Which capabilities to prioritize when implementing merchandising systems
As a retailer grows and matures, they will inevitably find themselves, at some point, needing to invest in technology to support the needs of their merchandising team. And when it comes to technology, the entire process chain from planning, right through to execution, is often taken as a whole.
Planning systems typically include support for financial, merchandise and assortment planning and there’s often additional supplemental planning around those. Execution systems, on the other hand, are about forecasting, allocation, and replenishment (DC & store) and may offer additional tools to support those processes.
Even if the decision is made to invest in all of these areas at once, it is unlikely that an organization can implement the technology and absorb the changes brought about by all of them simultaneously. That leaves us with the question; “Which capabilities should I implement first?”
Both intuitively and logically it would seem to make sense to start at the beginning of the process, get planning support in place and then work downstream. However, that may not work in all situations. Below I’ve outlined some things to consider when making decisions about where to start. There are many subsets of analytics and optimization that can influence these processes but, for simplicity’s sake, our focus will be on the core processes.
Traditional Merchandising Systems
These merchandising systems are ideally quite tightly integrated, with plans driving top-down decision making followed by bottom-up execution of those plans with the help of a statistical forecast. Understanding how the technology relates to the main decision points is an important starting point.
Below is one way to think of the typical decision points needed to procure and place products:
Each of these decision points is supported by technology. As technology and retail science has matured, the application of technology to these areas has started to shift. A traditional view of the technology mapping to these processes in a fashion retail environment would look something like this:
Purchasing budgets are set through a top-down process with a company financial plan being broken down into a more detailed merchandise plan that becomes each buyer’s, planner’s or allocator’s budget or “check book” of what they can spend.
They then build out the offer to the customer using assortment plans which effectively work from mid-level product planning criteria (I tend to refer to this as a “middle-out” process). This means using criteria such as style or color, rather than SKU-level (bottom) or department or division-level (top) criteria, and at mid-level location groups / clusters, rather than at individual stores / channels (bottom) or the entire company (top). This process aligns the planned product purchases with the merchandise plan that was decided upon at the top and with the allocations which are typically created at the bottom. The process is often supplemented with variations on these plans including open-to-buy plans, inventory-flow plans, key-item or ladder plans, etc. In a traditional setting, the assortment plan carries a lot of the decision weight in the process covering everything from what to buy, where to offer it, how much to buy, when to move it and how to place it initially.
In some cases, an allocation system will be used to provide the first real, bottom-up view of what specific stores or channels need by spreading the proposed purchase order among stores. This helps to ensure that what’s proposed in the assortment plan properly considers the needs and capacity of stores before committing to the order. The initial allocation is then either driven by the assortment or the allocation that validated the assortment.
Once a product begins its life, it is often supported by a replenishment system that tracks store- / channel-level selling, forecasts future activity and inventory requirements and makes sure inventory is in place to meet the demand as it occurs. This is often the first point in a traditional fashion environment where the art of merchandising meets the ‘science’ of retail in the form of technology. Many fashion retailers shortcut this process by using an allocation system to cover replenishment of products within their life.
Finally, at the end of a product’s life, before products are marked down, any remaining inventory is pushed out, usually by the allocation system, ideally to the stores most likely to sell it.
An Evolution of Merchandising Processes
Retail ‘science’, technology, and process have evolved to the point where many retailers are rethinking how these pieces fit together. This is in part because, as an industry, we have struggled to find a truly effective assortment planning tool. Part of that struggle is because we’ve been trying to make those tools do too much at once. In addition, technology has evolved to enable us to apply a more scientific approach and bottom up understanding of the process. Putting these two things together enables us to rethink and simplify the process as illustrated below:
Looking at this from right to left; final allocations, replenishment, and initial allocations utilize modern technology to optimize their results for their intended purpose. The forecast, which had often traditionally been embedded in (and was only useful for) replenishment, has now been optimized individually for different product types and can support allocation and planning processes in addition to replenishment. In fact, well thought out solutions have integrated these three processes, not just from a data perspective, but functionally as well. Functional integration unifies the look and feel of the processes and makes it easier for users to manage them end to end, whereas historically there was often a split meaning that different processes, such as replenishment and allocation, needing dedicated expert users.
This configuration also enables us to apply the logic needed to generate an order plan driven by the forecast. Then we can both analyze the details of more accurate store forecasts, and generate an inventory plan taking these into account along with real-world constraints such as presentation requirements, pack configurations and lead times. This becomes the new bridge between the ‘art’ and the ‘science’, where bottom-up details are married and automatically rationalized with the top-down plan, while generating a much more accurate order quantity.
The result of this is that the pressures on the assortment plan, in terms of generating order quantities, breaking that down into individual orders and managing the initial allocations, have been significantly reduced. Now the focus is on the ‘what’ to buy and the ‘where’ to make the offer. While sales volume can inform this decision, it’s no longer necessary to attempt to get to a precisely accurate volume to drive ordering. The order planning process will take the product and where it is being offered to generate a more precise order quantity. The characteristics of products and the relative volumes of the items in the assortment plan can guide the order plan to adhere to the buyer’s (or planner/allocator) intent when assembling the assortment. In some cases, higher level forecasts and/or iterations of the detailed forecast can be used to validate the assortment as well.
The top-down planning process remains in place, although now with support from forecasting mechanisms that provide a check and balance, or even guidance.
The Numbers Game
To get a perspective on what will ultimately be influenced, let’s compare the three major components of merchandising. Let’s take as an example a retailer offering a range of fashion and basic merchandise with three distribution centers (DCs) and 500 stores. Supply chain professionals working in different sectors, companies, and departments see the following activities in slightly different ways, but for the purpose of this example, they are broken down into assorting, ordering and allocating / replenishing as defined below.
Assortment planning (10 decisions)
Defined for purposes of this discussion as determining what products to offer. We generally have one major objective: determine what products to buy or not to buy. If we include decisions around ranging (which stores get the products we select), then we also make this choice for the stores. In virtually all fashion environments, stores are clustered into groups by sales volume, demographics, geography or similar groupings. If, for the sake of this example, we assume 10 of these groups, then we’re making 10 ‘include or exclude’ decisions per product. The difference in what products to order is unlikely to be significantly altered by using an assortment planning system. The decision that will typically have the greatest impact on the quality of the result is how the stores are to be grouped. Beyond that, the most substantial benefits tend to come from the consistency of a process and having all the data held in a single merchandising system for easy comparison.
Ordering (12 decisions)
Defined as determining how many of the items selected in assortment planning should be shipped to a warehouse or DC at any given time. Here we’re making the same number of decisions as the number of DCs. This is multiplied by the number of receipts we plan. In an environment with 1/3 of products being one shipment (only received once in the DCs), 1/3 being two shipments (having a second DC receipt for fill-in) and 1/3 being ongoing basics, we may have an average of, let’s say, 4 receipts per product. If we have 3 DCs, that’s a dozen decisions per product (3 * 4 = 12). Here is where technology can begin to have a larger impact on the actual results. Especially in the evolved process described above, we’ll be matching the quantity ordered more closely to the future demand by store / channel and the constraints around them.
Allocation & Replenishment (2,000 decisions)
Defined as determining how much available inventory goes to each store. Here we also have decisions to make for each receipt. If we use the average of 4 receipts from above, we need to make a store-specific choice for each store, for each of those receipts. In a chain with 500 stores, we’re now talking about 2,000 decisions (500 * 4 = 2,000). In the case of direct-to-store ordering, generally allocation is combining the ordering and allocation steps. Ultimately the difference in applying technology is to move a few units or case packs from stores that would sit on them to stores that would successfully sell them. That may not seem like much at first glance, but doing that with each receipt of each product typically represents millions if not 10’s of millions of dollars of additional profit to a retailer of this scale. This is before the capital savings that come from holding less inventory overall.
Using the above logic, there are clearly many more decisions in the process of allocation than in ordering and assorting. Obviously, there are multiple considerations for each activity, but ultimately allocation has more instances where good decisions can be helpful, or perhaps more importantly, where bad decisions can be detrimental.
Time to Value
Allocations are often carried out by less experienced personnel in many retail environments. When I started in corporate retail, the role of allocator was a junior position that was filled with people just out of school and with no retail experience. It always struck me as being a poor structure since a good product purchase can quickly be ruined by bad allocation decisions.
As technology and processes have evolved, more opportunities for the data to drive activity have been incorporated into allocation processes. By using parameters that can often be sourced and integrated from other existing processes and/or preconfigured, we can now reliably automate most allocation scenarios. Merchandising systems can make more detailed evaluations of the many small decisions made in allocation and replenishment scenarios. The results of these decisions are then systemically validated and only exceptional situations require attention from an end user. This can focus the activity into areas that either:
a) allocators can be better trained to address or
b) since less effort is required, that can be dealt with by more senior personnel.
Less time is spent and the results are better.
In addition to time saved on the process side of allocating, modern technology has made implementation timelines for allocation, replenishment and order planning technologies much shorter – especially in SaaS environments. This enables some modern allocation systems to be up, running and making a return on investment in a matter of months. They can be piloted within part of the organization or its operations and rolled out as quickly as the changes can be absorbed. In situations where existing processes have left some stores with far more inventory than they can sell and others with far less, immediate gains can be made through analytically driven stock balancing logic.
Compare this to typical assortment planning projects that still commonly take well over a year (and sometimes many years) to get up and running. This is partly due to the higher degree of “art” in the assortment process which varies dramatically from retailer to retailer – and often even across merchants (i.e. buyers, planners, and assorters) within retailers.
Measuring the value achieved in fulfillment and ordering is more straight-forward as well. Since decisions are now primarily based on data, we can easily measure the results against earlier data and determine how much value the new technology and processes are delivering. Measuring return from planning processes has historically been very subjective and therefore the resulting measurement of value is far from empirical.
So Which Comes First?
If you are in an environment where you need help in all three of these areas, what then? Which should you focus on first? Well, each situation is unique and these choices are dependent on your current capability and proficiency. Generally, there are two reasons why it makes sense in most situations to focus on allocation & replenishment first and work upstream to ordering followed by planning disciplines.
The first reason for this is illustrated in the numbers game section above. The more chances there are to improve the quality of the decision then, generally, the more impact there is on the bottom line. Of course, if you do a better job of choosing the “perfect product” it will likely result in better performance. However, it’s rare, that choices with such dramatic influence are missed by merchants in the process of assortment planning. It’s much more common that over assortment is an issue.
This leads us to the second reason to consider allocation & replenishment first. If you make the “perfect” assortment choices, and even create ideal orders to DCs, poor allocation can still irreparably damage the results you get. If, however, you make “good” decisions on assortment and ordering (which is common since there are fewer choices being made and therefore more thought going into each), improved allocation can make the best of what you ultimately end up with. These improvements, if done well, can almost always have more impact than changes to ordering and assorting processes. The possible exception is when applying the evolved order planning process describe previously, which can feed the allocation process with more appropriate available inventory levels.
Starting at the end and working up frequently generates enough return to fund investment in the other two areas as time permits and as your business can absorb the change.
Retail Continues to Change and Evolve
One final additional thought; complexity is the reality of today’s retail landscape. Customer behavior is changing at a pace never before seen in retail. Between economic influences, brand loyalties, channel competition, fashion preferences and other factors, today’s customer is more unpredictable than ever. This change is felt differently at each individual store, so it is important to have visibility to those changes and have the ability to respond to them immediately. The bottom-up processes of forecasting, allocation and replenishment are the last chance to identify and react to these and therefore it’s the closest you get to meeting the demand that your customers represent.
The last chance to get it right is logically the first place to invest in doing a better job.
Want to know when similar resources are published?
Subscribe to receive a monthly digest of our most valuable resources like blog posts, whitepapers and guides.
- Machine Learning in the Context of Retail Demand Forecasting
- Artificial Intelligence Drives More Efficient Goods Handling in Retail
- Decision Science and Pragmatic AI in Retail
- Why Category Management Should Always Start with Behavioral Clustering
- Pushing the Boundaries of Data Processing Requires Specialization