“Sorry, we don’t have your size in this store.” Sound familiar? This all too common issue of stores not having the right sizes is the problem fashion retailers are trying to solve with their allocations process. On face value it seems like a fairly simple challenge, but dive a bit deeper into the complexity of analysis required and the permutations in play and you can quickly appreciate the scale of the task.
Let’s say a 100 store fashion retailer has 3000 new products per year, each product has on average 5 sizes. That’s 1,5 MILLION different variables that require analysing and forecasting in order to get the correct store stock levels to make sure that no sale is missed and there is little to no inventory left over at the end of the season. Take into account the previous season’s stock holding and that number grows even further. Growing your store footprint? Great! But more stores adds exponential complexity and workload on the team.
Get this part of the value chain wrong and you sell out quickly in stores with high demand for sizes resulting in missed sales and slow selling stock in others which need to be moved between stores at a cost or worse, marked down at the end of the season impacting gross margin.
So where to start?
Data is key
As with everything the quality of the data used when forecasting and analysing inventory performance in order to generate optimal allocations is critical and the first and foremost area to focus is the product categorization hierarchies.
Product hierarchies are often set up without much thought to the allocations specific challenges and can have a wide range of non homogenous products in the same category. All products don’t sell at the same rate in a retail business. Higher priced items, for example, will sell less frequently then price entry items. Having a mix of these products in the same product category creates noise in the sales patterns and results in difficulties when trying to forecast demand at these higher levels in order to prepare an allocation.
Solution: Invest time periodically reviewing the product categorisation and level of homogeneity to work towards better understanding the sales flows and creating more uniform product categories.
Understanding the optimal size curve for a given category is essential in order to optimally allocate and re-order products for that category. This is another reason why creating homogenous product categories is important as without this you won’t be able to analyse and accurately manage an optimal size curve for the product category. This will impact your ability to purchase the optimal size curve at buying stage but also will limit your accuracy when analysing store level category size curves.
Ultimately you need to generate an ideal size curve by category by store is vital if you are to begin applying science to your allocation method. This is where the level of complexity typically starts to blow out and where the limitations of spreadsheet based allocations are exceeded.
Solution: Start by analysing your size curves in your homogenous product categories at the company level. Periodically reviewing these based on recent sales history will help with improved ordering and will set you up for beginning to analyse product category size curves by store. If you have 15 or more stores, this will quickly become too time consuming so consider looking at key stores in the business and starting to improve there. The next step is to look to the power of retail systems that can accommodate this level of complexity and volume of data crunching.
Every store will sell your set of products in varying degrees of difference. Each store also sells its sizes differently according to the demographics of the local customer base. Getting the stock levels right at a style/colour/size/store level is the objective of any allocation process and to do this you need to be able to forecast the demand at the lowest level possible.
Due to the limitations of spreadsheets and conventional allocations systems where demand is purely a function of historical sales, something new fashion products don’t have, businesses try to measure demand at a category level by looking at historical sales flows across the year. This method is pervasive in the traditional fashion allocation solutions and while it delivers good results it requires the consistent management of product hierarchies, store level seasonal and assortment planning and size curve planning which in turn requires an extensive team to manage this across each of the different product areas.
Solution: The future of allocations is machine learning and AI technologies. These new approaches unlock previously dreamt levels of computational power allowing retailers to forecast demand for new fashion products with no sales history right down to the lowest level and apply optimisation algorithms that take the numerous business rules and constraints into account to generate the most accurate ideal store allocation possible.