Break through to the other side
30 October 2023
AI could help fashion logistics break the constraints that have long held it, says Stuart Higgins. But while the technology is exciting, Stuart also urges caution.
HISTORICALLY, THE fashion industry has had to manage around a number of inconvenient truths that have shaped its approach to demand forecasting, replenishment, inventory management, as well as markdown and clearance.
Chief among these has been the trend, over time, to source increasing volumes of merchandise from the Far East, with the consequential impact this has on forecast accuracy. In short, the further out you forecast, the more inaccurate you are, and the inability to replenish mid-season from the Far East due to the long lead times typical of ‘far-shore’ manufacture is an issue.
Another significant factor that has influenced supply chain management within the fashion sector is the shear complexity of trying to forecast sales across multiple ranges, incorporating multiple colour / size variations within multiple stores. The mathematical complexity quickly outpaces the ability of the average business to cope. Within a typical range of 5000 products, sold across eight size combinations and six colourways in 350 stores generating a potential 84 million items/store combinations that require forecasting – something that is way too much for the average merchandiser to deal with.
So instead, fashion retail has adopted a simpler replenishment model that makes forecasting easier, but that also injects ‘planned inefficiencies’ and workarounds into the management of a typical season. Seasonal forecasts are generated at the aggregate level, largely based upon last season’s averages for equivalent products. The entire season’s volume is ordered ahead of time and brought into the country before season launch. Ranges and Store Assortments are created based upon groups of ‘similar’ stores to ease the burden of forecasting sales at an individual store level. The majority of the inventory is allocated to store ahead of season launch – often because there is insufficient space in central distribution centres to hold back stock for agile replenishment once actual sales patterns start to emerge. As a result, the industry accepts that it will probably generate no more than 60-70% full margin sell through and have to get into a pattern of markdown and clearance activity to clear stock levels that are primarily associated with managing around the inefficiencies and approximations injected into the supply chain by all of the above compromises!
One of the reasons the retail industry is so animated about the advent of generative AI is that it suddenly opens up the possibility of managing fashion supply chains in a very different way. Generative AI provides the mouth watering prospect of being able to tailor ranges and assortment for individual stores and to easily manage the mathematical complexities of managing tens of millions of forecasting and replenishment decisions on a daily basis.
Generative AI affords fashion retailers the luxury of being able to use many more, disparate, data sets to improve replenishment decision making. Store assortments can be tailored for the local customer demographic, including historical and future buying trends, catchment areas profiles on household expenditure, population density, competitor offerings, even size profiles, to create the prospect of a ‘perfect’ store assortment matched to the needs of the local population. Seasonal buying trends can be ascertained early in the season by comparing store / sku sales to previous season trends and forecasting future demand at a store / sku level for the balance of the season. Replenishment orders are therefore based on a fundamentally more accurate forecast and decisions can be made earlier in the season about under-performing products that may require markdown to clear, affording the opportunity markdown at a lower level of discount to stimulate demand earlier in the season and reducing overall cash investment in markdown and clearance to improve net-net margins across the season.
So we have the prospect of generative AI delivering better ranges, tailored to individual store needs, with higher on-shelf availability in-season, higher sales and reduced levels of markdown and clearance generating a more profitable season end-to-end – the panacea for fashion retailers.
AI in application
This isn’t just speculation. We are starting to see retailers using generative AI already:
• AI generated Store Assortment Planning is already being used by ASOS, among others, to tailor the range you see on screen when ordering online. By analysing customer data, including purchase history, browsing behaviour, and style preferences, ASOS utilises AI algorithms to curate the most appropriate product assortment for each customer, offering a personalised and tailored shopping experience to each of their customers online.
• As mentioned, AI brings with it the capability to both ‘educate’ the algorithms using many more databases of information, such as individual store footprints, weather patterns, local demographics, competitor mix, online sales presence, local events, etc. And at the same time to have the computing power to do this at an individual store level – providing much lower levels of granularity and tailoring the assortment to the needs of an individual store. Examples of retailers pioneering this capability in the UK include Tesco, M&S, Sainsbury and Boots and it is certain that the fashion sector will soon adopt this capability given the power to drive more accurate seasonal sell through.
It’s very early days for the application of the technology, is the cost of living crisis having an impact on fashion retailers’ appetite for the investment and experimentation needed to make a radical new tech work for them?
The retail fashion sector is an interesting one, where retailers are often ‘betting the business’ on the latest trend or fashion. Fashion retail balance sheets are not always as robust as one would imagine and often the retailers are literally betting the future of the business on the next season’s buy. The cash flow impact of ordering product for the next season’s launch is so great that, if their designs are not quite on-trend or if they order too much and have to invest disproportionately in markdown and clearance, then the cash flow impact can literally break the business. That is why we often see fashion retailers, who have not matched their planned Christmas trading performance, going under towards the end of January when quarterly store rental payments become due and they simply have not got the cash in the bank to sustain a large cash call on the business because Christmas under-performed and there is no money in the bank.
Given that context, it is not surprising that many fashion retailers are cautious about investing early in new technology trends, because of the uncertainty around benefits delivery. Many fashion retailers become more cautious in times of high inflation or economic hardship, where consumer confidence is falling, because it magnifies fears around the ability of the core business to hit its trading targets and magnifies concerns about cash flow - so speculative investment in technology trends such as AI are very mush pushed onto the back burner, even though they may provide very attractive potential returns on investment.
Stuart Higgins, partner, BearingPoint
For more information, visit bearingpoint.com