Retailing organizations need to equip themselves with the infrastructure to develop and manage customer databases. POS data has been a reasonable success story in the, western markets. This paper provides examples of simple analysis using POS data that can directly aid practising managers in their strategic and tactical decision-making.
Retailing organizations need to equip themselves with the infrastructure to develop and manage customer databases. POS data has been a reasonable success story in the, western markets. This paper provides examples of simple analysis using POS data that can directly aid practising managers in their strategic and tactical decision-making.
Retailing organizations need to equip themselves with the infrastructure to develop and manage customer databases. POS data has been a reasonable success story in the, western markets. This paper provides examples of simple analysis using POS data that can directly aid practising managers in their strategic and tactical decision-making.
Notes present short pieces which are research-based, experience-based, or idea-based.
Effective Retail Promotion Management: Use of
Point of Sales Information Resources Arindam Banerjee and Bibek Banerjee India has witnessed a surge in organized retail- ing in the recent past. While the retailing industry is still in its infancy stage and growth opportunities are significant, the lessons from more developed markets suggest that increased competition in this area will demand better operational efficiencies to remain viable in the long run. To be able to support good decision- making for the future, retailing organizations need to equip themselves, among other things, with the infrastructure to develop and manage The increasing availability of customers' customer databases which can be effectively transaction level data at the point of sale mined in the future to help drive strategy- (POS) in electronic form in various stores building. Investing in Point of Sales (POS) data in India is opening up important arenas in has been a reasonable success story in the, marketing analytics that can enhance busi- western markets and this paper attempts to ness dicision-making process. This paper highlight some of the plausible applications of discusses the strategic role that manage- such data. It also emphasizes the areas in which ment of customer information resources can retailing majors could possibly invest to reap play for the growth and sustenance of the benefits of market information in the future. business. This paper also provides examples of simple analysis using POS data that can The use of POS data in developing analyti- directly aid practising managers in their cal models that provide insights into managerial strategic and tactical decision-making. decision-making in the context of retailing has been well-documented in marketing science Arindam Banerjee and Bibek Banerjee are Associate literature. A recent article on this issue (Bucklin Professors in the Marketing Area of the Indian Institute and Gupta, 1999) articulates the progress made of Management, Ahmedabad. in the consumer packaged goods industry in the US in the past 20 years in harnessing POS data both in academic research as well as in devel- oping effective industry practices in the area of retail promotion management. Some of the decision areas under retail promotions that have been extensively researched are: • Product pricing at retail outlets which in cludes decisions on timing and depth of discounts. • In-store display planning which includes issues
Vol. 25, No. 4, October-December 2000 51 Vikalpa
on what brands/product categories to display POS data availability has been on the increase, and location of such displays. especially with most retailing giants having electronic check out in their retail outlets which • In-store communication and advertising facilitate easy recording and storage of data. management — issues on design of ceiling signs, mobiles, inflatable, floor graphics, danglers, instant redeemable coupons, etc. Why is POS Data an Important Source of Managerial Information? • Local area advertising decisions to promote products that build store traffic. POS provides a rich source of behavioural data • Decisions regarding shelf space allocation on the customers' purchase characteristics. It is and arrangement of brands to command the most disaggregate form of behavioural data optimal visibility. available from the retail outlet. In comparison, • Decisions regarding store coupon circulation the monthly retail audit data collected by ORG- for various product categories. MARG aggregates brand sales over a period of a month. Therefore, the details of the purchases made on every transaction are not available at What is POS Data? this higher order of aggregation. POS data has been used to develop various analytical frame- As the name suggests, POS data is information works that have provided important insights into collected at the retail store that provides vol- consumer retail purchase behaviour. In mar- umetric information on every transaction made, keting science literature, a seminal paper on i.e., the quantity sold, both in numbers and modelling brand choice (Guadagni and Little, value, the price at which the transaction was 1983) with marketing mix variables using con- made, and any added information regarding sumer panel data marked the initiation of a spate promotional programmes which the store ran of research on consumer behaviour using panel at the time the transaction took place. POS as well as store level retail transaction data. provides marketers with direct behavioural information on what consumers buy. In the developed countries, especially in the POS data has been recorded in a systematic US, research companies such as AC Nielsen and IRI have established businesses collecting both manner in India for over a decade now, albeit store level (POS) and consumer panel level at a modest scale. ORG-MARG has an' exten- information. The information collected is used sive operation in India that conducts a monthly primarily for tracking studies of market share, retail audit on a sample of stores for a limited but over the past 12 years, consumer data has number of brands/Stock Keeping Units (SKUs). been systematically analysed to generate value The information recorded has been primarily added insights on consumer reaction to various volumetric data, which has been used by Fast marketing mix initiatives. The SCAN*PRO Moving Consumer Goods (FMCG) companies model (Wittink et al., 1988) and its variants have to track market share over time. While third been extensively used by practising managers, party data collection practices at the retail store especially in the packaged goods industry in the have been around for a while in India, the US to develop an understanding of the causal fragmented nature of retail industry has inhib- effects of price and retail promotion. Some of ited the collection of individual transaction level the areas which have been extensively re- data at the retail store. With the coming of searched, both in academic institutions as well organized retailing in India in the past three as in industry, are discussed below. to four years and its projected spiralling growth (Rs 160,000 crore by 20051 ), transactional level Analysis of Key Drivers of Sales 1.Economic Times, March 18, 2000. ET Interactive, Interview with Simon Bell, Principal, A T Kearney, New Delhi. Price and in-store promotion variables have
Vol. 25, No. 4, October-December 2000 52 Vikalpa
traditionally captured the attention of both retail • Optimal trade allowance package to be offered store managers and brand managers. What to retailers by manufacturers with the objec- brands to promote and when, for how long, tive of maximizing manufacturer's category which will drive profitability of the store or the contribution. brand, has been of prime importance to man- While the first two areas are primarily retail agers in this industry in the US for long. To management issues and are of prime importance evaluate the relative importance of various to store managers, brand and category manage- promotional vehicles, like temporary price ment specialists in manufacturing organizations reductions, coupon offers, in-store special dis- are necessarily concerned about optimizing their plays, freebies, and local area feature advertising own brand portfolio contributions. Sometimes, of price-offs, the SCAN*PRO model (Wittink these objectives may work against each other, et al., 1988) has been extensively used. Man- for example, if promotion of brand A in the agers across various companies ranging from store brings in high volume share for the Kraft, Coca Cola, Pepsi and Procter & Gamble, particular brand, which is good for the manu- to name a few, have used this modelling facturer of brand A; however, promoting brand technique to examine the effect of various trade A cannibalizes sale of other brands in the promotion initiatives. The end objective has category in the retail store to such an extent been to develop normative models that help that it causes a fall in the total contribution for make decisions regarding optimal promotional the product category in the retail store. spending — how much to spend, on what brand to spend, and what specific promotions to run. With retailing dominated by large organized Promotional price elasticity and base price corporations in the US who have the leverage elasticity are estimated using the SCAN*PRO to negotiate better terms with manufacturers to model output to make pricing decisions, wherein maintain brand visibility in the store, manufac- issues regarding the long-term impact of price turers have been forced to devise promotion changes are weighed against short run promo- mechanisms that harmonize retailer's category tional price effects. profit objectives along with their own. A very effective managerial tool developed In the context of using POS data for by market research organizations specializing in developing analytical frameworks for resolving the retail industry using POS data has been the category management issues, brand and cate- Sales Rate Chart. This is simply a distribution gory managers in the US have constructed of sales volume generated at a retail outlet across causal models which determine not only the various price points. The chart is simple but direct effect of own price and promotional mix provides important information on sales spikes elements on sales of the brand, but also the cross due to price changes. Sales rate charts have their effects of marketing mix elements of other limitations in terms of analytical rigour; how- brands which are deemed to be direct competi- ever, they have proven to be a successful tool tors. Own price/promotional elasticities and in making broad pricing decisions both in the cross price/promotional elasticities elements are long and short run. inputs to build category profit simulators which are typically scenario building tools which help Analysis of Category Management Issues in designing effective promotional programmes Three specific decision areas that have benefited which maximize contribution. by the use of POS information analysis are: Such exercises are carried out quite regu- larly at large packaged goods companies. In the • Optimal product portfolio size at the retail beer category, extensive use of this simulator outlet and retail shelf space optimization. has been made by the second largest brewery • Optimal retail promotion initiatives to max in the US market. The specific purpose of this imize retail category contribution. project has been to convince retailers across the
Vol. 25, No. 4, October-December 2000 53 Vikalpa
US markets that promoting their brand of beer other agencies (presumably not having access on key holiday weeks maximizes profits for the to store level data) have argued for the appro- retailers compared to promoting the market priateness of using market level data to estimate leader brand. the impact of what they call market level Rationalizing on portfolio and shelf space phenomenon such as advertising. Readers should management requires additional inputs in terms note that market level sales data is obtained by of layout design of retail stores (planograms) aggregating POS data across all stores in a which when integrated with POS sales informa- market over an appropriate time window. Using tion can provide significant insights into optimal appropriate projection factors that account for portfolio size as well as the area and shelf the population of stores the store level data can location to be allotted to different product be projected to get an estimate of the overall categories with the objective to maximize the market offtake. retail sales or profits. Exercises of this nature While brand managers are concerned about are performed regularly with fair degree of the effectiveness of their advertising budget, success in developing appropriate marketing retail managers also need to know about the decisions. Several research companies have impact of advertising on brands, especially developed syndicated analytical models which advertising campaigns of new brands and new have had limited success in resolving issues product categories. across the board. They go by nomenclature such, The Retailing Environment in India as Portfolio Manager or Category Manager; however, based on our experience, developing With her population touching a billion, India customized solutions using POS data has had is working at the doorsteps of becoming one significant potential in resolving product assort- of the world's foremost consumer markets. ment and shelf space management issues. About a quarter of this huge mass of consumers live in towns and cities and the remaining in Advertising villages. Over the years, the retailing infrastruc- POS data along with media exposure data as ture that has proliferated in India is character- collected by research agencies in the US, such ized by a high degree of fragmentation as as Nielsen Media Research (ORG-MARG col- compared to many developed nations. A recent lects media exposure data in India) have been estimate puts the figure at 10 million operational used by brand managers to calibrate their long- retail outlets in India, 32 per cent of these being term advertising spendings. There have been in urban areas. Small stores (about 300- 400 attempts at studying the short-term effects of sq.ft.) accounted for 64 per cent of the retail advertising on sales; however, the impacts outlets in the country whereas very large stores estimated have been fairly low (0.1 to 0.12 for (about 800 sq.ft.) constituted only 3 per cent established brands, 0.2 to 0.4 for new brands). of the establishments. FMCG stores accounted There has been moderate success at estimating for nearly 75 per cent of these retail outlets. advertising impact on sales in the long run (1.5 According to a CMIE2 forecast, total retail sales to 2 times more than short run effects). in India is likely to exceed Rs 9200 billion by An explanation for these lukewarm results the year 2002, of which about 73 per cent is may rest with the characteristics of data re- expected in the food sector. Therefore, it is sources used to construct the analysis. Research obvious that the great Indian retailing revolution agencies such as AC Nielsen and IRI favour is not waiting for the size of the business the use of store level data (POS) to study opportunity. The challenge lies in identifying the advertising effects, which they claim is consist- key drivers that steer the Indian consumers' ent with the analysis conducted to study the perception and behaviour when it comes to their effect of store promotions (Bucklin and Gupta, shopping needs. 1999). This point is, however, debatable since 2 Centre for Monitoring Indian Economy.
Vol. 25, No. 4, October-December 2000 54 Vikalpa
In this context, market enthusiasts are crys- Herein lies the opportunity of utilizing the tal gazing on the 'fate' of large format and/or POS data to model consumer response to organized retailing in India. There are interest- marketing mix variables at the retail level. With ing trends by way of statistics, e.g. the friendly increasing competition, both at the manufac- neighborhood mom&pop stores have increased turer level and the retail level, it is obvious that per 1000 population as per an ORG study, and there are distinct advantages of being the first large format and specialty retailing is also on mover in harnessing information resources to the increase (though their absolute numbers are drive the marketing strategy building exercise, still quite small). This is at the expense of both for retailers and manufacturers. The latter, perhaps the middle-sized-middle-value shops3 . it is hypothesized, will increasingly feel the The reality is that every retailer has to "under- pressure of building strategies that are in con- stand his customers" more discerningly than gruence with the retailers' business objectives. ever before and make strategic choices to pursue the right target (customer) with the right propo- The rapid changes in the retailing environ- sition. Also, the reality is that every retailer ment are currently quite evident at least in some today is unanimous in his appreciation that he of the metropolitan pockets of the country. A needs to "deliver value" to his customers. fairly extensive review of the expanding retail- ing sector in India is available in prominent The final reality is that in today's retailing trade publications.4 Competition has set in from environment in India, the sheer complexity of multiple sources. In the Chennai market, com- the product-market matrix is posing 'mental- petition in food retailing is multi-faceted with model' based decision-making a real challenge. the neighborhood (kirana) stores facing direct This exponential change, started during the competition from more than one organized early 90s, is arising essentially out of two retailing chain: FoodWorld, Nilgiri's, Subhiksha, sources. First, in pursuit of the proverbial 200 and Vitan. With such intense competition in the million strong Indian middle class, the manu- market, newer retailing chains are forced to facturers have been continually adding new adopt significantly differentiated POS strategy products in the market place. For instance, in to make customers change their shopping habits. the FMCG sector, there were 57 core categories For instance, the specific trade publication of products in 1990, which grew to 76 by 1996 reports that Subhiksha adopts a blanket policy (and the trend continues). These 19 new cat- of discounting prices by selling less than MRP, egories boasted of 1378 brands and 2579 SKUs. similar to the textbook definition of an Everyday Furthermore, the number of SKUs in the erst- Low Pricing (EDLP) strategy. "You have to give while 57 categories also grew from 7715 to 15160 customers a solid reason to change their shop- during the same period (Banerjee et al., 1999). ping behaviour. One which conventional stores like the kirana can't duplicate," says the director The second change agent, and the most of Subhiksha chain of supermarkets. important, is the consumer who has become vastly discerning through rapid exposure to the The above instance corroborates our asser- global business environment. The Indian con- tion that retailers of the future will have to sumer today reflects a strong preference for rapidly employ differentiated marketing strategy imperatives such as evaluating choices from based on customer response information. Uti- among large assortments of products, a plea- lizing customer behaviour data at POS to deve- surable shopping experience, and a shopping lop effective pricing strategies, EDLP or other- experience that would provide the maximum wise, will certainly be an imperative for long- "value" per rupee spent. term survival and growth.
3 The Hindu Business Line, "Responding to Modern 4 Brand Equity, The Economic Times, April 26 - May 2, Retailing Formats," May 23, 1998. 2000.
Vol. 25, No. 4, October-December 2000 55 Vikalpa
Status of POS Data in Retail Table 1: Type of POS Data Available Management in India * Item Files • Bill Date, Bill Number, Bill Time, Product As was pointed out earlier, ORG-MARG has Code, Cost, MRP, Sold Value, Quantity operated a retail store audit in India for many * Product Master years. The data primarily consists of actual stock • Product Code, Product Name, MRP, Group movement data at sample stores. Collection of Code corresponding marketing mix elements has been * Group Master minimal, and only very recently has some effort • Group Code, Group Name been made to collect price and promotion data. * Campaign Files AC Nielsen is also reported to be planning to • Product Code, Campaign Start Date, Campaign develop an audit infrastructure. However, none End Date, Cost, MRP, Selling Price, Campaign of these organizations is capable of fully col- Number lating the rich transaction level data that exists Source: Raghuram et al, 1999. at the retail point of sales. AC Nielsen and IRI claim to provide census data on store sales, but the investment in infrastructure that organized only in a few geographic regions in the US. retailing chains have made in collecting trans- action level data. An ad hoc survey in the A typical source of POS transactional data Ahmedabad market, which is in the infancy will have information on the following: stage of the retailing revolution compared to its • Wider range of SKUs. counterparts in the south, revealed that a sig- nificant number of-supermarket stores collected • Tracking of generic/store brand sales. transaction data at the cash register. Yet, no • Ability to handle consumer panels through significant efforts are made currently to "mine" loyalty programmes. this database to reveal insights on customer behaviour. Even with limited information avail- Ability to conduct product basket analysis. ability, these databases have enough data to • Ability to generate information on optimal understand market level drivers of store sales. "loss leader" promotion strategy. The reasons for this perceived apathy may be varied; lack of expertise in mining data may A recent study document prepared on this be an important reason but more often it is the topic (Raghuram et al., 1999), highlights some issue of prioritization of initiatives that has kept of the analytical capabilities of this type of developmental projects such as database mining information. However, a detailed assessment of a back burner. This is most unfortunate since a typical POS transactional level database we predict that with rapid entry of competition, available today shows the limited amount of players who understand customer psyche, information collected currently (Table 1). This through its manifestation of overt behaviour, is not surprising given that use of the data has will dictate the future course of the retailing been primarily to generate simple accounting industry. reports. In order to develop the full potential of POS data for decision-making, the future managers have to proactively think of informa- An Example of a Potential Analysis tion bytes that need to be systematically col- We used a database obtained from a store of lected which can be used along with the a large retail food chain based in South India transaction level data in the future to help to identify specific "category-driven" transac- making better store level decisions. tions that generate more revenue for the store. One area that conventional grocery stores The purpose of the study was to identify like the small kirana stores cannot replicate is customer segments that, on an average, gener-
Vol. 25, No. 4, October-December 2000 56 Vikalpa
ated more revenue for the store compared to Table 2: Average Transaction Value (in Rs) others. Customer segments were identified based Category Key Purchase1 Minor Purchase on their main purchase category as exhibited Staple 517 160 by the nature of their transaction. Spices 445 130 Transaction data over a four-month time Cooking Oils 620 153 period across all recorded product categories in Pulses 556 140 the store was used for the analysis. Five major category purchases were identified for the study Soap 475 152 1 based on the frequency of purchases made When the key purchase is of the corresponding category. (weighted by the value of the purchase). These Similarly, the average transaction, which was categories were: a) staple, b) spices, c) cooking driven by spice purchase, was valued at Rs 445 oils, d) pulses, and e) toilet soap. A sixth compared to Rs 130 that was spent by custom- category was formed clubbing sales from all the ers5 who did not have spices as a major category other categories. The number of categories in their shopping list. chosen for the study was kept at six to avoid unnecessarily complicating the study, but the The largest difference in the transaction analysis could be done at any level of detail. value occurs between transactions that record It must be pointed out that while this type of significant purchase of cooking oils versus the data has been available in India only in the past transactions that do not record significant three to four years, there does not seem to be purchases of cooking oil (Rs 620 versus Rs 153). a scarcity of transaction data in electronic form This result throws some light on the type in the organized retailing sector currently. of customers that bring in more value to the In our analysis each individual transaction store in terms of revenue (Chen et al., 1999). in the store (identified as all purchases made If store managers were to make decisions about during one store visit based on a unique bill promotions in the local area to draw more number) was categorized as either a particular customers, the typical output obtained from this category driven purchase or not (e.g. staple analysis may be used to identify the "right" driven purchase or not staple driven purchase) destination category to be featured in the local based on the rupee value spent on the category area promotion to increase store traffic of high in the specific transaction. To control for vari- valued customers. In the present case, high ation in average value spent across categories, propensity of using cooking oil as a promotion the purchases were categorized based on their category will attract the highest revenue cus- relative value compared to the maximum value tomer. recorded in that category across the entire Designing Proper Information Resource to database. This is a limitation, since without any Help Better Decision-Making in the Indian customer-related characteristics in the database, Context it is impossible to identify the drivers of indi- vidual purchases and hence one has to resort There are limitations in the analysis described to approximations. above given the approximations made in cate- gorizing transactions into specific category driven SAS software was used to manipulate the purchases due to the lack of customer descrip- data and the results obtained are presented in tors. Also, the results obtained are not statis- Table 2. The direct interpretation of the results tically robust enough to pass a test of technical is that customers that came into the store rigour. Although technical rigour is needed, primarily to buy staples spent on an average Rs 517 compared to customers who did not have "The same transaction may be classified as driven by staples as a major category in their shopping multiple category purchases — staples, spices, etc. This is list (average spending per transaction is Rs 160). true since many consumers make monthly purchases in bulk across major consumption categories.
Vol. 25, No. 4, October-December 2000 57 Vikalpa
often, appropriate data analysis (even without true value of this loyalty programme is to passing the litmus test) provides a higher level provide individualized promotions to customers understanding of consumer behaviour which can which enhances perceived service quality at the shape better decision-making. store. What is critical to drive the process is There are a number of non-food departmen- availability of the right information/data re- tal store chains in Mumbai and Delhi, which source that can be mined. In this respect, have introduced loyalty programmes and are FoodWorld (C) (Banerjee et al, 1999) touches tracking purchase data on these customers. on the need to "IT-enable" operations for Proper design and mining of these databases can smooth running of the retail operations. We yield significant insights to develop direct mail would suggest that retail organizations of the promotion campaigns. future do need to go beyond envisioning IT There is tremendous opportunity to enhance integration. They will benefit by adopting a the scope of in-store promotions and to capture proactive stance in designing and developing an the information at the most disaggregate level. IT-enabled information warehouse for both In-store displays, whether they are end of aisle customer and operational information, which displays, within aisle or front of store, have been can significantly drive strategic decision-making recorded to have varying effects on sale spikes. in the future. As a starting point, an assessment A deeper understanding of the drivers of in- of the type of decisions that one needs to take store sales will help retail store managers plan in more advanced stages of the business life their space utilization more effectively in con- cycle is necessary. Lessons from the developed junction with the most appropriate marketing markets of the west and the type of decision programme. It is important to invest in an initial problems that they have faced may provide period of testing various innovative marketing leads to the type of information requirements options and recording them religiously, such as of the future. one would do in conducting a pilot test pro- gramme. After the initial phase of testing, Some Broad Level Initiatives managers would have a sufficiently large infor- mation base to fine tune their programmes Initiating customer loyalty programmes is a according to the market needs. However, this pragmatic way of collating customer level trans- latter stage does make the created database actional data that can be linked to customer redundant. Effective ongoing decision-making characteristics. Segment level analysis of cus- demands continuous building and mining da- tomer likes and dislikes drive most marketing tabases as well. decision-making and this type of database provides ample opportunity to customize store Conclusions promotional programmes based on the type of consumer segments who frequent the store. An Retailing is entering an active phase in its effective way of utilizing this behavioural infor- business life cycle in India. This note has mation is to generate customized coupons and attempted to address a dimension of this busi- promotional features at the check-out counters ness, which has a strategic role to play in the based on the customer characteristics and cur- growth and development of its constituents — rent purchases made. Catalina Corporation, a the management of information resource. Like Chicago-based research agency has evolved many other service industries which have be- such coupon generating machines that have come intensely focused on information resource been set up at several retail food stores across management, for example, consumer finance, the US. Manufacturers have also availed of this insurance and hospitality, survival in the long service at retail counters to induce brand run will depend upon smart management of switching with limited success. However, the market information resources. Unfortunately, Vol. 25, No. 4, October-December 2000 58 Vikalpa developing marketing information resources Chen, Yuxin; Hess, James D; Wilcox, Ronald T and needs a significant lead-time and there is a Zhang, John Z (1999). "Accounting Profits vs natural advantage of being the first off the block. Marketing Profits: A Relevant Metric for Category Large retail chains have a definite advantage Management," Marketing Science, Vol 18, No 3, pp 208-229. with respect to the available infrastructure that they have in place. They simply have to initiate Raghuram, G; Banerjee, Bibek; Jain, A K; Koshy, a proactive process of investing in appropriate Abraham and Bhatt, Gunjan (1999). Retailscope '99, information that will yield the right kind of Retail Sales Data: The Hidden Treasure, Draft marketing insight for the future. Report, Ahmedabad: Indian Institute of Manage- ment. References Guadagni, Peter M and Little, John D C (1983). "A Banerjee, Bibek; Raghuram, G and Koshy, Abraham Logit Model of Brand Choice Calibrated on (1999). "FoodWorld (C): The Road Ahead," Scanner Data," Marketing Science, Vol 2 (Summer), Working Paper, Ahmedabad: Indian Institute of pp 203-238. Management. Wittinik, Dick R; Addona, Michael; Hawkes, William Bucklin, Randolph and Gupta, Sunil (1999). "Com- and Porter, John (1988). SCAN*PRO: The Estima- mercial Use of UPC Scanner Data: Industry and tion, Validation and Use of Promotional Effects Based Academic Perspectives," Marketing Science, Vol 18, on Scanner Data, Working Paper, AC Nielsen, No 3, pp 247-273. Schaumburg, IL.
Frederick Cooper (Editor), Randall M. Packard (Editor) - International Development and The Social Sciences - Essays On The History and Politics of Knowledge-University of California Press (1998)