RFID-enabled visual merchandising in apparel retail

Abstract

RFID item-level tagging is a powerful tool for improving inventory accuracy, which is a prerequisite for both omnichanneling strategies and store floor replenishment from the backroom. This paper explores how retailers can leverage RFID item-level tagging further to optimise product merchandising in the store area, to map store floor performance and, ultimately, to increase sales.

To this extent, we introduce new original metrics that are enabled by RFID item-level tagging. These metrics aim at monitoring the net profit (NP) of sales areas (store money mapping) and classifying product lines (PL), understood to be a combination of model and colour, in terms of generated profits and costs to display (CD). We introduce three PL categories—money makers, indifferent and money losers—based on the NP/CD ratio, and we propose different strategies for each category to aid store managers in optimising sales performances.

We have been testing the proposed approach for more than a year in three flagship stores of a major fashion retailer. The results have shown consistent sales increases in RFID managed stores compared to non-RFID test stores.

The proposed RFID visual merchandising is specifically tailored for small-sized fashion retail and luxury outlets. Usually located in very expensive areas, these stores have limited display space available. Therefore, retail managers seek to understand what is actually worth displaying and where, to reduce costs and sustain consistent store sales. Future works could also address whether the proposed approach could boost sales in large retail outlets.

Keywords

RFID visual merchandising retail money mapping apparel supply chain management

1 Introduction

In fashion and apparel retailing, consumer buying decisions seldom rely on large-scale information or articulated decision-making processes. Rather, when shoppers enter a store to purchase an item, their decision-making is impulsive, often based on strategic visual presentations and merchandise assortments that are presented in the store area by the retailer.

Through visual merchandising, retailers strive to present an attractive sales environment that, on the one hand, has an emotional impact on customers and keeps them in the store as long as possible, and on the other, triggers the buying process through well-designed product displays. A high level of excitement and interest on the part the consumer reduces his or her ability to assess the purchase and increases the likelihood of impulse buying.

Thus, a thorough understanding of store layouts, furniture, colours, sounds, and people is crucial for retailers. Even more important are the products and how, when and where they are displayed. These factors have a significant impact on consumers’ perception of brand image and the retailer’s ability to keep the customer engaged in the store—and ultimately on triggering impulse sales.

Currently, a large amount of literature on visual merchandising as well as lengthy practitioner experience are available.

In apparel stores, arrangement decisions, that is, which/how many stock keeping units (SKUs) should be exposed not only in windows and corners but also on shelves and racks, is usually carried out by store associates, retail managers or visual merchandisers, the latter being an ad hoc position that some retailers have created for this purpose. However, these activities are rarely tracked, and there is little information regarding which SKUs are shown, in what quantities, where, and when in the IT legacy systems. This is because retail associates by nature pay less attention to tracking, as they are more focused on sales and view other activities such as tracking as a sort of ‘burden’ that absorbs time without adding consistent value.

Moreover, there is a technology gap that prevents retailers from accurately and efficiently tracking the success of their displays. Barcodes suffer from inaccuracies related to manual operation. For example, if a store associate scans the wrong code or simply forgets to do scan an item, inaccurate data are fed into the system and incorrect decisions may be taken in consequence. It is largely agreed that once operatives no longer trust the accuracy of the data and information provided, even the most sophisticated system is dropped. In other words, visibility is the main issue that retailers must tackle in implementing and monitoring their visual merchandising strategies. The key challenge is obtaining consistent information on product displays and correlating it to sales.

Evidence from the field (see the following paragraphs) showed us that even in luxury fashion retail stores, customers buy what they see in the store. In fact, up to 86% of sales are triggered by the models/colours that are displayed in the store area, while only 14% involve models/colours that are not displayed.

RFID item-level tagging can easily close this technology gap and provide both the accuracy and ease of use that retailers need to support their visual merchandising strategies and store display execution.

Apparel retailers first deployed RFID in the early 2010s to better replenish items from the backroom, a process that has been found to be the weakest link of the whole supply chain. Then, some years later, a second wave of RFID emerged, helping apparel retailers worldwide to increase their inventory accuracy to superior levels. This is crucial since inventory accuracy is a prerequisite for enabling omnichanneling, which boosts both on-site and online sales. This paper shows that a third wave of RFID adoption could come, with retailers utilising RFID not only to increase inventory accuracy and avoid out of stocks but also to track what, how many, where and when a specific SKU is displayed and correlate this information about displays to sales by determining what is worth displaying and what is not.

The main ideas the authors put forward in this paper are as follows:

In apparel retailing, only PLs that generate positive margins (roughly speaking this is sales margins minus the total costs of the space they consume; this point will be thoroughly addressed later) are worth displaying. PLs that do not should either be replaced or worked on to increase their contribution. The longer the PLs have been exposed, the truer this equation becomes;

Not all store areas are equal. The higher the store area value (for instance, measured in terms of the number of customers visiting it), the higher the expected sales from a PL that is displayed.

The remainder of the paper is organised as follows. In the next section, we explore relevant scientific literature on RFID and visual merchandising in fashion retailing. Then, we detail our original model for RFID-enabled visual merchandising. A case study is then described, showing the potential benefits. Concluding remarks, implications, limitations and future work suggestions close the paper.

2 Literature review

In the following sub-paragraphs, we examine the literature related to visual merchandising and RFID item-level tagging in retail.

2.1 RFID in fashion and apparel retailing

As stated by Azevedo and Carvalho (2012), fashion and apparel are among the most dynamic supply chains, characterised by highly unpredictable demand, product range explosion and extremely short product life cycles, usually weeks. Apparel products may become suddenly obsolete and suffer steep price markdowns due to ultra-rapid changes in customer needs, driven by fashion trends, weather and other unpredictable boundary conditions. At the same time, sudden demand spikes can create out of stocks and lost sales.

In such a variable environment, effective supply chain management becomes a competitive edge. Randall et al. (2011) states that fashion companies should strive to smooth the product flow in the supply chain. They should try to balance the need with anticipated demand, reduce production costs and stockpile products downstream to better serve customers. There are risks associated with high market mediation costs, including widespread obsolescence markdowns or lost sales caused by demand surges for certain SKUs.

According to Fisher (2003), best-in-class fashion supply chains are thus those that are able to trim market mediation costs, that is, serving customers with exactly what they want (style/colour/sizes), where they want and when they want it, and optimise the associated physical costs (procurement, operations, distribution, holding stocks) through supply chain management techniques enabled by product flow synchronisation, full supply chain full transparency and data sharing. Above all, postponement, inventory pooling and transhipment, economies of scales and scope based on volumes are important, as well as the ability to respond quickly and agilely (Christopher et al., 2004; Brynjolfsson et al., 2013).

The earliest RFID item-level tagging deployments in apparel retailing date back to the early 2000s. According to Madhani (2011), the first RFID projects were primarily focused on upstream activities in an attempt to reduce downtime in inbound/outbound logistics processes. The main RFID features exploited at this stage were full process automation and accuracy improvements. However, as observed by Moon et al. (2008) and Bottani et al. (2009), RFID item-level tagging soon emerged as an extremely powerful tool for better supply chain management, whose value is at its highest in downstream processes, particularly in retailing.

Bertolini et al. (2012) thoroughly addressed the return on investment that RFID item-level tagging can spark in apparel and fashion retailing. The authors provided quantitative results regarding logistics and store processes and suggested different areas and processes that could be affected by RFID, including both former operational aspects (e.g. shipping and receiving, inventory counts) and innovative strategic issues related to garments try-ons, store replenishment, inventory management, customer satisfaction and sales volume.

The main benefits related to item-level RFID adoption in fashion retailing have been addressed also by Al-Kassab et al. (2011), Delen et al. (2007), Hardgrave et al. (2009) and Hargrave (2009):

superior accuracy in store inbound processes and inventory counts

availability of new sets of data (i.e. backroom vs. store area inventories)

superior accuracy along with real-time inventory visibility, which makes it possible to reduce out of stocks, both in the shop floor area and in the store as a whole

enabling of new business models (omnichanneling)

This is why Rizzi et al. (2015) state that apparel and fashion retailing is one of the industries relying increasingly on RFID use, with applications ranging from supply chain management to marketing, from shop floor management to brand protection and from customer experience to omnichanneling. The RFID Centre at the University of Auburn, formerly at the University of Arkansas, is widely recognised as one of the most productive research groups with regard to RFID item-level tagging in apparel and retailing.

The impact of RFID on out of stocks, inventory accuracy and omnichanneling has been addressed in the literature (Goyal at al., 2016; Hardgrave et al., 2009; Hardgrave, 2012). To our knowledge, however, few or no authors have examined the impact of RFID item-level tagging on visual merchandising.

Therefore, the purpose of this paper is to bridge this gap, investigating how accurate, real-time, selective data about the location of items in the store area enabled by RFID can help managers to make better decisions regarding displays, boost store sales and increase store turnover.

2.2 Visual merchandising

Retail scientists define visual merchandising as the ability to communicate a product and/or a brand to customers through a sensorial experience. This visual message should be appropriately delivered to motivate a positive psychological and behavioural outcome and ultimately lead to purchase (Kerfoot et al., 2003).

Visual merchandising enhances the attractiveness of a store and its perceived image from the customer’s viewpoint. Through visual merchandising, retailers create an attractive environment that increases the customer’s desire to enter a store, to stay in the store area as long as possible and the likelihood of impulsive purchase, giving the customer a feeling of excitement and pleasure in return (Kerfoot et al., 2003; Law et al., 2012; Mehta et al., 2013; Chang et al., 2014).

In order to impact buyers’ emotions and stimulate purchasing behaviour in stores, retailers leverage a combination of different elements, such as people, store layouts and equipment, materials, perfumes, colours, lighting, sounds and graphics—but above all, in-store product displays (Ebster, 2011).

Customers react when they are faced with a different combination of stimuli, and the overall picture is always greater than the sum of its parts. Mehrabian and Russell (1974) first proposed a conceptual model based on the stimulus–organism–response (SOR) paradigm to study how the combination of many visual merchandising elements create a stimulus (S) on the individual organism (O)—that is, the customer—that affects his or her emotional state and therefore his or her positive or negative response (R) to purchase.

Among the different types of stimuli in visual merchandising, windows and in-store product displays play a pivotal role. Morgan (2008) underscored the function of either a window or in-store display: ‘if Selfridges were a magazine, the windows would be the front cover’. A survey carried out by Gudonavičienė et al. (2015) suggests that one of the most powerful tools for impulse buying in apparel and fashion retail is the combination of windows displays and in-store layouts and product displays. Stylish, original shop windows attract customers and compel them to stop, look and enter the store. Once inside, charming in-store product displays in strategic store areas as well as glamorous total looks increase the product/brand value perception and therefore the likelihood of impulsive purchases.

Practitioners usually use conversion rate as a key metric, which relates the number of sales to the traffic in the store. In the apparel industry, Footfall (2016) determined that the combined online and in-store conversion rate in 2015 was 14.9% in North America and 13.9% for Europe, APAC and North America (combined).

Today, to compute conversion rates, the vast majority of retailers rely on daily point of sales data, provided by checkout scans, as well as information on the number of customers entering the store, which can be obtained through ad hoc customer-counting systems. However, while conversion rate ‘as it is’ may provide some valuable insights, it actually delivers only a partial view of the whole picture since it fails to relate sales to the visual merchandising policies adopted by store managers. In other words, as windows and in-store product displays have a pivotal impact on emotional purchases, the conversion rate does not provide crucial information that is needed to optimise visual merchandising actions. Managers need to answer a number of important questions: What combination of windows and in-store displays results in the highest conversion rates? What happens to conversion rates when product displays are changed? Once products have been given visibility by ‘lending’ them a certain amount of valuable store space, are they able to generate consistent store turnover? That is, is it worth keeping a product on displaying? Should it be replaced with a more profitable one?

In traditional retail environments, these questions remain unanswered, as accurate, punctual and specific pieces of information about product displays cannot be gathered. To get these answers, visual merchandisers would have to manually input daily in-store and windows displays into legacy systems, but this is impossible both in terms of efficiency and accuracy. However, thanks to RFID technology, punctual, accurate data can now be easily and smoothly generated and matched with sales, giving store managers valuable insights regarding what is worth displaying to increase conversion rates and boost sales.

3 The RFID visual merchandising approach

In this paragraph, we detail how RFID can enable effective visual merchandising policies. Our assumptions have been tested in three flagship stores of a major Italian retailer, to test both the robustness and practical implications of our model. For confidentiality reasons, punctual values have been omitted.

3.1 Theoretical model

3.1.1 Net profit vs. display costs

The store sales area, as well as a portion of it, can be considered as a kind of a firm, where turnover, once direct costs have been covered, should generate a profit that covers at least the indirect costs. In the following paragraphs we will examine the difference in store profit and loss lines in more detail.

The only profit line for a store is sales. Here, we can consider either gross or net sales. If net sales are taken into account, sell-out (meaning the product price multiplied by the quantity of products sold) has to be reduced to consider product sell-in costs (the respective logistics costs—procurement, manufacturing and distribution—to bring the same product to the store). We will follow the latter approach in the following paragraphs.

The store’s net profit NP can be obtained by summing up daily net profits as NP_j for j = 1,... M, where M is the number of open days in a year. Thus, $NP = \sum_{j = 1}^{M} {NP}_{j} [\frac{€}{year}]$ (1) Again, daily net profit NP_j can be determined by summing up the net profits of single store areas NP_k,j, with k = 1,... ,K being the single store areas (floor, corner, shelf, window). The following Equation (2) can be applied: ${NP}_{j} = \sum_{k = 1}^{K} {NP}_{k, j} [\frac{€}{day}]$ (2) Finally, Equation (3) can be adopted to calculate the area net profit NP_k,j ${NP}_{k, j} = \sum_{i = 1}^{N} x_{i, j, k} * \frac{1}{\sum_{k = 1}^{K} x_{i, j, k}} * S_{i, j} * (p_{i, j} - {Cu}_{i}) [\frac{€}{day}]$ (3) where

PL_i,j is the single product line i = 1,... ,N – that is, a given model/colour/all sizes – sold on day j, j = 1,... ,M

S_i,j [item] sales of product line PL_i on day j

x_i,j,k = 1 if Q_i,k,j>0, that is, PL_i is displayed in store area k on day j

x_i,j,k = 0 if Q_i,k,j = 0, that is, PL_i is NOT displayed in store area k on day j

The backroom has a different behaviour: x_i,backroom,j = 0 if ∃ at least k^* < > backroom |x_i,k^*,j = 1

p_i,j is the unit price of PL_i on day j [€/item]

Cu_i [€/item] is the unit purchase price of the sell-in

Then, $NP = \sum_{j = 1}^{M} \sum_{k = 1}^{K} \sum_{i = 1}^{N} S_{i, j} * \frac{x_{i, k, j}}{\sum_{k = 1}^{K} x_{i, j, k}} * (p_{i, j} - {Cu}_{j})$ (4) It can be stressed that

RFID item-level tagging enables punctual computation of x_i,j,k coefficients in Equation (4), which means the ability to allocate a single sale to the store area/s where that product line was displayed

In Equation (4) there is a direct proportion between net profit, sales and product margins, which is well known to retail managers, and this will be recalled later.

With regard to store costs, the following considerations are important.

First, as mentioned in the literature review paragraph, the store area works as a big display window for the vast majority of fashion retail stores. Items in the store area are managed through visual merchandising strategies to work as “sales generators” based on the emotions and shopping experience they are able to elicit. These are in fact the models/colours the customer can see, touch and feel and which trigger purchase intention. Usually, the displayed items are not sold; rather, another item, the same model but perhaps in a different colour/size, is taken from the backroom and given to the customer to complete a sale. In this way, the store area remains untouched, and the same customer experience can be delivered to the following customers.

Second, any store is characterised by an annual total cost [€ /year], comprising the amortisation of capital expenditures (Capex) and yearly operating expenditures (Opex). Capex encompasses owned facilities, furniture and technologies, while Opex usually comprises labour, rents, general services, etc. We do not consider sell-in costs as part of Opex, as they have already been taken into account in profits.

Store costs should be allocated only to items displayed in the store area because they are the ones that generate sales and therefore revenues. These revenues, once capital and operating expenditures are covered, generate a net profit. As tackled later, even for a fashion retail boutique, the vast majority of customers buy the models/colours they can see and touch in the store area. The likelihood of selling a product line that is not displayed is very small.

Based on these premises, starting from annual total cost TC (Capex + Opex), we can easily calculate the daily store cost TC_j. For the sake of ease, assume that costs do not change against days. Therefore, $TC = \sum_{j = 1}^{M} {TC}_{j} [€ / year]$ (5) ${TC}_{j} = \frac{TC}{M} [€ / day]$ (6)

Where M is the number of store open days per year.

Different approaches can be utilised to share store costs among store areas. The simplest method is to use space (that is, square meters) as a driver for cost allocation. Thus, the bigger the zone area, the higher its cost. A more sophisticated approach may take into account the average number of customers visiting an area: then, the higher the number of customers, the more expensive the area and thus the greater the cost share that should be allocated there.

For the sake of simplicity, let us adopt an approach based on space only, where

A_k is the space of the store area k [m²]

A is the overall store area [m²]

Combining with Equation (6), the total cost for store area k is ${TC}_{j, k} = \frac{TC}{M} * \frac{A_{k}}{A} [€ / day]$ (7)

Starting from the total cost of area k on day j, we can easily obtain the cost to display CD_i,k,j [€ /PL], which is the cost to display product line PL_i in area k on day j., where

Q_i,k,j [item/day] is the number of items of PL_i displayed in area k on day j

Q_k,j [item/day] is the number of items displayed in area k on day j

PS% _i,k,j [% ] is the percentage of space in area k given to product line PL_i on day j

$PS %_{i, k, j} = \frac{Q_{i, k, j}}{Q_{k, j}}$ (8) The cost to display can therefore be calculated as ${CD}_{i, k, j} = {TC}_{j, k} * \sum_{k = 1}^{K} PS %_{i, k, j} = \frac{TC}{M} * \sum_{k = 1}^{K} \frac{A_{k}}{A} * \frac{Q_{i, k, j}}{Q_{k, j}} [€ / product line / day]$ (9) The reader can notice the following:

Only RFID item-level tagging enables the punctual calculation of Q_i,k,j, which is needed to compute the cost to display. In fact, RFID inventory counts (either through fixed or handheld readers) make it possible to assess inventory levels for every product line in every store area on every day.

For any store area k, the higher the quantity Q_i,k,j displayed and the longer that product line is displayed, the higher its cost to display. Therefore, its sales should be higher to cover those costs.

In conclusion, it is worth stressing once again that RFID technology enables both net profit and the calculation of the cost to display.

As far as net profit is concerned, RFID makes it possible to correlate sales to displays, that is, to attribute every sale to a product display in a specific store area. This process, also known as ‘money mapping’ of store areas, highlights ‘hot zones’ where displayed products are sold rapidly and in quantity, compared to ‘cold’ ones, which are characterised by slow-moving product lines. With bar code technologies, this information cannot be obtained since punctual, real-time display data for store areas are not available.

RFID also enables real-time cost calculations to be displayed. It also provides real-time inventory count data either through fixed readers (ceiling readers or smart shelves) or mobile handhelds to calculate CD_i,k,j for every single PL _i in area k on day j.

Next, we show how to leverage all of this to optimise store visual merchandising.

First, let EBIT [%] be the company’s expected margin as measured by the ratio between EBIT and net profits. Since every store is expected to have a profitability at least equal to the company’s EBIT, the same ratio should apply to every product line PL_i displayed every day j in every store area k.

For every product line, thanks to RFID technology, we can compute CD_i,k,j and NP_i,k,j using equations (8) and (4), respectively. Then, we can position product line i on the bubble diagram shown in Fig. 1, where the product line net profit is on the x-axis and the related cost to display is on the y-axis. The bubble area is proportional to sales.

Fig.1

We can then highlight three main product line categories:

Money makers (1): these product lines are represented with number 1 in the bubble diagram in Fig. 1. These are the product lines whose NP> CD*(1+EBIT): money makers generate high net profits, far higher than the related cost to display. Roughly speaking, these are usually the best-selling models, characterised by consistent sales growing trends. Therefore, these are the models/colours that the store manager has to display, not only to generate sales but also often to attract customers into the store. At the same time, these models need to be carefully monitored to check for early signs of evolution toward the indifference line.

Store manager inputs for money makers include the following:

Maintain display levels in the store area

Either try to reduce the displayed inventory or move a product line to a less visible store area and see if consistent sales can be maintained even with lower inventories or in less ‘precious’ areas. This will free high-value space for slower product lines, which could benefit from being placed close to money makers or being moved to an area with higher visibility. If sales should drop, return to the starting point.

Indifferent (2 and 3): The line NP = CD*(1+% EBIT) represents the indifference area, where the net profit generated by a product line equals its cost to display. Roughly speaking, all product lines characterised by a median ratio between sales and displayed quantities fall around the indifference line.

The indifferent product lines are represented in Fig. 1 with the numbers 2 and 3, and the input for store managers may vary depending on profits.

If profit is high (2), store managers may try to move the products to the money-maker zone by reducing the quantity displayed and seeing whether sales volumes are maintained. Alternatively, items should be moved to less precious areas to see if the same sales levels can also be achieved in a lower cost to display area. If product lines evolve to the left-hand side zone, the solution can be maintained; otherwise products should be taken back. These are in fact high sales product lines, which have a high impact on store turnover, and they are eventually worth displaying.

When product lines are characterised by low net profits (3), the store manager should explore how long these product lines have been displayed in the store area. Low profits may be due to the fact that these product lines have only been recently displayed, and therefore it may be worth waiting to make a decision. Conversely, long-displayed product lines should be removed and either returned upstream or transhipped to other stores or simply held in the backroom and displayed again later in the season.

Money losers (4): Money losers are denoted by the number 4 in Fig. 1. They are product lines lying below the indifference curve. This means that their cost to display is higher than the related profits. These product lines may be characterised by either low sales or low margins, and/or they may be displayed in high quantities in high-value sales areas. An indicator for money losers is usually a low ratio of sales compared to the quantity displayed.

Store manager input for these models/colours are as follows:

In any case, reduce the quantities displayed in the store area and/or move items to low-value sales areas; see if the same sales levels can be sustained with lower inventories in the same store area or with the same inventories in less precious store zones. Otherwise, the products should be replaced.

Are these models characterised by missing central sizes? It is worthless to display models when central sizes are missing, as even though customers may like the model, they will be unable to their sizes. Are these models out of season? For example, heavy spring models may be difficult to sell if the weather turns hot, or vice versa; light fall models do not sell well in cold wintertime. If so, the products should be replaced.

3.2 Sales vs. space

The cost to display vs. net profit calculation may not be straightforward for every product line, as the main Capex and Opex figures must first be determined. Moreover, according to (4) sell in and sell out, punctual values should be available. Therefore, integration between RFID and other legacy systems may be required.

However, as highlighted earlier, since NP depends on the quantities sold (see S_i,j in Equation 4) while cost to display depends on inventories in the store area (see Q_i,k,j in Equation 8), retail managers can leverage RFID data to develop an easy to assess KPI, which is useful for the analysis described above.

According to the simplified procedure, the following steps can be followed:

Given a time interval j, typically a day

Given a store area k (a floor, a corner, a window a rack)

Using RFID, calculate Q_i,k,j [item/day], that is, the quantities displayed on day j in area k for every product line i

Either through RFID or through BC scans at the point of sale, calculate S_i,j [item/day], that is, the quantities sold on day j for every product line i

Then, for every product line PL_i calculate:

% SP_i, that is, the percentage of space (with reference to overall quantities displayed that day in that area)

% SA_i, that is, the percentage of sales (with reference to overall store sales that day)

Position every product line in the simplified matrix % SP – % SA

The four quadrants can be defined according to the respective percentiles. While the 50th percentile can be adopted on the % SP axis, the percentile for % SA can vary according to the area k considered. The higher the value of the sales area, the higher the percentile for the product line to become a money maker (i.e. product lines displayed in a window or in an entrance area can be considered money makers only if they are in the 75th percentile or higher).

Once PLs are set in the Fig. 2 matrix, the same conditions from the previous paragraph apply to the money makers (upper left quadrant), indifferent (lower left and upper right ones) and money losers (lower right quadrant).

Fig.2 4 Evidence from the field

The theoretical model described in the previous paragraph has been field tested to validate it and assess its practical applicability and implications on store performances.

We deployed an RFID test pilot in a retail chain of a leading luxury fashion and apparel maison. The company has more than 1,000 stores worldwide, selling more than 15 million items per year. For our tests, we selected three flagship stores representative of the whole chain, two in Rome and one Milan, in the heart of the respective fashion districts.

The RFID supply chain is rather straightforward. Garments supplied from the main logistics hub to RFID stores are RFID tagged at the point of shipping. A passive UHF pre-programmed hang-tag is added to each garment. The pre-serialised RFID tag is associated to the SKU at the DC via a double BC scan (SKU and pre-encoded serial number).

In the stores, we deployed the following RFID processes:

receiving: carried out with both mobile readers and fixed readers, making it possible to automatically cross check received goods with the bill of lading and manage errors.

replenishment from the backroom: in two out of three RFID stores, RFID fixed gates track the replenishment from the backroom to the store area and vice versa and update the inventories of the related business locations accordingly.

inventory counts: store associates carry out this process with handhelds daily. Inventory counts can be taken for the store floor as a whole as well as in specific areas, namely, floors, rooms, corners and/or windows. Once the area to be inventoried is selected, the RFID middleware automatically sets the handheld RFID power accordingly. As a rule of thumb, the bigger the area, the higher the RFID power.

checkout: carried out with fixed readers at the checkout counters.

other RFID processes: includes other processes such as returns, B2C sales and transhipments from/to other stores that may impact product inventories both in the backroom and in the store area are taken into account and properly RFID reengineered.

The RFID data collected from the field are stored in EPCIS-compliant data warehouses (DWs). Local DWs synchronise with a cloud-based central infrastructure in real-time, serving as the engine for a business intelligence dashboard. The dashboard imports raw RFID data from all stores and makes it possible to calculate punctual inventories in different store areas, and thus to calculate % sales and % space for every product line, according to the model presented in the previous paragraph. In agreement with retail and store managers, we decided to set the analysis at the model/colour level (that is, product lines PLs)

Testing campaigns have been carried out for four seasons, starting with spring–summer 2016, and are currently on-going. To date, about half a million tags have been tracked through the RFID supply chain described above.

As a result of the RFID technology, it is possible for a company to give an objective, quantitative answer to previously unanswered questions, which had always puzzled retail managers, leaving room for personal interpretations and positions:

Does a flagship fashion retail store, with a strongly assisted sales service such as those selected for the RFID deployment, sell what customers can see and touch on the shop floor, or are store associates able to induce sales of product lines directly from the backroom?

Is it true that the higher the number of sizes displayed, the higher the sales volume—or are sales insensitive to the number of sizes displayed?

To this end, we carried out focused tests both in Milan and Rome stores, correlating product line sales with RFID inventory data on the store floor. The results indicated that, regardless of the highly assisted sales through dedicated store associates, the vast majority of customers purchase the models/colours they are able to see and touch in the store area. The likelihood of selling a product line that is not displayed, that is, directly from the backroom, is very small. This is true not only for high-space retail stores and outlets (this point is highly agreed upon by company retail managers) but also for the luxury fashion flagships, such as the one tested during the pilot. Evidence from the field shows that up to 86% of sales are triggered by the models/colours that are displayed in the store area, while only 14% involve models/colours that are not displayed. Therefore, the appropriate displaying of products in the store area is paramount to sustain consistent sales.

We addressed the second question through dedicated tests in the Milan store. For example, during one test, store managers displayed the same product lines in different corners on alternate days, placing either three central sizes (40, 42, 44) or a full mix of all sizes (from 36 to 48). We correlated daily sales with the display arrangement adopted. The three central size displays usually gave similar or sometimes even better performances in terms of sales than the full mix. Remarkable indeed was a test where the same jackets were displayed in one day, following the three central size arrangement but on a slow selling floor, and on the next day in a full mix on a high selling floor. The test went on for three weeks, and unexpectedly the 3x display yielded higher sales.

As far as the application of the model described in the previous paragraph is concerned, before RFID deployment, store area displays were managed as follows (the same process applies for the non-RFID test store):

Displays on floors, corners, racks and windows are chosen by a visual merchandiser, who briefs store managers about the job to be done. The store managers are then able to make slight adjustments, based on his or her local and punctual sales perception.

Non-RFID stores do not have visibility on:

Whether the visual merchandiser’s instructions have been followed;

What adjustments have been made and what, where and how much merchandise is currently displayed in the store area;

Sales maps in the store areas;

Product line behaviour (money makers, indifferent, money losers).

In RFID stores, visual merchandising follows a new “to be” procedure:

Every day, or even twice a day, store associates take RFID inventory counts through RFID handheld devices of floors, corners, racks and/or windows. Counting using RFID handhelds takes a matter of minutes and has an accuracy of almost 100%;

The retail manager updates the NP CD matrix weekly, every Monday, as well as the % SP – % SA matrix for every store area;

High-selling and slow-selling areas are identified;

For every area, PLs are placed in the diagram shown in Fig. 2 (money makers, indifferent and money losers);

Store managers make changes to improve the sales of indifferent and money losers;

The agreed actions are taken;

During the week:

Sales are monitored to check the real effectiveness of the actions proposed in terms of sales increases;

Money makers are also monitored to verify that they maintain their remarkable sales performances;

If the results are consistent, the new display is kept; otherwise new solutions are proposed.

Tests have been carried out on the three RFID stores, sometimes at the category level and sometimes at model level, down to single product lines.

We measured the results in terms of incremental sales against projected store sales, the latter taking into account external factors, such as the number of people visiting the stores (this data was available in the three stores because the appropriate systems were installed), adjusted according to other external factors (i.e. the weather, national holidays, etc.).

4.1 Quantitative outcomes

Table 1 shows some of the most significant results gathered during the RFID visual merchandising testing campaigns. Activities took place in spring–fall 2016 in Rome and Milan stores. We measured incremental sales achieved thanks to the RFID visual merchandising process.

Table 1
Table1

Test Store Area Focus Store sales

1 Rome Entrance PLs +3,2%

2 Rome Ground floor Knitwear +2,3% ()

3 Milan First floor Knitwear +3,8% ()

4 Non-RFID stores All Knitwear +0,5% (*)

5 Rome Ground floor PLs +1,7%

6 Milan Windows First floor Area, some PLs +11,3%

7 Milan Ground floor Knitwear +3,8%

8 Milan Basement Jackets +0,4%

Test	Store	Area	Focus	Store sales
1	Rome	Entrance	PLs	+3,2%
2	Rome	Ground floor	Knitwear	+2,3% (*)
3	Milan	First floor	Knitwear	+3,8% (*)
4	Non-RFID stores	All	Knitwear	+0,5% (*)
5	Rome	Ground floor	PLs	+1,7%
6	Milan	Windows First floor	Area, some PLs	+11,3%
7	Milan	Ground floor	Knitwear	+3,8%
8	Milan	Basement	Jackets	+0,4%

(*) for knitwear only.

In the following paragraph, we detail the tests and comment on the results.

Test #1 was carried out in Rome, were RFID data highlighted the store entrance as a critically slow-moving area (0.31 sales/day per item displayed, on average), despite the high number of customers passing through. The store managers tried to figure out how improve this performance. They decided to rearrange the merchandise, giving more visibility to heavier models and moving lighter ones behind. In the next two weeks the entrance sales velocity increased from 0.31 to 0.50 sales/day/item displayed. Store turnover improved as well, scoring +3.2% against the corrected forecasts.

Tests #2 and #3 were carried out in Milan and Rome and were focused on knitwear. In both stores RFID and sales data put knitwear into the money loser category (0.06 sales per day/item displayed in Rome and 0.12 sales per day/item displayed in Milan). The merchandisers investigated and found that both the Milan and Rome displayed knitwear in slow-moving areas. Therefore, they suggested replacing some indifferent product lines, which have long been displayed in high-selling areas, with the same knitwear, the latter therefore moving from a slow to a hot area. A lying visual merchandising concept was adopted in one store while hanging visual merchandising was adopted in the other. The results were remarkable, especially for the hanging displays. Knitwear moved towards money makers (0.13 sales per day/item displayed in Rome and 0.21 in Milan). The knitwear turnover also increased in the two stores: by 2.3% and 3.8%, respectively.

Test #4: RFID stores passed the information obtained from the RFID visual merchandising tests about knitwear (which models to display and how to display them, that is, hanging vs. lying) to six other stores, which were not equipped with RFID. In the following weeks, knitwear sales rose slightly in the non-RFID stores, but as a slower rate (+0.5%). The reason was further investigated, and it seemed likely that not all of the non-RFID stores had adopted the suggested policies.

Test #5 took place in Rome. In this test, which was based on RFID inventory data combined with sales data, the store managers moved some of the money makers displayed on the ground floor (high-sales area) to a slow-moving sales area far from the entrance. In this way, they were trying to force customers to walk through the whole store to reach the popular items, thereby increasing sales in the slow-moving areas. In the subsequent weeks, the sales of the money makers remained strong, even in the slow area, while the overall store sales benefited as well.

Test #6: In week 31, the order lines displayed in the Milan store windows emerged as money losers. The store managers decided to remake the windows, replacing previous product lines with fast-selling ones. At the same time, new window models were also displayed on floor 1, forcing customers to walk through the whole store and up to floor 1 to see and touch them (see test #5 in Rome). The results were remarkable, with an increase of 11.2% compared to projected sales, the most significant improvement of the whole campaign.

Test #7: RFID visual merchandising analysis in Milan identified knitwear as a potential money loser. The RFID data showed that those PLs were scattered on all floors. The store managers decided to group them and create a focused corner on the ground floor to increase sales. The category improved and moved towards being a money maker. The overall store performance benefitted as well, with an increase of 3.8% against projected turnover.

Test #8 focused on jackets in the Milan store. The RFID data showed slow sales but also very few items displayed in the basement area. The store manager decided to try to increase sales simply by increasing the number of product lines displayed. Projected sales increased 0.4% in the following two weeks.

One additional test can be mentioned, although it was not focused on increasing sales. Some money loser product lines that had long been displayed on a slow-moving floor were moved to a high-selling one to appraise the effects of this display shift. After a week, we calculated the new positions of the product lines in the NP CD matrix. We observed that more than 50% of the PLs migrated toward the money-maker quadrant, while all the others remained either in the indifference or in the money-losers quadrants. The insight taken from this test was the need to replace these money loser PLs. Moreover, information was passed on to the style department about the main features of these money losers so that they would not be replicated in future collections.

In conclusion, the model developed and detailed in the previous paragraph has shown to be robust and applicable in a real environment. Information gathered combining RFID inventory data (% space) with turnover (% sales) information provides valuable insights to store managers for optimising visual merchandising and balancing categories and assortments in the store area. In addition, during the testing campaigns we achieved significant results in terms of sales increases (+3.1% on average). These figures demonstrate that the RFID visual merchandising use case supports full RFID rollout in fashion retail.

5 Conclusions

Visual merchandising in the store area plays a pivotal role in affecting consumer buying decisions. Through visual merchandising, retailers attempt for present an attractive sales environment that impacts customers emotionally and keeps them in the store as long as possible while triggering the buying process through well-designed product displays.

However, even in high luxury, fashion retail stores, customers buy what they see in the store. Evidence from the field proved that the vast majority of sales are triggered by items whose PLs (understood as a unique combination of model/colour) are displayed in the store area.

Fashion stores currently do not track their display decisions, that is, what/how many PLs have been exposed not only on shelves and racks but also in windows, corners and floors; therefore, this piece of information reflects a significant gap in knowledge.

RFID item-level tagging can easily close this technology gap and provide the punctual visibility retailers need to get consistent information about PL displays, correlate this information to sales and optimise their visual merchandising strategies.

We introduced an original, quantitative model to analyse the performances of the store and its store areas as well as of the PLs displayed daily. RFID data make it possible to categorise PLs according to the original framework proposed, that is, on the one hand according to the net profit generated by sales, and on the other hand, based on the cost to display every PL in a specific store area.

Only PLs whose net profits overcome the costs to display by an EBIT percentage are worth displaying, and the higher the value of the store area where they are exhibited, the higher the net profit (that is sales) they should produce or the lower the cost to display (that is inventory) they should require.

Based upon these premises, in our approach we divide PLs into three categories: the money makers, that is, those PLs whose profit far exceeds their cost to display; the money losers, that is, the PLs whose cost to display exceeds the net profit; and the indifferent, referring to those PLs whose net profits equal the cost to display. For every category, we propose tailored actions that store managers should take to optimise store performance (maintain, replace, reduce or move).

In addition, we propose a simplified model, based on RFID data only, where PLs are categorised based on their percentages of sales generated and space taken. The same grouping—money makers, losers and indifferent—then applies, as well as the same optimisation strategies.

Our model has been tested in a real RFID deployment, comprised of three flagship retail stores of an Italian fashion maison. Two stores in Rome and one Milan, in the heart of the respective fashion districts, were the case subjects, and they can be considered as representative for the whole chain. The testing campaigns have been carried out for four seasons, starting with spring–summer 2016, and they are currently on-going. To date, more than half a million tags have been tracked through the RFID supply chain.

The RFID data demonstrated the importance of displaying at least one model/variant in the store area in order to induce sales, even in stores featuring a deeply assisted sales model, such as flagships. Meanwhile, the number of sizes displayed appeared to play a secondary role in sales.

As far as the RFID visual merchandising approach is concerned, the results were measured by comparing the actual sales in the categories affected by RFID-driven actions with store projections without RFID. Moreover, we compared RFID store sales with similar non-RFID test stores. The results showed that the proposed model is consistent and robust. Additionally, a significant turnover improvement in RFID stores has been consistently achieved thanks to the RFID-enabled visual merchandising model. They are nearly the same order of magnitude of optimised RFID replenishment from the backroom.

Although the three RFID stores are representative of the whole chain and are among the most significant ones in terms of importance, location and characteristics, the small number of stores tested is a significant limitation to this study. In the near future, this scope will likely be increased in an attempt to determine the implications of the RFID-enabled visual merchandising model for either smaller retail stores or larger retail outlets.

References

Al-Kassab

, Blome

, Wolfram

, Thiesse

, & Fleisch

(2011). RFID in the Apparel Retail Industry: A Case Study from Galeria KaufhofGaleria Kaufhof. In Unique radio innovation for the 21st century (pp. 281–308). Springer Berlin Heidelberg.

Bertolini

, Bottani

, Ferretti

, Rizzi

, & Volpi

(2012). Experimental evaluation of business impacts of RFID in apparel and retail supply chain. International Journal of RF Technologies, 3(4), 257–282.

Bottani

, Ferretti

, Montanari

, & Rizzi

(2009). The impact of RFID technology on logistics processes of the fashion industry supply chain. International Journal of RF Technologies: Research and Applications, 1(4), 225–252.

Brynjolfsson

, Hu

Y.J.

, & Rahman

M.S.

(2013). Competing in the age of omnichannel retailing. MIT Sloan Management Review, 54(4), 23.

Christopher

, Lowson

, & Peck

(2004). Creating agile supply chains in the fashion industry. International Journal of Retail & Distribution Management, 32(8), 367–376.

, Hardgrave

B.C.

, & Sharda

(2007). RFID for better supply-chain management through enhanced information visibility. Production and Operations Management, 16(5), 613–624.

Ebster

(2011). Store design and visual merchandising: Creating store space that encourages buying. Business Expert Press.

Fisher

M.L.

(2003). What is the right supply chain for your product. Operations management: Critical perspectives on business and management, 4, 73.

Footfall, 2016, full report available at http://uk.shoppertrak.com/wp-content/uploads/2016/03/Fashion-Conversion-Benchmarking-Report-2015.pdf

10.

Garrido Azevedo

, & Carvalho

(2012). Contribution of RFID technology to better management of fashion supply chains. International Journal of Retail & Distribution Management, 40(2), 128–156.

11.

Goyal

, Hardgrave

B.C.

, Aloysius

J.A.

, & DeHoratius

(2016). The effectiveness of RFID in backroom and sales floor inventory management. The International Journal of Logistics Management, 27(3), 795–815.

12.

Gudonavičienė

, Alijošienė

(2015). Visual merchandising impact on impulse buying behaviour. Procedia-Social and Behavioral Sciences, 213, 635–640.

13.

Hardgrave

B.C.

, Miles

R.B.

, & Mitchell

(2009). Item-Level RFID for Apparel: The Bloomingdale’s RFID Initiative. Information Technology Research Institute, University of Arkansas, Feb., retrieved, 12, 2009.

14.

Hardgrave

B.C.

(2009). Item level RFID for apparel: The Dillard’s RFID initiative. Information Technology Research Institute.

15.

Hardgrave

B.C.

, Waller

, & Miller

(2005). Does RFID reduce out of stocks? A preliminary analysis. Information technology research institute.

16.

Hardgrave

B.C.

(2012). Omnichannel Retailing You can’t do it without RFID, RFID Journal, available online at, http://www.rfidjournal.com/articles/view?10247

17.

Jung Chang

, Yan

R.N.

, & Eckman

(2014). Moderating effects of situational characteristics on impulse buying. International Journal of Retail & Distribution Management, 42(4), 298–314.

18.

Kerfoot

, Davies

, & Ward

(2003). Visual merchandising and the creation of discernible retail brands. International Journal of Retail & Distribution Management, 31(3), 143–152.

19.

Law

, Wong

, & Yip

(2012). How does visual merchandising affect consumer affective response? An intimate apparel experience. European Journal of marketing, 46(1/2), 112–133.

20.

Madhani

P.M.

(2011), RFID deployment: Fast fashion retailing.

21.

Mehrabian

, & Russell

J.A.

(1974). An approach to environmental psychology. the MIT Press.

22.

Mehta

, & Chugan

P.K.

(2013). The impact of visual merchandising on impulse buying behavior of consumer: A case from Central Mall of Ahmedabad India.

23.

Moon

K.L.

, & Ngai

E.W.T.

(2008). The adoption of RFID in fashion retailing: A business value-added framework. Industrial Management & Data Systems, 108(5), 596–612.

24.

Morgan

(2008). Visual merchandising: Window and in-store displays for retail. Laurence king publishing ltd.

25.

Randall

W.S.

, Gibson

B.J.

, Clifford Defee

, & Williams

B.D.

(2011). Retail supply chain management: Key priorities and practices. The International Journal of Logistics Management, 22(3), 390–402.

26.

Rizzi

, Romagnoli

, & Thiesse

(2016). A new framework for RFID use cases in fashion and apparel retailing. International Journal of RF Technologies, 7(2-3), 105–129.