIC Scrub sponges 10 pieces

Sales price €0.79 Price per piece €0.08

Index = 100

1) AH Scrub sponges 6 pieces, Sales price €0.82, Price

per piece €0.14, Index = 175 2) AH Scrub sponges soft 2 pieces, Sales price €0.89, Price

per piece €0.45, Index = 562 3) AH Scrub sponges hard 2 pieces, Sales price €0.89, Price

per piece €0.45, Index = 562 * Indices are based on price per kg

4.3 Shelf transformation

(29)

Figure 4: Shelf transformation of the four product categories

4.4 Price distance measurement

(30)

5. RESULTS

In this section, the focus will be on the results of our field study which is proposed in the previous section. First, we provide an overview of the conducted field experiment. Second, we distinguish between effects on category sales and effects on substitution patterns. Section 5.2 describes the effects on category sales. For this, descriptive data of the dataset is computed. Lastly, using SPSS statistical software, linear regression analysis is performed and results are described for the effects on substitution patterns.

5.1 Introduction

Due to time limits, the delisting experiment took place in two categories which have been classified as being a hedonic product category and two categories which have been classified as being a utilitarian product category. According to the outcomes in section 5.1, we chose the two hedonic product categories with the highest mean score and the two utilitarian product categories with the lowest mean score. This resulted in a choice set of the following product categories: melba toast (M=6.07), mozzarella (M=5.30), aluminium foil (M=1.71), and scrub sponges (M=1.62). The shelves were transformed for a period of three weeks before measuring the data. In Table 4, we showed the delisted items for this experiment and the corresponding residual items including sales price, price per unit (kilograms or pieces) and price distance as index between delisted item and corresponding item, based on the price per unit. The average price distance between economy private label and consecutive priced item is above 200%. This means that a consumer who regularly buys an item in one of the four categories now has to pay twice as much for the cheapest alternative. As mentioned, we expected people to switch to an alternative with the smallest price distance. In all tables and figures, we distinguish different phases, described in Table 6. In Table 7, we computed average sales of residual products in the experimental weeks to give a general overview of collected data.

Table 6: Time block, weeks and phases

Time block Calendar weeks of 2017 Phase

Pre-test weeks 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 Shelves in original state

(31)

5.2 Descriptive data analysis

The dataset has been checked on missing data and outliers. No missing data and outliers have been found. Table 7 contains descriptives to have an overview of turnover and unit sales in the product categories before, during and after the experiment. The turnover and unit sales are weekly averages, rounded down. For all collected data, consult Tables 8-11. Relative changes are shown between brackets. First, we notice that turnover has increased for all four product groups during the experiment. The biggest increase in turnover was caused by delisting the cheapest mozzarella item, possibly because of the big price distance between EPL and SPL item. After the experiment, the turnover in the hedonic product categories still remains higher than before the experiment. In the utilitarian product categories, turnover decreases below the initial level after restoring the shelves to original state. Second, we notice that unit sales have a different impact per product group. In the mozzarella product group, unit sales drop during the experiment. However, these differences are not significant (see section 5.3).

Table 7: Descriptive data of product groups, turnover and unit sales before and during experiment. Note that all numbers are weekly averages. Product group Turnover before experiment Turnover during experiment Turnover after experiment Unit sales before experiment Unit sales during experiment Unit sales after experiment Mozzarella €103.54 €146.07 (+41.1%) €119.20 (-18.4%) 121 102 (-15.7%) 114 (-5.8%) Melba toast €50.79 €51.76 (+1.9%) €61.24 (+18.3%) 68 80 (+17,6) 78 (+14.7%) Scrub sponges €22.95 €24.96 (+8.8%) €18.31 (-26.6%) 27 28 (+3.6%) 23 (-14.8%) Aluminium foil €45.82 €61.17 (+30.2%) €34.12 (-44.2%) 28 29 (+3.6%) 20 (-28.6%)

Delisting EPL items decreases the choice set. The total turnover increases because EPLs are the cheapest items in the product category. However, some people might postpone their purchase or switch stores, because they do not accept to pay a higher price for an alternative. The next section aims at specific results per product group.

(32)

5.2.1 Experiment group 1: mozzarella

Mozzarella cheese is a hedonic product group with the biggest changes in both turnover and unit sales when comparing those numbers before and during the experiment. There are two reasons for this. In the first place, the price distance between economy private label and standard private label is the biggest, namely 264 (EPL= €0.45 vs. SPL=€1.19). Second, regular mozzarella buyers may remember the price and existence of the EPL item and find higher prices of the residual items unjustifiable. This effect may be strengthened because most mozzarella items had a similar package of 125 grams, so price differences are more quickly noticed. Based on these short-term results, delisting the EPL item of mozzarella does not negatively influence category turnover. Most people shift from AH BASIC to AH brand, with an increase in average weekly sales from 12 units to 47 units (+291%), but which is also an enormous shift in turnover from €0.45 to €1.19 (+164%) per item. As the margins of SPLs are generally higher than of EPL products, this seems a profitable situation.

Table 8a: Turnover (€) and unit sales (Qnt.) of mozzarella cheese items

AH BASIC Mozzarella

AH Mozzarella Galbani

(33)

Tabel 8b: Share changes before experiment (pre), during experiment (test), and after experiment (post)

Brand Unit sales share (pre) Turnover share (pre) Unit sales share (test) Turnover share (test) Unit sales share (post) Turnover share (post)

AH BASIC 61.88% 32.56% N/A N/A 48.25% 20.98%

AH 11.05% 15.38% 46.10% 38.56% 12.28% 14.12%

Galbani 13.79% 22.01% 30.52% 29.47% 17.32% 23.27%

Galbani maxi 9.66% 22.47% 18.18% 25.43% 17.11% 32.90%

AH Biologisch 3.62% 7.58% 5.19% 6.54% 5.04% 8.73%

Figure 5: Visualization unit share (left) and turnover share (right) per phase

5.2.2 Experiment group 2: melba toast

For melba toast, there is no private label alternative available. The choice set of consumers consist of two national brands, Buitoni and Van der Meulen, and an organic private label brand item. Here, most consumers switch to Van der Meulen, which has the smallest price distance compared to AH BASIC (PD=194).

Table 9a: Turnover (€) and unit sales (Qnt.) of melba toast items

AH BASIC Melba toast

(34)

12 25.56 36 14.22 18 6.21 9 0 0 45.99 63 13 20.59 29 10.27 13 7.59 11 2.88 3 41.33 56 14 29.82 42 22.12 28 7.59 11 7.68 8 67.21 89 15 0 0 44.24 56 21.39 31 10.56 11 76.19 98 16 0 0 18.96 24 13.11 19 5.76 6 37.83 49 17 0 0 29.23 37 37.95 55 1.92 2 69.10 94 18 19.88 28 26.86 34 14.49 21 3.84 4 65.07 87 19 20.59 29 11.85 15 22.08 32 2.88 3 57.40 79 20 23.43 33 11.85 15 21.39 31 1.92 2 58.59 81 21 19.88 28 11.85 15 14.49 21 1.92 2 48.14 66

Tabel 9b: Share changes before experiment (pre), during experiment (test), and after experiment (post)

AH BASIC 47.86% 45.59% N/A N/A 37.70% 36.55%

Buitoni 27.20% 28.83% 48.55% 50.48% 25.24% 27.23%

Van der Meulen

19.53% 18.63% 43.57% 39.56% 33.55% 31.61%

AH Biologisch 5.42% 6.94% 7.88% 9.96% 3.51% 4.61%

Figure 6: Visualization unit share (left) and turnover share (right) per phase

5.2.3 Experiment group 3: scrub sponges

(35)

Table 10a: Turnover (€) and unit sales (Qnt.) of sponge scourer items AH BASIC Scourers AH Scourers soft (2 pcs.) AH Scourers hard (2 pcs.) AH Scourers (6 pcs.) Total Week € Qnt. € Qnt. € Qnt. € Qnt. € Qnt. 2 13.43 17 1.78 2 3.56 4 5.74 7 24.51 30 3 11.06 14 1.78 2 3.56 4 4.10 5 20.50 25 4 11.06 14 1.78 2 1.78 2 9.84 12 24.46 30 5 8.69 11 5.34 6 2.67 3 9.02 11 25.72 31 6 13.43 17 1.78 2 4.45 5 7.38 9 27.04 33 7 12.64 16 0.89 1 1.78 2 3.28 4 18.59 23 8 11.06 14 6.23 7 1.78 2 2.46 3 21.53 26 9 12.64 16 1.78 2 3.56 4 4.92 6 22.90 28 10 18.17 13 0.89 1 4.45 5 4.10 5 27.61 24 11 20.54 26 1.78 2 4.45 5 4.10 5 30.87 38 12 9.48 12 5.34 6 1.78 2 4.92 6 21.52 26 13 11.85 15 4.45 5 0.89 1 7.38 9 24.57 30 14 0 0 1.78 2 2.67 3 4.10 5 8.55 10 15 0 0 4.45 10 0.89 1 18.04 17 23.38 28 16 0 0 6.23 8 2.67 3 18.04 17 26.94 28 17 0 0 3.56 4 4.45 5 15.58 19 23.59 28 18 3.16 4 6.23 7 1.78 2 4.10 5 15.27 18 19 8.69 11 2.67 3 1.78 2 8.20 10 21.34 26 20 9.48 12 2.67 3 1.78 2 8.20 10 22.03 27 21 7.90 10 2.67 3 1.78 2 7.38 9 19.73 24

Table 10b: Share changes before experiment (pre), during experiment (test), and after experiment (post)

AH BASIC 52.26% 51.63% N/A N/A 38.95% 37.30%

AH (Soft 2pcs) 11.30% 11.93% 26.19% 19.27% 16.84% 18.17% AH (Hard 2pcs) 11.86% 12.52% 10.71% 10.84% 8.42% 9.09% AH (6pcs) 24.58% 23.91% 63.10% 69.90% 35.79% 35.57%

(36)

5.2.4 Experiment group 4: aluminium foil

As expected, in the aluminium foil product group, there is a huge increase in turnover for the private label brand. After delisting AH BASIC, the private label brand increases its unit sales share and turnover share to more than 80%.

Table 11a: Turnover (€) and unit sales (Qnt.) of aluminium foil items

AH BASIC

Aluminium foil AH Aluminium foil

Toppits aluminium foil

Total Week € Qnt. € Qnt. € Qnt. € Qnt. 2 25.02 18 11.94 6 5.18 2 42.14 26 3 34.75 25 9.95 5 5.18 2 49.88 32 4 23.63 17 7.96 4 0 0 31.59 21 5 50.04 36 27.86 14 5.18 2 83.08 52 6 18.07 13 19.90 10 2.59 1 40.56 24 7 16.68 12 9.95 5 0 0 26.63 17 8 33.36 24 11.94 6 2.59 1 47.89 31 9 29.19 21 17.91 9 0 0 47.10 30 10 27.80 20 13.93 7 2.59 1 44.32 28 11 29.19 21 21.89 11 2.59 1 53.67 33 12 26.41 19 7.96 4 5.18 2 39.55 25 13 22.24 16 9.95 5 5.18 2 37.37 23 14 31.97 23 19.90 10 0 0 51.87 33 15 0 0 51.74 26 15.54 6 67.28 32 16 0 0 39.80 20 10.36 4 50.16 24 17 0 0 55.72 28 10.36 4 66.08 32 18 4.17 3 13.93 7 7.77 3 25.87 13 19 12.51 9 29.85 15 0 0 42.36 24 20 11.12 8 29.85 15 7.77 3 48.74 26 21 11.12 8 21.89 11 0 0 33.01 19

Table 11b: Share changes before experiment (pre), during experiment (test), and after experiment (post)

Brand Unit sales share (pre) Turnover share (pre) Unit sales share (test) Turnover share (test) Unit sales share (post) Turnover share (post)

AH BASIC 70.67% 61.84% N/A N/A 34.15% 25.95%

AH 25.60% 32.07% 84.09% 80.24% 58.54% 63.69%

Toppits 3.73% 6.09% 15.91% 19.76% 7.32% 10.36%

(37)

5.2.5 Turnover effects on the long-term

Based on Table 7, which shows the turnover of product category before, during and after the experiment, we can compute the effects on the bottom line caused by delisting EPL items. By extrapolation, a gross estimate shows us that the experiment store misses about €3164,72 of turnover in only those four product categories per year (Table 12). Due to confidentiality, margins could not be obtained from the business intelligence system.

Table 12: Turnover effects extrapolated

Product category Average weekly turnover during experiment Extrapolating (x 52 weeks)

Mozzarella €42.53 € 2211.56 Melba toast €0.97 €50.44 Scrub sponges €2.01 €104.52 Aluminium foil €15.75 €798.20 TOTAL €60.86 €3164.72 5.3 Operationalization of variables

The variables below are part of the dataset from which the hypotheses are tested (Table 13).

Table 13: Overview of variable operationalization

Variable Notation Operationalization

Dependent variables

Substitution patterns

TurnoverIndex VolumeIndex

True turnover divided by expected turnover in certain week

True unit sales divided by expected unit sales in certain week

Independent variables

Characteristics which identify what is influencing substitution patterns

ExpItem EPL SPL NB BioPL PG_Mozzarella PG_Melba PG_Schuurspons PG_Aluminiumfolie Teststore Pretestphase

Dummy indicator, 1 equals when the product item is delisted in the experiment store, otherwise 0

Dummy indicator, 1 equals when the product item is an EPL, otherwise 0 Dummy indicator, 1 equals when the product item is a SPL, otherwise 0 Dummy indicator, 1 equals when the product item is a NB, otherwise 0 Dummy indicator, 1 equals when the product item is organic PL, otherwise 0 Dummy indicator, 1 equals when the product item belongs to mozzarella, otherwise 0

Dummy indicator, 1 equals when the product item belongs to melba toast, otherwise 0

Dummy indicator, 1 equals when the product item belongs to scrub sponges, otherwise 0

Dummy indicator, 1 equals when the product item belongs to aluminium foil, otherwise 0

Dummy indicator, 1 equals when the store is the experiment store, otherwise 0

(38)

Testphase Posttestphase VerwOmzet VerwVolume Marktaandeel_Volume Marktaandeel_Omzet

store is the experiment pre-test phase, otherwise 0

Dummy indicator, 1 equals when the store is the test phase, otherwise 0 Dummy indicator, 1 equals when the store is the post-test phase, otherwise 0 Expected weekly turnover of product item, calculated as average turnover of that item in the pre-test phase

Expected weekly unit sales, calculated as average unit sales of that item in the pre-test phase

Average market share of product item based on unit sales in the pre-test phase Average market share of product item based on turnover in the pre-test phase

Moderating variables Product type Price distance Type_Hedonisch Type_Utilistisch Pricedistance PricedistanceCorrContent

Dummy indicator, 1 equals when the product group is seen as hedonic, otherwise 0

Dummy indicator, 1 equals when the product group is seen as utilitarian, otherwise 0

Price distance w.r.t. delisted item in that product category (index)

Price distance w.r.t. delisted item in that product category (index), corrected to package content

Control variables

Other stores Store2 Store3 Store4

Store 2, where delisting has not taken place

Before estimating the coefficients, an initial analysis on the means shows us that effects will certainly occur. The dataset includes data from 16 different stock keeping units, which come from four different product categories, and contains data of four traditional grocery retail stores, from which one is the experimental store. During the test phase, an increase in turnover is shown in the bold marked columns (Table 14). From the other side, this did not affect unit sales negatively. In the post-test weeks, it seems that both turnover and unit sales revert slowly to the situation before the experiment.

5.4 Hypothesis testing

(39)

5.4.1 Effect of delisting on category unit sales

Table 14: Mean turnover and unit sales

An overview of the mean unit sales per store is shown (Table 14/15). It is remarkable here that all stores have an increase in mean unit sales during test phase, except Store 1 where the experiment took place. This means that the experiment caused a decrease in average unit sales from 15.33 to 14.06. In order to analyze whether or not unit sales differ per phase, we performed a one-way analysis of variance of the unit sales per week by test phase for all four stores (Table 16). This one-way ANOVA was not significant. Another one-way ANOVA for the experiment store only (Table 17), also shows no significant results, F(1, 254) = .205, p = .325.

Table 15: Mean unit sales per phase in the four stores

Store Phase Mean SD

Store 1 Pre-test phase 15.33 17.995 Test phase 14.06 15.183 Post-test phase 14.72 13.693

Store 2 Pre-test phase 8.39 10.322

Test phase 9.67 11.820

Post-test phase 8.39 10.907

Test phase 8.73 10.191

Test-phase 8.00 10.884

Table 16: One-way ANOVA of unit sales per week by test phase

Store Source df SS MS F p

Store 1 Between groups 2 70.03 35.01 .124 .442 Within groups 317 89675.86 282.89

Total 319 89745.89

Total 319 36181.89

Total 319 34596.39

(40)

Table 17: One-way ANOVA of unit sales per week by test phase in the experiment store

Total 255 77925.75

For hypothesis 1 (H1), the expected effect was that delisting EPL items in a product category will negatively influence category sales in units. According to the descriptive data analysis and ANOVA results, we cannot accept H1. The analysis shows that unit sales do not drop significantly (all p-values are > .10).

5.4.2 Effect of delisting on category turnover

Regarding category turnover, we again computed a means table (Table 18). The mean turnover during the test phase is higher than the pre-test phase in all four stores. This is caused by a higher weekly store turnover in the control stores, which was between +2% and +4% compared to the pre-test phase. In order to analyze whether or not turnover differs per phase, we performed a one-way analysis of variance of the turnover per week by test phase for all four stores (Table 19). The next table shows a one-way ANOVA for the experiment store only (Table 20), which is highly significant: F(1, 254)=3.071, p = 0.04.

Table 18: Mean turnover per phase in the four stores

Store Phase Mean SD

Store 1 Pre-test phase 13.94 10.880 Test phase 17.57 19.489 Post-test phase 14.50 10.987

Store 2 Pre-test phase 7.82 7.320 Test phase 9.43 7.158 Post-test phase 7.70 6.676

Store 3 Pre-test phase 7.39 6.441 Test phase 8.56 6.062 Post-test phase 6.66 6.163

Store 4 Pre-test phase 5.79 5.986 Test-phase 7.45 6.680 Post-test phase 5.13 5.183

Table 19: One-way ANOVA of turnover per week by test phase

Store 1 Between groups 2 512.95 256.48 1.627 .099 Within groups 317 49957.91 157.60

Total 319 50470.86

Total 319 16417.01

Total 319 12806.90

(41)

Table 20: One-way ANOVA of turnover per week by test phase for experiment store

Total 255 42865.56

For hypothesis 2 (H2), we expected that delisting EPL items in a product category will positively influence category turnover. According to the descriptive data analysis and one-way ANOVA, we accept H2. An increase in category turnover was expected because the sales prices of SPL and NB items are way higher than EPL items. But the data also shows that more consumers choose to buy an item from this product category. Concerning the post-test effects, the unit sales share and turnover share are partially restored to their initial values. The item to which most people switched during the experiment still has more share than before the experiment and this applies for all product categories. Consumers seem to be satisfied with a smaller assortment setting as effects are measured for all product groups. This said, the experiment both has positive effects during and after the experiment. However, it is not clear if these post-test effects will continue on the longer term as we were unable to draw results of a bigger set of post-test weeks.

5.4.3 Substitution patterns towards residual items

Delisting economy private labels resulted into a smaller choice set for consumers. According to hypothesis 3, we expected that delisting those items will provoke more substitution towards standard private label items than to other brands. Reason for this is that standard private label items have a smaller price distance with EPL items than national brands and organic PL items Second, these SPL items also have the store brand logo on the package (AH instead of AH BASIC) so the products look relatively similar. In order to analyze whether or not delisting leads to substitution to SPL items, we performed a regression analysis of brand choice (coded as several dummy variables) on volume index. The results of this regression, R2_{= 0.38, F(3,}