Franchising in the Convenience Store Industry - Research Paper Example

Topic: A Empirical literature Review of Organizational Design and Control across Multiple Markets: The Case of Franchising in the Convenience StoreIndustry Introduction Companies get into expansion mode by having a number of business units for many reasons: to leverage from available competencies, to cover risks better way, to promote company’s brand name, and get economies of scale etc. Mostly companies serving multiple markets are chain organizations. Benefits of serving multiple markets kept aside; there are significant challenges of control across different market types that come in the way of attending to and serving diverse customer base. Taking the example of a bank, it diversifies risks by offering credit and savings services to different groups and communities. Similarly, a hotel chain can achieve economies of scale by opening at new locations, as a departmental store repeats its business model by operating many stores in markets that serve different customer range. Organizational Design and Control across Multiple Markets Crucial control problems surface such companies because of the vast network creating information asymmetries among the units answerable to customers and the headquarters. Such situations demand expertise of the local management more than the headquarters comparatively. Companies try to resolve the issues faced in controlling multiple markets through their organizational design and control choices. There are market type dispersions due to changing customer demand; it is not related to geographical dispersion necessarily. It is a possibility that there are strongly different types of customers in any two units of a company although they are not far from each other geographically. For example convenience stores in the ethnic population in Chicago attend to customers of totally different ethnicities although distance from one store to the other is just some blocks. The same cannot be true always, as it also happens that units are situated far away from each other geographically but customer base is the same. Some of the operational decisions can be delegated to unit managers like maintaining stock levels and customer-relationships but crucial operational and strategic decisions need to be taken by the headquarters like spending on advertisement, purchase and choosing the service or product mix. It becomes challenging to take these decisions by the headquarters when market-type dispersion is high. It creates two control problems, first, due to difference in local conditions, it becomes difficult to keep a check on managers and control them, as serving different markets. Secondly, wide customer base demands won’t be easy to satisfy from a demand perspective. One way of dealing with information asymmetries between the headquarters and a business unit manager with local market knowledge could be transferring of authority and incentives to local managers. In chain organizations, such problems can be managed internally by awarding higher responsibility to managers and distributing higher compensation risk to the local unit or externally by franchising some units. Franchising in the Convenience Store Industry Franchising provides the answer to control problems by at once decentralizing of decision making and offering of incentives. It has been substantiated even by franchise managers that franchising mechanisms have leverage over internally controlling the chain units in market-type dispersion. Franchising mechanisms attract franchisees with management and market knowledge that is in any case better than the skill of company-appointed store managers. Risk taking entrepreneurs are offered franchise. Risks are transferred from the company to the franchise that can manage them more competitively than headquarters. According to the Chief Operating Officer of a big convenience store chain in the Northeast U.S.: “Franchisees are more entrepreneurial and would have more control over difficult situations such as adapting merchandising to local market conditions, and managing theft, shrinkage, and other aspects required to maintain performance.” Review of Empirical Literature Campbell, Datar & Sandino (2008) have examined why chains use organization design solutions to reduce control problems coming from market-type dispersion by conducting analyses on convenience store chains at the chain and the store level. Analysis was made by using 2004 TDLinx data from 420 convenience store chains to find the effect of market-type dispersion on the chains’ decisions to franchise some of its stores or all stores, and even none. By using the ordinal logit model, it is attempted to know whether there exists a positive relation between market-type dispersion and decision to franchise. This analysis is better than other traditional methods of monitoring issues created by geographical dispersion. Campbell et al. (2008) analyzed 43 chains that have their own stores and franchises as well as 34,892 stores managed by those chains. It was found that the association between the decision to franchise and store’s market-type dispersion was sound to the store’s demographic features. The findings show that there is a positive relation between franchising individual stores and high market-type dispersion. Chain level survey data was also used from the National Association of Convenience Stores (NACS) to know how non-franchise chains control market-type dispersion. For this purpose, data was used from 53 non-franchisor chains to prove that such chains facing higher market-type dispersion were more prone to decentralize their functions than the chains with lower market-type dispersion by not using the corporate and inspection rank workforce in comparison to the store-level employees. Store managers are paid higher variable pay, which was found by dividing store managers’ bonuses with total reward. The tests were found to be more consistent than other determinants of decentralization and variable pay, the chain’s geographical dispersion and size. Measurement of Market-Type Dispersion and Divergence Campbell et al. (2008) measured market-type variation at chain level and at store level. At chain level, the extent of market-type dispersion is the total of local features among stores within a chain, for example, a chain with all of its stores in Santa Monica, CA should have less market-type dispersion in comparison to a chain with stores in Downtown LA, Santa Monica, and Sherman Oaks. The extent of market-type divergence at store level depends on the store’s location demographic features that are not similar to the chain’s average demographic features. For example a chain having one store in Harvard Square and the remaining stores in Roxbury, the store in Harvard Square should have higher market-type divergence in comparison to the stores in Roxbury. Measurement of Chain Level Market-Type Dispersion Market-type dispersion was defined as the variation of certain location features in all stores within a chain. The related location features impact negatively the stores’ demand conditions. Before developing a measuring parameter for market-type dispersion, the most applicable location features to the demand of a convenience store were found before discussing the hypothesis for measuring dispersion (Campbell et al. 2008). The development of market-type dispersion is specifically crucial, as this aspect has not been analyzed in earlier literature. So, market-type dispersion in four different ways has been developed to compare the regression results of different measures. As per the marketing literature, the drivers that impact the consumer buying trends the most in grocery and retail stores are population density, average household size, per capita income, age, and ethnicity (Campbell et al. 2008). Examples of findings from these five dimensions besides previous research include: ________________________________________________________________________(Population (i) stores situated in densely populated areas have higher sales realized; Density) (ii) customers of not densely populated areas go farther from their local market areas, causing loss of sales realization; ________________________________________________________________________ (Household (iii) Shoppers from larger families shop generic products and bigger brand Size) sizes, they have the tendency to use off-price coupons, as availability of income in hand to spend in grocery stores is sufficient with them; _______________________________________________________________________ (Per Capita (iv) young, highly educated, and prosperous customers have greater income & age) opportunity costs, and therefore, is less concerned on price and promotion; (v) rich customers prefer the displays and advertisement relevant with their desires to save them costs and time on finding; ________________________________________________________________________ (Ethnicity) (vi) African-Americans spend comparatively more money on generic grocery purchases than Whites; (vii) the percentage of African-American and Hispanic consumers in an area is related to greater price sensitivity. ________________________________________________________________________ All of these location features have been taken from the ESRI dataset at a zip-code level. The U.S. Census’ definition of per capita income and average household size has been used. Population density has been calculated as the number of residents per square mile. For ethnicity the percentage of white population has been measured, while using the median age of the population to measure age. Campbell et al. (2008) have developed dispersion measures besides the five location features stated above. One set of measures is based on the variation of individual location features, while another set of measures makes use of cluster analysis. Market-Type Dispersion Based on Variation of Location Features Dispersion can be expressed (Campbell et al. 2008) by getting the standard deviation of each location feature L over all stores in a chain I but there is a negative input of expressing standard deviation, which is that it is expressed in the same units as the location feature L (e.g., the standard deviation of the per capita income is expressed in dollars), and that’s the cause of it not being matched to other location features. To overcome this hindrance, dispersion of each location feature L is expressed by normalizing the standard deviation by the average value of the Location feature over all stores within the chain i: The normalized dispersion is without units so that to be compared beside various location features. This way we get an idea of the amount of differences among stores in a chain for a specific location feature, balancing it with the typical value of that location feature within the chain. For developing the chain’s total market-type dispersion, the normalized dispersion expression is aggregated of all chains by addition or multiplication, as shown below: The total dispersion expression would assign equal weights to the five location features and take them exclusively. The multiplied aggregated dispersion is not same to added aggregated dispersion because of its interaction effects. Therefore, MDispersion_Mult is more susceptible to very high or low dispersion values on any single location feature. For example, this term would place more (less) weight on the average household expression when income dispersion or some other location feature is also high (low) (Campbell et al. 2008). Market-Type Dispersion Based on Cluster Analysis Taking the lead from the clustering approach of many of the segmentation studies in marketing research (Punj and Stewart 1983; Chatuverdi et al. 1997)), Campbell et al. (2008) has divided the markets through cluster analysis and then viewed the number of clusters or market types, attended by all chain stores. A total of 29,827 zip-codes were clustered taken from the ESRI database into sets with matching location features like population density, average household size, per capita earning, ethnicity, and age. Clusters were formed by following the three steps (Punj and Stewart 1983; Sharma 1996)): 1. All location variables are ranked over the complete sample of zip-codes by mean centering and then dividing it by the location feature’s standard deviation. Thus all location variables are expressed in terms of the number of standard deviations they depart from the mean. 2. By using a hierarchical centroid clustering technique13, a priori cluster solution is measured, to (a) find out a candidate number of clusters, and (b) zero-in a starting division of the clusters that would not be impacted by outliers. 3. By making use of a priori cluster solution incurred from step 2 as an initial point, a non-hierarchical (K-means) technique14 is employed to get a final cluster solution. Following this method, 25 sets of zip-codes with an R-square15 of 81.1 percent is achieved, and R-squares that rank between 77 and 85 percent for the five location features, indicating that the cluster solution has a fair level of homogeneity within sets and the clusters are well divided over the location variables. Campbell et al. (2008) by using the 25 zip-code clusters have construct two extra market-type dispersion standards expressed as the number of clusters where the stores in chain i operate (MDispersion_NClusti), and a dispersion standard based on a herfindahl index shown as: The MDispersion_NClusti values differ between 1 and 25, while the MDispersion_HHIi exhibit has a minimum value of zero, if all stores in a chain are situated in a single type of market, and a maximum value of 96 if a chain has precisely 1/25th of its stores in all the 25 clusters. The market dispersion efforts based on clusters are different from the dispersion efforts reached by the range of each of the location features in various ways. First, MDispersion_NClust and MDispersion_HHI are less responsive to outliers in location features. For example, having a store location with average per capita income quite higher than the rest of the chain would crucially increase the standard depart of per capita income within the chain, but would have no impression on the Nclust or Herfindahl computations of market-type dispersion. Second, MDispersion_NClust and MDispersion_HH1 suffer information loss on how stores are divided within each cluster. Finally, MDispersion_NClust and MDispersion_HH1do not differentiate among which clusters the stores come. For example, if there are two chains A and B, where chain A has its stores equally distributed in cluster 2, while chain B has half of its stores in Cluster 1 and half of its stores in Cluster 3, then both chains would have the same MDispersion_NClust and MDispersion_HH1 values. But, if the difference between the location features of Clusters 1 and 2 is more than the difference in location features in Clusters 1 and 3, then according to MDispersion_Sum and MDispersion_Mult, chain A would be more dispersed than chain B (Campbell et al. 2008). Measurement of Store Level Market-Type Divergence The market-type divergence at the store level is the limit to which the store’s location features vary from the average location features in the chain. To tell apart these measures from those at the chain level, they are called as “market divergence” measures. In the matter of the chain level dispersion measures, Campbell et al. (2008) have used population density, average household size, per capita income, ethnicity, and age, as the related location features. Two sets of market divergence measures are developed: one is based on the divergence between the store’s individual location features and the chain’s average features, while the other is explained in terms of the clusters outlined for each chain. Market-Type Divergence Based on Individual Location Features Campbell et al. (2008) have calculated the absolute value of the difference between the value of a location feature L for store j, and the average value of location feature L over all stores in chain i to account market divergence at the store level. As there are different units for different location features like per capita income divergence is measured in dollars but average household size is measured in number of persons. To make the measures of different location features compatible, the divergence measure is normalized by the standard deviation of the L values for all stores in chain i. Thus, Campbell et al. (2008) have defined for each location feature L: This measure signifies how many standard deviations store j’s location feature L is away from the chain i’s average. This normalized divergence metrics is aggregated by either multiplying or adding the Normalized Divergence measure over the five different location features: Like the chain level dispersion measures, the aggregated divergence metric developed via addition would handle the five market feature dimensions as not subservient to one another, while the aggregated divergence measure developed via multiplication would include interaction effects for the various location features. Market-Type Divergence Based on Cluster Analysis By using the 25 market clusters created at the chain level, Campbell et al. (2008) have created a measure of store level divergence that appropriates the divergence between the store’s cluster and the most random cluster in the chain, weighted by the randomness of the main cluster. The measure is calculated as given: where, %mainclust = is the percentage of stores in the most random cluster in chain i distance = is the distance between the centroids of the store js cluster and the chains most frequent cluster. Summary & Conclusion The two methods for computing dispersion at the chain level and divergence at the store level are: (1) measures based on the variation of individual location features that take to the MDispersion_Sum and MDispersion_Mult measures at the chain level, and to the MDivergence_Sum and MDivergence_Mult measures at the store level; and (2) measures based on 25 different clusters (types of markets), which include MDispersion_NClust and MDispersion_HH1 at the chain level, and MDivergence_mainclust at the store level. It has been substantiated with empirical literature that chains passing through greater levels of market-type dispersion are more prone to transferring authority and increase incentives by opting for the organizational design choice of franchising, to tackle control problems. A chain can franchise some or all its outlets related to market-type dispersion, gauged through many possible methods, and that this result is dependable to other factors discussed in literature like the chain’s geographic dispersion. The decision to franchise at a store level for the subset of chains that have some stores and franchise others has been examined empirically. It is found that stores whose location features are more expanded from the most common location features of the chain overall, have the greater possibility of being franchised. Bibliography Campbell, Dennis., Datar, Srikant M., Tatiana, Sandino. (2008). Organizational design and control across multiple markets: the case of franchising in the convenience store industry. Retrieved from http://www.hbs.edu/research/pdf/08-091.pdf Hendrikse, G.W.J. & Jiang, T. (2007, August). An incomplete contracting model of governance structure variety in franchising. ERIM. Rotterdam School of Management (RSM). Retrieved from http://publishing.eur.nl/ir/repub/asset/10462/ERS-2007-049-ORG.pdf Kranz, Sebastian & Lewin-Solomons. (2008). Decision structures in franchise systems of the plural form.” Retrieved from ftp://web.bgse.uni-bonn.de/pub/RePEc/bon/bonedp/bgse8_2008.pdf Read More