The bigger a retailer is, the greater the impact of branch performance will be on its business success. In the UK alone, the Post Office has closed 2,500 of its 14,000 branches since 2007, RBS has planned the closure of 300 branches since 2008 and Jessops, a high-end photographic retailer, has closed 81 of 400 stores. Since all store closures affect the overall business model and the corporation‘s employees, it is critical that managers have the best data and tools available to them when deciding who to remove from the network.
The economic slow down has renewed retailers’ focus on their branch network. As less money is spent in the economy, any operational weaknesses are becoming exposed and investment decisions are subjected to closer scrutiny. Areas seen to give a poor return are examined by managers to ascertain whether it is better to risk additional investment in order to support a turnaround or whether it might be more efficient to just trim them from the network. However, if the decisions on where to focus improvements are unwise, any well intentioned turnaround efforts will fail to achieve improved performance while opportunities with more potential may be missed.
The majority of retailers are still using the same old metrics to benchmark performance, resulting in wasted resources and effort. The comparison of branch performance has traditionally centered on segmenting branches into like groups and comparing key performance metrics such as sales per square meter, return on assets, return on capital, profit and revenue. Though benchmarking has many well known pitfalls (see figure 1) it is still widely used and useful within its limitations, providing a two-dimensional picture of performance.
Using a novel approach to branch analysis, Data Envelopment Analysis (DEA), we can remove some of the traditional limitations of benchmarking and provide an insight into not just comparative statistics but levers of performance and the potential impact if changes are made. The extra dimension of information gives managers three key benefits:
What is Data Envelopment Analysis (DEA)?
DEA is a linear programming-based benchmarking technique used to explicitly consider the relative efficiency of a group of similar branches. This 100 % objective approach to branch analysis allows the identification of best practices and under-performing branches.
“Unlike many benchmarking approaches that rely on managers to observe, compare, and identify best-practice techniques; DEA helps the user to find those best practices which are too complex to be located through observation and traditional analysis. It enables management to objectively determine the optimum approaches in complex service operations. To be more precise, the bestpractice service provider is the one that uses the fewest resources to provide its volume and mix of service at or above the quality standard of the business. Service costs decline as the less productive service operations are improved to the best-practice level, guided by DEA.” 1
DEA can be applied to any branch network and is particularly useful when comparing servicebased branches, such as banks (see figure 2).
DEA benefits from being a purely data-driven and therefore objective tool. Organizations undertaking this assessment can easily obtain a quantitative picture of the branch concerned, provided the person who is in charge of the data – possibly a management accountant – and who understands how the branch is run, has the time to provide access to the data available in the organization. The following procedure should be observed when collecting and implementing the data.
Step 1 – Understanding the inputs and outputs that characterize all branches
The DEA requires a breakdown of all the inputs to a branch. Typically this would include factors such as labor hours, management hours, open times, rent levels, raw materials used, equipment spend over 12 months (including spending on refits) and any factors related to the specific type of business. Similarly, outputs such as revenue, customer service levels (off the shelf availability, mystery shopper ratings etc.), waste (expressed as a negative) should be gathered.
Step 2 – Segmenting the branches
In order to avoid the pitfall of comparing inappropriate branches, some attempt must be made to segment the branches into like groups. For example, branches based in busy office areas should not be compared with those in airports, branches in out of town malls should not be compared with those on a city centre high street.
Step 3 – Building and running the model
The DEA model uses linear-based programming within Excel to crunch the inputs and outputs together and assess the overall productivity of each branch. The tool uses the solver function within Excel to provide a sensitivity report that gives the result of the modeling – the background mathematics are quite sophisticated but the process is quite straightforward and the model can easily be adapted to suit all businesses.
Productivity assessment and branch comparison
1. Identifying the best performing branches
The first output from the data analysis is to identify which branches are the most productive – that is to say those that have the best output mix compared to the inputs. The most efficient are given a 100 % productivity rating. Further down the line, these branches will be used as a best practice pool to share with underperforming branches.
2. Identifying comparison groups
Those branches which don’t hit 100 % productivity are then identified along with a reference group of branches that have similar profiles. The groups of similar branches (known as Hypothetical Comparison Units [HCUs]) provide specific information on the areas other branches need to change in order to improve their productivity. For example, the data may show that branch A achieves its 100 % productivity by having twice as many transactions as branch D. Since the branches have similar profiles, this information can be used to understand which practices in branch A drive transaction levels and whether the same practices could be adopted in branch D.
There are two key things that should be noted at this point. Firstly, the productivity rating will usually be a combination of known information and surprising information. If a branch is known to perform significantly worse that others then this will almost certainly be reflected by low productivity. However, there is also potential for high performing branches to get a low productivity rating. Which should then lead to a reassessment of their potential to improve.
Secondly, the HCUs can reveal previously unknown relationships, for example that the branch in Windsor is similar to the branch in Gretna Green. As these are seperated by 300 miles it is unlikely that it would have occurred to anyone in the organization to compare the two branches and share best practice.
3. Identifying which branches to invest in
Once the productivity is understood, decisions can start to be made about where to invest turnaround effort. The Profitability vs. Productivity Matrix (see figure 4) shows a basic way in which actions should be taken.
Ideally, all branches would be in the green quartile of the chart, high profit with high productivity. Those on the “left hand side” of the chart, with low productivity, have the potential to improve and move across and up toward the green quartile, as an improvement in productivity should result in an increase in profits. Those branches in the red quartile, high productivity with low profits can be seen as having no where to go in terms of productivity and therefore to be at the maximum profit level.
Where productivity is high we know that the branch is performing as well as it can i.e. there is no value or potential return on investing in change activity – the decision is simple: if we are happy with the level of profit then we keep the branch open, if not, this is a branch we should remove.
Where the productivity is low, we know that the branch is performing below its potential and that investment in change is likely to result in an improvement in performance. The decision here requires more analysis on which branches are likely to give the highest return on investment.
At this stage, careful consideration needs to be given to the people affected within the branches. To be more specific, closures in productive branches that have low profitability could result in good people and ideas disappearing from the business while investment is put into poor performing managers in low productivity branches. Ideally, some attempt would be made to identify high performing staff members and efforts made to retain them.
DEA is a purely quantitative tool and part of its strength is the objectivity of the numbers, but having a snapshot of performance is only interesting if it can be turned into solid actions that deliver a significant benefit to the business.
Step 4 – Sense check the outputs
It is important to first critically analyze the outputs and to challenge those results that look out of place. There may be factors or inputs that were not properly considered when setting up the model which could affect the results – if this is found adjustments should be made and the model rerun.
Step 5 – Understand the data, build an improvement strategy
Once the data is fully understood, then an improvement strategy needs to be established. This typically takes two forms; first a network assessment identifying whether to close any branches and a priority assessment of those that will receive turnaround help. At this point a business case and target are defined along with a high level project plan.
Step 6 – Pilot an improvement intervention
Piloting the change within a few branches is essential. A pilot gives further insight into how to approach the change initiative and typical actions. This learning can be fed back into the target and business case for the project as well as the roll out plan for the other identified branches.
Step 7 – Roll out the improvement intervention
It is common to move from branch to branch rolling out an improvement initiative but this approach is slow and will leave potential benefits on the table. An approach that is likely to yield the best results is to learn from the pilot and then roll out the change to all other branches in the program in parallel.
As mentioned earlier on, decisions made regarding the closure of branches can have serious consequences for a business and for the people within the branches. Using DEA allows the business to take a fresh and unbiased look at the performance of its network, ensuring that any branches selected for closure are not leaving behind unfulfilled potential.
Using the DEA enables the business to identify the branches that will give the highest return on change investment, it provides pointers for what to change and which branches to look at in terms of best practice and it also provides a target for improvements that can drive a change project forward.
By Graham Page, Partner, Tefen UK
Pete Caldwell, Partner, Tefen UK
1 References: Sherman David H. Ladino George. (1995) Managing Bank Productivity Using Data Envelopment Analysis (DEA), Institute for Operations Research and Management Science, March-April, pp 60-73