Store grading is something that everybody understands! At a simple level, a store grade is an attribute shared by a group of similar stores. Most often this attribute is numeric as the idea of grading normally carries with it the notion of relative measurement. Thus we often see stores grouped together by turnover band or space. We can distinguish these store “grades” from store “clusters” which are groups of stores that share non-numeric attributes like outlet type, climate or location type. We have all seen store grades used in retail reporting systems, and many of us will have used store grades to help us in Assortment Planning. I am going to concentrate here on the uses of grading in Assortment Planning.
In Assortment Planning we use store grades as a way to reduce the number of decisions we have to make. If we had to plan every store individually we would end up repeating similar assortments many times. Planning at the store grade level allows us to be most effective as a result of our efforts.
What measures should we use to create the grades? Retailers commonly use sales value as the basis for store grading. This allows them to group together similarly performing stores, on the basis that they should have similar ranges assorted to them. As they become more sophisticated many retailers begin to incorporate space into the equation. This often results in a two tier grade system with space sub grades within each sales grade. Superficially, this would appear to be a reasonable approach, and it generally does provide more efficient planning that not using grading at all. However, it places its emphasis on the wrong element – sales.
When we are making decisions about assortments we are primarily deciding which items will go to which stores in which periods. The first question that we need to ask ourselves is how many items we should be sending to each store or store grade. The factor that limits the number of items is not primarily sales velocity (derived from sales value), but space available for display. Space is a limiting factor in bricks & mortar stores in the same way as production capacity is in manufacturing. If we are going to send similar ranges to groups of stores it makes far more sense to group these by space available for display than by sales value. Of course, this assumes that you have accurate records on space at the product level you wish to grade by, and that you have systems in place to keep these records accurate and up to date. But what about sales? Well, we can allow the replenishment systems to pick up on rates of sale and to refill accordingly. We are not ignoring sales, we are merely saying that there are more relevant measures that we can use for grouping stores for assortment planning.
Of course there may be instances where you have more space in a store than is warranted by the sales performance. This is not a desirable outcome, but is one which does exist from time to time. Where this is the case, you may consider reducing the density of assortment. If this widespread you may want to consider subgrading by sales within space.
Once we have decided which measures to use for grading, we also need to decide at what level we wish to grade. There are 4 factors that will influence our decisions here:
- The availability of data for the selected measure
- The availability of a system that will calculate the grades at the selected level
- The availability of a range & assortment planning system that can use the grades at the selected level
- The ability of our transactional systems to execute plans based on the selected level
It is common for retailers to grade at a department level, but as data becomes easier to access and as planning and transactional systems become more powerful, we are seeing more and more people dropping the level down to category.
The next question that arises is “How do we calculate the breakpoints between the grades?” There are three methods that are commonly used:
- Even split
- Exponential growth
- Statistical deviation
Even split, as the name suggests, would create breakpoints at even intervals. Thus for a universe of stores with category level space measuring between 17 and 90 space units, and 10 grades being created we would see breakpoints every 9 units. This method is easy to implement and understand, but tends to be insufficiently discriminating at the lower end and over discriminating at the high end. For example, a store with 80 space units may be ranged quite similarly to one with 90. However, a store with 10 units will have a greater difference to one with 20. Exponential growth splits the grades so that the difference between each grade is constant in percentage terms. In the example shown below each grade split is 1.18 times higher than the previous lower one. This is more complex to design but has the advantage that the gaps between the grades are smaller for the lower grades and larger for the higher grades. The standard deviation method uses the assumed distribution of the space to create grades. The example below uses a normal (bell shaped) distribution to create the breaks. The main limitation with this method is that the space distribution is not going to be “normal” and therefore basing our breaks on a normal curve is inherently “wrong”. However, the results tend to create more evenly populated grades as can be seen below. It is possible to create break points based on the actual distribution, but this is not for the faint-hearted. The table below shows the breakpoints created in our set of sample data using the three methods, and the number of stores in each grade that results from each method.
| Even Split | Exponential Growth | Standard Deviation | ||||
|---|---|---|---|---|---|---|
| Space Grade | # Stores | Space Grade | # Stores | Space Grade | # Stores | |
| Grade A | 81 | 9 | 76 | 15 | 74 | 21 |
| Grade B | 72 | 25 | 64 | 61 | 67 | 39 |
| Grade C | 63 | 50 | 54 | 71 | 62 | 24 |
| Grade D | 54 | 63 | 45 | 44 | 58 | 35 |
| Grade E | 45 | 44 | 38 | 22 | 54 | 23 |
| Grade F | 36 | 46 | 32 | 31 | 50 | 18 |
| Grade G | 27 | 12 | 27 | 5 | 46 | 19 |
| Grade H | 18 | 12 | 22 | 5 | 41 | 22 |
| Grade I | 9 | 1 | 18 | 7 | 34 | 37 |
| Grade J | 0 | 0 | 15 | 1 | 0 | 24 |
So what do we do with these grades? Once we have created our space grades we need to decide how many options we can merchandise effectively in each grade of stores. This is normally done by calculating display units available, average display densities per space unit and average display unit quantities per option as shown below.
| Category X | Grade A | Grade B | Grade C |
|---|---|---|---|
| # Stores | 40 | 60 | 80 |
| Total Space Units | 1200 | 1500 | 1600 |
| Space Units per Store | 30 | 25 | 20 |
| Display Units per Space Unit | 6 | 6 | 6 |
| Display Units Total | 180 | 150 | 120 |
| Units Per Style/Col | 6 | 6 | 6 |
| # Style Colours | 30 | 25 | 20 |
This is not to say that the decision on number of options is based on numeric calculations alone, but we have to ensure that the qualitative decisions are constrained by reality. Once we have done this we can then assign the planned number of options to a grade.
At this point we come up against the micromarketing paradox. To be efficient multiple retailers we need to take account of the different characteristics of our outlets (e.g. demographic and climate variations). However, to plan effectively with limited resources we have little option but to assort at the grade level. A recent survey by the Aberdeen Group underlines this by indicating that micromanaging at the store level is more a trait of laggards than it is of best-in-class retailers. They go on to say “The traits for an excess of store clusters are too time consuming to create and maintain, and are more likely to result in users giving up in frustration than adding meaningful data into the forecast engine.”(1)
One response would be to create clusters within grades using the store attributes (City Centre, Young Metropolitan, Mild Climate etc). However, when we start to do this we end up with an explosion in the number of planning decisions that have to be made. If you had 10 space grades and 5 attribute clusters. You would potentially have 50 assortments to create. If you had 200 stores you would only have an average of 4 stores per cluster. Even though not all of the potential clusters will be populated, this simply is not efficient planning.
The most effective response to this is to plan modular ranges at the grade level, explode these to store level plans, and then to edit individual store assortments for those outliers that really do differ from the norm created b y the grading process. In that way we can maximise the effectiveness of our planning whilst taking account of the most important individual requirements of our stores.
As we can see, well executed store grading is therefore fundamental to the creation of an effective range and assortment plan.
1. “The Business Benefits of Advanced Planning and Replenishment” December 2005 – Source: Aberdeen Group, Inc
.jpg){kind=link}