I n today’s highly competitive retail industry, it is crucial to recognize the factors which help retailers to thrive in the long run. One of the major factors which distinguishes a successful retailer from others is information regarding the efficiency of its stores. Like Charles Darwin’s ‘survival of the fittest’ theory, the efficiency of a store enables the retailer to function effictively in the market, with minimum inputs and maximum output, to be the fittest amongst all the rest.

A retail store is a critical asset in the retail industry. Customers interact directly with the retail store and most of the revenue in the retail industry comes from stores. Hence, efficient management of store operations is essential, to achieve and sustain customer satisfaction, as well as cost-effectiveness.

Techniques to evaluate a store’s productivity and its relative efficiency involves evaluating many factors-such as inventory turnover, sales per selling space, employee productivity, customer traffic, effectivity (retail conversion rate), average sales, gross margin (sales profit before costs), among other factors. The current industry practice generally involves a simple, logical, efficiency-based approach which is not mature in terms of leveraging solid analytical techniques. Thus, generally, store efficiency is evaluated by considering only sales density, that is, sales per square foot or number of employees. This involves only two metrics at a time, ignoring other factors. These standalone factors contribute individually and independently towards the evaluation of a store’s performance. This leads to information throw in various random directions, while business leaders are attempting store performance evaluation and goal setting. Various store performance dashboards showcase different charts as these can’t be clubbed together to get one performance index for a store. Combining them into one, to get the overall performance in terms of an efficiency score, which can then be used to compare various stores, is possible through a non-parametric approach from data mining and analytics called Data Envelopment Analysis (DEA). DEA uses a linear programming procedure to evaluate store efficiency by considering parameters-such as staff information (employee count and payroll), operating expenses, fuel consumption and space as inputs and sales, EBITDA, gross margin as outputs for a store.

Advantage DEA

DEA is a very powerful technique to evaluate the relative efficiency of business units when multiple inputs and outputs are involved. This framework has been adapted from multi-input, multi-output production functions and applied in many industries. In general, DEA minimizes inputs and maximizes outputs: In other words, lower levels of the former and higher levels of the latter represent better performance or efficiency. Instead of central inclinations, DEA focuses on boundaries. It doesn’t try to fit a regression plane through the center of the data as in statistical regressions; rather, DEA proves particularly proficient at detecting relationships that would remain unseen with other evaluation approaches. DEA benchmarks units against their best peer groups. It can be applied to explore production data as well as cost data. Based on the business problem, one must be able to differentiate the various factors into inputs and outputs to apply DEA analysis efficiently. Some benefits of this technique are:

• It is assumption-free
  
• It’s one of the most scientific methods of calculating efficiency
• There’s no need to explicitly specify a mathematical form for the production function
• Multiple inputs and outputs can be considered for calculating efficiency
• Inputs and outputs can have different units
• It’s an useful technique to identifying future goals for business units
• It can conduct comparisons directly against peers, and reward efficient experimental units based on past performance

DEA impact: A retail case study

About 750 stores of a top US retailer had to be compared in terms of features such as area, sales, gross margin, payroll and incentives, operating expenses, EBITDA, revenue etc. To get the best and most significant factors influencing the efficiencies of the stores, correlations among the inputs and the outputs were checked. Among all the variables, number of employees working in the store and the area of the store, represented the best inputs, whereas sales and EBITDA represented the best outputs. DEA algorithm was leveraged to analyze data and get relative efficiencies across stores.

Key insights were derived from the results (given as findings 1, 2, 3, 4 and 5) which helped store managers and operation leaders undertake critical business action plans to improve store performance.

Findings 1: Insights based on relative efficiencies across stores

• All stores having efficiency levels of 80-100% are in rural areas
• In suburban, high-density and urban areas, the maximum number of stores are 40-49% efficient
• In almost all regions, many of the stores have efficiencies between 30-39%, which indicates great scope of improvement in the store’s performance

Findings 3: Sales and area vs. relative efficiencies across stores

• Sales are increasing with store area
• Stores with less area and lower sales show the highest efficiency because of better space utilization
• Average efficiency is high with lower store area but as the store area increases efficiency remains mostly constant

Findings 4: Sales and number of employees vs. relative efficiencies across stores

• Sales are increasing with the number of employees
• Stores with less number of employees as well as sales, have higher efficiency

Findings 5: US states where top ranked stores are located

• 25 stores across 17 states are 70-100% efficient
• Most efficient stores are concentrated in Central and Pacific regions

From data to insights

DEA solution framework is a powerful tool that enables identification of production frontiers and efficiency of firms in various sectors. It has been recognized as a valuable approach, useful in uncovering relationships that remain hidden with other methodologies. DEA provides relative efficiencies, that is, efficiencies relative to the data considered rather than absolute efficiencies. After applying the DEA algorithm, one can begin to tackle, with a clear degree of insight, issues such as identification of best and poor practices, target setting, resource allocation and monitoring efficiency changes over time.

Sandipan Bhattacharyya - Lead Consultant - Analytics, Wipro

Sandipan is part of Data Discovery Platform team and leads an initiative which involves creation of industry-specific applications across verticals. With over 12 years of advanced analytics experience, he has worked in multiple domains including Banking, Financial Services and Insurance, Manufacturing and Technology, Retail, Energy and Utilities. Sandipan has a Ph.D. in Statistics from the Indian Agricultural Statistics Research Institute, New Delhi.

Shivanjali Garg - Data Scientist - Analytics, Wipro

Shivanjali is part of the Data Discovery Platform team. An expert in multiple analytical tools and data mining methodologies, she has worked on various analytics projects in Manufacturing and Technology, Retail, Energy and Utilities domains.