Sorry, but copying text is forbidden on this website!
One of the challenges for companies that have invested heavily in customer data collection is how to extract important information from their vast customer databases and product feature databases, in order to gain competitive advantage. Market basket analysis (also known as association rule mining) is one of the data mining methods (Berry and Linoff, 2004) focusing on discovering purchasing patterns by extracting associations or co-occurrences from a store’s transactional data. Several aspects of market basket analysis have been studied in academic literature, such as using customer interest profile and interests on particular features of the product for the product development and one-to-one marketing (Weng and Liu, 2004), purchasing patterns in a multi-store environment (Chen et al., 2004), or point at certain weaknesses of market basket analysis techniques (e.g. Vindevogel, Van den Poel and Wets, 2005). Market basket analysis has been intensively used in many companies as a means to discover product associations and base a retailer’s promotion strategy on them.
When different additional brands are sold together with the basic brands, the revenue from the basic brands is not decreasing, but increasing. “Buy two, get three”sales promotion campaigns are very successful, if market basket analyses are used in order to determine the right products to be promoted. “Buy a product, get a gift” sales promotion campaigns are successful, if a basic product and a gift are related and the basic product has high margin rate. Based on market basket analyses, sets of products are defined and sold together with discount.
Limitedbrands organizes internal competition in up-selling.
Our paper – a case study – presents and analyses the application of market basket analysis in a major trade company in Slovenia.
2 The company Merkur, d. d.
Merkur, d. d. is a trading company (Merkur, 2005) that has for years ranked among the top companies in Slovenia dealing in items for home improvement, home services as well as lawn and garden. Merkur, d.d. has recently strengthened its position on the foreign markets through the supplies of goods to industrial enterprises, and by the establishment of its own retail network abroad. Merkur, d.d. is the mother company of Merkur Group. The Group consists of two Slovenian subsidiaries and six subsidiaries abroad (Zagreb, Sarajevo, Skopje, Munich, Milan and Warsaw). Besides that, the group also includes two offices (Moscow and Belgrade). Merkur plans to further strengthen its position on the domestic market, spread its sales to the foreign markets, especially to the markets of former Yugoslavia, and develop a high-quality range of products.
The company is organised in several large departments: Wholesale, Retail Sales, Sales to Foreign Markets, Purchasing, Logistics and Supporting Services. Customers include construction companies, trading organisations, installation companies, industrial enterprises, craftsmen and small entrepreneurs, as well as end consumers. The company makes almost 60% of its sales revenues by selling goods wholesale. To make the sales quick and efficient, the Wholesale Department has been divided into four sales sub-divisions. At present, Merkur has 38 retail sales centres in Slovenia. Specialisation increases the effectiveness of sales, so two types of Merkur sales centres were developed: MERKURDOM focusing on ordinary households, and MERKURMOJSTER intended for DIY (do-it-yourself) users. More information about MerkurDom and MerkurMojster is available on Merkur internet site: www.merkur.si.
2.1 Characteristic figures of the company
The scope of the company Merkur, d.d. can be shown through the following figures: The sales programme consists of about 200.000 active items (more than 120.000 items on stock), divided into 5 sales programmes, 74 lines of goods, 720 groups of goods and 5.600 basic goods classifications. Around 80% of sales are done with the top 12.000 items and 80% of stock is held on the top 20.000 items. The Purchasing Department issues more than 250.000 purchase orders with 1.200.000 items annually. Merkur purchases goods from more than 2.000 suppliers. About 80% of purchases are done with the top 200 suppliers. Wholesale has business relations with more than 2.500 buyers – organizations.
About 80% of wholesale sales are done with the top 800 buyers. Wholesale issues approximately 400.000 invoices with total 2.200.000 items annually. Retail sells goods to 13.000 buyers / organizations and to about 500.000 end consumers. More than 70% of sales to end consumers are personalized with the Merkur loyalty card called the “Merkur Card of Trust”. Retail issues 6.000.000 invoices with more than 20.000.000 items to end consumers annually. In the period from 1993 to 2004 Merkur achieved 19% average annual growth in revenues, 20% average annual growth in net margin and 27% average annual growth in profit from operations. Today Merkur is the sixth largest Slovenian company in revenues.
3.1 The history of DW&BI in Merkur
Merkur started to implement data warehousing and business intelligence (DW&BI) in 1999 with a project called KAS (Commercial Analytical System) (Svetina, 2002). Before 1999, different analyses and reports were performed in Merkur’s transactional information systems, much of the analytical data was held in Excel spreadsheets and Access databases. In the past, Merkur twice attempted to implement DW&BI technology, but failed because proposed technology was still too difficult to use for the majority of the users. In 1999 Merkur started with a major business process reorganization and, therefore, better and new business analyses were needed in order to make better decisions. The need for a DW&BI system emerged, so the KAS project was given high priority. Merkur started to design analytical data models for sales data and succeeded in integrating sales data from wholesale, retail and sales to foreign markets in one unified data model.
The IT department proposed Microstrategy DW&BI technology, which was installed and tested in the beginning of the year 2000. The technology was found to be appropriate and the decision was made to implement DW&BI with Microstrategy solutions. The first power users (sales analysts) were educated and the first KAS sales analyses were used in the decision-making process. In the beginning the ETL (extract – transform – load) process was carried out on monthly basis, but by autumn of 2000 the company started to perform ETL process daily. Later in the year 2000 the purchasing analytical system was introduced as well. In 2001, the data warehouse was upgraded with data on Merkur’s business plans. Sales and margins were planned on a very low organizational level. The annual plan fact table has more than 1.000.000 records, so the salespersons’ performance is measured very accurately. Because the technology is easy to use, the number of KAS users increased up to 100. In 2002, the implementation of a very large and complex analytical module followed, containing inventory data.
The inventory levels of each item in every warehouse on a monthly basis is stored in KAS and enables detailed inventory analyses and detection of critical items. Also, data on Merkur’s partner’s debts and liabilities was added to data warehouse, which enables accurate cash flow management. Item price calculation elements and different prices were imported in KAS in 2003, so critical prices can be detected and all inconsistencies eliminated. Many minor additions to the system were also made over the last few years. All the time Merkur tries to use adequate analytical and data mining methodologies in order to improve the whole system of business reporting. From the DW&BI history we can see a controlled step-by-step development of the KAS system.
Such way of development gives opportunity for good definition and implementation of analytical contents and enables Merkur to make many better business decisions. The KAS system brings Merkur an important competitive advantage, which enables the growth of the company. Improved decision making can be demonstrated through different measurable key success factors which are improving constantly. Key success factors such as net margin, net margin per item, net margin per customer, number of new customers and others are measured in KAS. These factors are always accessible for KAS users and help them to make better decisions.
3.2 DW&BI technology
Since 2000 Merkur has used the Microstrategy DW&BI technology. Microstrategy provides ROLAP solutions, which enable a step-by-step approach in data warehouse development and processing large amounts of data. The data warehouse is implemented in an Oracle relational database. This means that the same database technology is used in both transactional and analytical information systems. Therefore, Merkur’s IT department can focus in one database platform instead of two or even more. Oracle technology was used in Merkur before the implementation data warehouse was started, so the implementation of this technology was fast and smooth. In Merkur the following Microstrategy tools (Microstrategy, 2005) are used: MicroStrategy Intelligence Server is the heart of the BI system and provides reporting and analysis for the whole enterprise. This BI server provides the full range of BI applications through unified metadata and a single integrated server. MicroStrategy Administrator consists of a suite of tools that provide the systems management environment for business intelligence.
It maximizes uptime of BI applications. Its tools give an environment for developing, deploying, monitoring and maintaining of systems. MicroStrategy Architect is a rapid development tool that maps the physical structure of the database into a logical business model. These mappings are stored in a centralized metadata repository. MicroStrategy Desktop is the business intelligence software component that provides integrated query and reporting, powerful analytics and decision support workflow with a desktop PC. MicroStrategy Desktop provides an arsenal of features for on-line analysis of corporate data. Reports can be viewed in various presentation formats, polished into production reports, distributed to other users and extended through a host of ad hoc features including drilling, pivoting and data slicing. The interface itself is customizable to different users’ skill levels and security profiles. In Merkur, the Desktop solution is used by 13 power users (analysts).
MicroStrategy Web provides users a highly interactive environment and low maintenance interface for reporting and analysis. Using this intuitive HTML-only Web solution, users access, analyze and share corporate data through any web browser on any operating system. MicroStrategy Web provides ad hoc querying, quick deployment and rapid customizability, making it even easier for users to make informed business decisions. In Merkur, Microstrategy Web is used by 90 end users of KAS. MicroStrategy Narrowcast Server is a proactive information delivery server that distributes personalized business information to users via email, pagers and cell phones. It includes an intuitive self-subscription interface that enables users to specify what information they want to receive, as well as when and how they want to receive that information. Narrowcast Server is becoming more and more important in Merkur because of its efficiency.
3.3 Merkur’s DW&BI system today
Presently, KAS; Merkur’s DW&BI system, is five years old. The development of the system continues constantly and there is still much content throughout the organization which must be implemented in the BI system. The most important content to be implemented in the future are the following: Integral data from Merkur’s finance and accounting system (the finance and accounting analytical system) Relevant business data from Merkur’s subsidiaries Data from Merkur’s human resources analytical system Data from Merkur’s e-business analytical system Data from Merkur’s logistic analytical system
Presently in KAS (Merkur Commercial Analytical System – KAS, 2005): • 13 power users (analysts) and 90 end users; of both groups, 50 users have the ability and knowledge to set-up their own reports. • Up to 30.000 reports are run on KAS on monthly basis. • KAS consists of the following objects: o 137 tables o 433 attributes o 1.195 metrics o 5.611 reports • Over 35 automated services are run on the Narrowcast Server
The KAS system enables many sophisticated business analyses such as market basket analyses, described later in this paper.
4 Market basket analysis and the used methodology
Market basket analyses are an important component of analytical system in retail organizations. There are several definitions of market basket analysis. In a broader meaning, market basket analysis targets customer baskets in order to monitor buying patterns and improve customer satisfaction (Microstrategy, 2003). The following analytics can be used: attachment rates, demographic baskets, brand switching, customer loyalty, core items, items per basket, in-basket price, revenue contribution, shopper penetration and others. In a narrower meaning, market basket analysis gives us the answer to the following question: which goods are sold together within the same transaction or to the same customer? By analysing this information, we try to find out recurring patterns in order to offer related goods together and therefore increase the sales.
We can track related sales on different levels of goods classifications or on different customer segments. In this paper, the narrower meaning of market basket analysis will be taken into consideration, focusing on the use of these analyses in Merkur. It has to be noted that several other terms are also used to describe market basket analysis: related sales, cross-sell, up-sell. The distinction between these terms is very unclear and the same terms are often used in different meanings. What can we gain from market basket analysis (Limitedbrands, 2004)We get the ability to learn more about customer behaviour. We can make more informed decisions about product placement, pricing, promotion and profitability. We can find out which products perform similarly to each other.
We can determine which products should be placed near each other. We can find out which products should be cross-sold. We can find out if there are any successful products that have no significant related elements. 1. Discover the selling documents (transactions) with the item, for which we want to perform market basket analysis. This logic is valid, if we want to carry out item-related market basket analysis. We can also perform good classification or even loyalty card holder-related market basket analyses, which will be shown later in this paper.
2. Discover all the items in relevant selling documents and their selling quantities, prices, number of transactions and other relevant data. As an example, an item related market basket analysis will be presented. We want to analyse sales related to item ‘209525 Decorative lamp Saturn II’. In the first step we determine the selling documents with this item. The partial result is shown in the table 1. Further, the result of the first step is used as a filter in the second step, which results in a table with items, sold together with item 209525.
5 Areas of market basket analyses
In Merkur different kind of market basket analyses are done. Analyses are adapted to various business needs, and some of them are discussed in the following sections. In every section, the relevant examples of analyses are presented and opportunities for business action discussed.
5.1 Marketing and sales promotion campaigns
When sales campaigns are prepared, promoted items must be chosen very carefully. The main goal of a campaign is to entice customers to visit Merkur’s retail centre and buy more than they usually do. Therefore, we must choose the right items and offer the right prices or other conditions. Margins on promoted items are usually cut, therefore, additional non-promoted items with higher margins should be sold together with promoted items. As we could see from the example in Section 3, item ‘209525 Decorative lamp Saturn II’ is quite adequate to be included in a promotion. Together with it many other items are sold, so we can allow a lower margin of promoted item. Of course, there are some other criteria for an item to be included in a campaign, such as: • Where on the item life cycle curve is the item situated? • What is our brand promotion policy? • Can we reach an agreement with the supplier (producer) to assure larger quantities and better prices?
Table 4. Sales promotion market basket analysis In table 4, data from a New Year’s promotion campaign is shown. The: campaign was done through public advertising. Paper catalogues of promoted items were sent to households, there were also commercial spots on TV and radio, and advertisements in newspapers. Because of advertising a certain number of customers came in Merkur retail centres in order to buy the promoted items. Additionally, they also bought many non-promoted items (70% opposed to 30% of revenues and 75% opposed to 25% of margins) with much higher % of margin (29,08% opposed to 21,81%). This means that promoted items generated sales of non-promoted items. There are also many possible ways for organizing campaigns using direct marketing tools for the interaction with Merkur loyalty card holder. This issue will be discussed in Section 5.5.
5.2. System solutions offering
Market basket analyses are also used to combine more items in a set or a system, because the majority of customers are interested in buying and using them at a time or in a short period of time after the purchase of a particular item. By designing sets and systems of related items a company can increase sales and also cut down costs of sales transactions, so that various discounts can be offered to customers. This results in a typical win-win situation. A retailer must know the needs of customers and adapt to them. Market basket analysis is one possible way to find out which items can be put together in sets and systems.
Table 5. Classification Group ‘Kitchen extractor hood’ market basket analysis In Table 5 we can see groups of goods which were sold together with the group ‘Kitchen extractor hood’. In the related groups are also different kitchen appliances like refrigerators, dish washers, kitchen-ranges, taps, dishes etc. This means that Merkur should design and offer the customers different kitchen systems. These systems should include kitchen furniture, major and small kitchen appliances and kitchen utensils. Such a system should be displayed in one place in a retail centre where customers could choose from whole system solutions to just several parts (items) of these solutions.
5.3. Placement of goods in retail stores
Market basket analyses give retailer good information about related sales on group of goods basis. As we can see in Table 5, the majority of kitchen appliances groups are related. Customers who buy a kitchen appliance often also buy several other kitchen appliances. It makes sense that these groups are placed side by side in a retail centre so that customers can access them quickly. Such related groups of goods also must be located side-by-side in order to remind customers of related items and to lead them through the centre in a logical manner.
In Merkur, two basic concepts of retail centres are used: MerkurDom specialises in high-quality items for home improvement and garden, MerkurMojster specialises in high-quality products aimed at DIY users, craftsmen, and entrepreneurs. Centres are also classified by size as small and large centres. For each of these concepts, standardized placement plans were developed. Market basket analyses represent one segment of tools for decision making considering placement of goods. It can show us where we should change the placement of goods. After the change we can measure the business effects of the change.
5.4. Education of salespeople
The interesting results of market basket analyses must be presented to the salespeople in retail centres, because the employees must be aware of them and they should use them in the process of selling. Every salesperson has some knowledge about related items from his or her experience. With market basket analyses we can structure this knowledge and use it to teach less experienced personnel. Merkur invests a lot in education of salespeople through both internal and external sources. Knowledge from market basket analyses is widely used in internal education.
5.5. Segmentation of customers
As mentioned in Section 1.1., more than 70% of sales to end consumers are personalized with the Merkur loyalty card called “Merkur Card of Trust”. This data enables us to answer the following question: What did consumers who bought item (group) X in period 1, buy in period 2? If we identify customers who bought item X today, we can anticipate what they will buy, for instance, in next three months, and we can advertise them the right products. A typical example is shown in Table 6. We analysed loyalty card holders who bought ceramic tiles in the period from April to June 2004. In Table 6 we can see product groups which were bought by the same card holders in the period from July to November 2004. They bought different bathroom and kitchen accessories and central heating elements.
It would be very useful, if Merkur organized a targeted marketing campaign for this specific group of customers in July 2004 and promoted these products. There are many other possibilities and opportunities in Merkur to use loyalty card-based market basket analyses as a support tool for direct marketing campaigns. Merkur usually organizes non-targeted common campaigns, in which the majority of Slovenian households are included. But lately Merkur also started to implement direct marketing methods and therefore an effective data warehouse and business intelligence system is essential. This helps many interesting marketing ideas to be implemented.
The practice in Merkur proves that market basket analysis is a very useful for marketing campaigns, good placement definition and education of sales personnel. Merkur uses market basket analysis throughout the promotion campaign process. When a sales promotion is prepared, market basket analysis is used to define the right products and the right prices for the campaign. Related non-promoted items are also defined in order to place them in the vicinity of promoted items and therefore increase sales. When sales promotion finishes, its results are carefully analysed in order to discover opportunities for next promotions. Merkur widely uses market basket analyses to manage the placement of goods in retail centres. Related products and product groups are placed together in such a manner that customer can logically find items he/she might buy.
The findings of market basket analyses are an important part of the process of teaching the salespeople of Merkur. Sales personnel must be aware of related products in order to increase satisfaction of customers and intensify sales. Market basket analyses are just a part in the holistic approach to the execution of marketing development strategy in Retail in Merkur. The analytical process is integrated in other marketing activities and analysts are an important part of Merkur marketing development team. Team work is crucial for successful use of such analyses. Beside of the organization of the Merkur marketing process, a capable DW&BI system is needed.
The BI system must have good performances when processing large amount of data. It also has to be scalable and flexible, but, above all, the BI system must be user-friendly so that different marketing specialists can use it without any problems. Fortunately, Merkur’s KAS is such a system. But there is still much work to be done. We demonstrated that market basket analysis in Merkur can be done and that it brings useful results. In the future a working direct marketing strategy must be developed based on data already available in KAS. Then an organization and information systems for efficient execution of this strategy have to be established.
Berry, M.J.A., Linoff, G.S.: Data Mining Techniques: for Marketing, Sales and Customer Relationship Management (second edition), Hungry Minds Inc., 2004 Chen, Y.-L., Tang, K., Shen, R.-J., Hu, Y.-H.: “Market basket analysis in a multiple store environment”, Decision Support Systems (article in press), 2004, accessed through www.ScienceDirect.com Limitedbrands: Achieving Greater Efficiencies with Market Basket Analysis, Microstrategy World 2004 Conference, Miami, 2004 Microstrategy: Business Intelligence in the Retail Industry, Microstrategy World 2003 Conference, Las Vegas, 2003 Microstrategy Web Site: http://www.microstrategy.com/Software/, Microstrategy, 2005 Merkur Commercial Analytical System – KAS, internal document, Merkur, 2005 Merkur Web Site: http://www.merkur.si/ang/podj.html, Merkur, 2005 Svetina, Marko: Izdelava in uporaba market basket analiz, http://www.muson.net/Konferenca_login.asp?mni=12, Konferenca MUS 2004, Ljubljana, 2004 Svetina, Marko: Komercialni analitski sistem v podjetju Merkur d.d., Konferenca Poslovna inteligenca in upravljanje odnosov s strankami, Ljubljana, 2002 Vindevogel, B., Van den Poel, D., Wets, G.: “Why promotion strategies based on market basket analysis do not work” (article in press), Expert Systems with Applications, 2005, accessed through www.ScienceDirect.com Weng, S.-S., Liu, J.-L.: “Feature-based recommendations for one-to-one marketing”, Expert Systems with Applications,
Vol. 26, 2004, pp. 493-508.