Article originally posted on The Interline.
2020 has demonstrated that fashion retail cannot build its forecasting models on historical data alone, and unpredictability is set to become the new normal. Digitisation can help, but how?
Fashion retail has long operated on the understanding that yesterday’s data is the best tool we have to predict tomorrow’s market. Macro trends evolve over longer timelines, but from one quarter to the next, one winter to another, aggregated data collected from past performance has been retail’s preferred indicator of potential future performance. From full-price sell-through to the success of a particular channel, campaign, or flagship store, the way to move forward has been modelled by looking back at a broad, blended picture.
This year has revealed the inherent inflexibility of that model. In early spring every industry had its expectations completely confounded. And as 2020 wore on, unpredictability went from being measured in years to being a daily force – introducing new hurdles, and at the same time accentuating the effect of industry-wide shifts that have been simmering beneath the surface for some time. The combined effect? It’s now clear that fashion’s future will not look like its past, and that constant disruption could become the new normal as retail reconfigures itself in real-time to respond as the world changes around it.
But while fashion has been exposed to considerable risk this year, as last week’s high street casualties attest, ongoing uncertainty can be turned into opportunity if retailers can rethink the rigid backend systems and processes they use for buying, planning, allocation, balancing and customer engagement.
Disruption is a definite. Digitisation should be, too.
From widespread vaccination unleashing a wave of pent-up demand on brick-and-mortar stores, to the very real possibility that the shift to online shopping could prove to be permanent, there are many potential scenarios for 2021 and beyond. And whether the culprit is COVID or an unknown future shock with similar – or even greater – impact on consumer behaviour and economic stability, this year’s figures will not be able to paint an accurate picture of the future any more than 2019 numbers could have prepared any retailer for 2020.
So how can a retailer make smart, informed decisions if future unpredictability is undermining the reliability of past evidence? How can they have confidence in timing promotions, managing store allocations and balancing stock, or simply trying to plan for what shoppers are going to want to buy, and through what channel, if nobody really knows where tomorrow’s customer journey is going to begin or end?
Last week’s The State of Fashion 2021 report captures this quandary, and makes a strong case for the retail industry’s pressing need to shift from producing products to meet expected demand – inferred from a blend of past insights – to being led by real, current demand: “Brands should secure high-quality and reliable production capacity and make the long-overdue shift to a demand-focused model to operate in this fluid environment.”
But what steps can retailers take to make that shift a reality? And how can digitisation – which is routinely cited as the key to unlocking a more resilient, demand-led model – help retailers not just manage disruption, but also prepare for a future where predictable consumer demand could very well be a thing of the past?
As this year has demonstrated, a lot of retail’s ability to thrive in the face of uncertainty relies on retailers having access to accurate, actionable information when it counts. And this is where the difference between the way retail has typically captured and analysed data – historically and in aggregate – and the potential of machine learning and mathematic modelling for forecasting demand emerges.
These are buzzwords that retailers and direct to consumer brands are right to be wary of, because they have been mis-applied in the past. However, they are technologies with real applications that can deliver tangible results – especially in the face of uncertainty.
Consider the impact of overstocks and deadstock – with the latter being retailers’ primary exposure to risk in 2020, with stores being forced to close, and inventory left to languish if it wasn’t able to be transferred to online channels. This may be an extreme version of the problem, but the problem has existed in other forms for years and is likely to become even more pronounced as the balance of demand across physical retail networks and online becomes harder to predict.
Both of these issues – along with the related challenge of moving to predictive rather than reactive replenishment, allocation and distribution – can be mitigated, or even solved, by replacing generalised demand curves with more detailed, probabilistic analytics for forecasting. That might sound like a daunting name, but the essence is that, rather than relying on intuition, the right data collection and modelling can instead evaluate every possible outcome and assess which has the highest probably of occurring. For retailers, this can make it possible to determine the probability of any single SKU being sold, across channels and locations, whilst also factoring in commercial considerations such as target margins, markdown strategies and so on.
What makes this sort of model – which is the one used in Nextail solutions – uniquely suited to the current and future state of retail is that they embrace uncertainty. By accepting that anything can happen, probabilistic forecasts factor in the unpredictable, and can give retailers the confidence to hold less inventory, calculate the timing of events and promotions on a dynamic rather than a fixed calendar, and rapidly re-assess their baseline inventory requirements for different channels once those events have ended. And this can be done for individual channels, specific stores, and particular product categories, rather than in a blanket way.
According to statistics gathered by Nextail, large retailers who shift to this model for decoding demand and forecasting based on more detailed data are able to reduce the amount of stock they buy by 20% and cut the incidence of stockouts of popular products by 60%. So, the value is far from theoretical, and it also extends into sustainability, where over-production is an important metric to reduce, and where having a better idea of what will sell can give retailers the confidence to procure or produce less in the first place.
From single slices to a whole.
While online channels have been able to continue trading with comparatively little downtime this year, brick and mortar stores are bearing the brunt of retailer’s need to retool their backend systems, at the same time as needing to become, post-COVID, the standard-bearers for the unified, channel-less “new retail”.
Shoppers now expect smart fulfilment, and the ability to buy online and either pick-up or return in store – that much is obvious. But the heightened demands of the retail revolution also run much deeper: in place of the fixed demand patterns, stock intake structures, and predictable return rates of the past, retailers now need to build assortment, allocation and replenishment systems that can address change in an instant, at an extremely granular level. Because customers conditioned by the responsiveness of digital-only retailers will expect products to be available where they are, with a seamless sale and post-sale experience.
This poses a problem for retailers, especially those basing current or future performance solely on historical data, because physical store data is typically structured from the top down, with stores grouped into fixed, inflexible clusters. Sliced this way, fluctuations in demand and other metrics at the single store level can be lost. In the short term this matters because previously top-performing clusters may be fully or partially affected by localised disruptions, or by other trends that affect performance in a way that past, regionalised data cannot capture. And it matters in the longer term because successful stores that relied on tourism or passing trade from office workers may not see anywhere near the same footfall figures as a result of changes in working patterns, so allocating the same amount of stock to these stores will end in a glut of unsold inventory.
One solution is for retailers to take advantage of advanced analytics in order to structure retail data differently – in a unified, bottom-up way – and then apply machine learning with a purpose to create predictive, and even prescriptive, recommendations. This is the approach offered by merchandising analytics and automation platform, Nextail. By collecting and analysing information at the lowest granularity and translating this into real-time insights, retailers can build an accurate picture of the stock they hold across their entire network (multiple stores, channels, and fulfilment centres)– and then more accurately forecast the stock needed and move products in a more dynamic way.
With store performance grouped into static clusters, retailers’ ability to pivot to capitalise on short windows of opportunity, or to make predictions in the hope of avoiding the risk of inventory remaining unsold, is limited. With more granular, timely insights, retailers gain the ability and the confidence to automate many of these processes, allowing them to take bold steps such as reserving warehouse space dynamically to pivot between stores and channels in the face of micro-fluctuations in demand.
Road-testing an agile approach to the new retail.
Some retailers have already made the jump to becoming more agile and data-driven as a way to cope with disruption and to prepare for a retail market where customer demand is constantly evolving. The latest of these is River Island.
A fixture of the UK high street for more than 65 years, River Island has recently worked to transform its core merchandising processes with technology, allowing the company to keep pace with changes in customer demand and to provide standout, seamless experiences that transcend channels.
To achieve this, River Island has implemented data-driven, cross-channel forecasting, first allocation, and replenishment solutions from Nextail. These have allowed the fast fashion retailer to intelligently automate their key merchandising processes, which has led to a measurable decrease in stockouts and a reduced risk of overstock. And prescriptive analytics now power more granular forecasting and inventory management decisions at the SKU level.
River Island and other retailers have also turned real-time insights to their advantage by using solutions like those pioneered by Nextail to empower both their headquarters and store teams. On top of improving agility and coordination at the operational level, retailers can use the same stock insights to automate inventory inspection and deliver the final component of a seamless, omnichannel customer experience.
And by reducing unsold or wrongly-allocated inventory, retailers that choose to replace their rigid backend systems with a data-driven alternative can also cut the logistics costs of shipping stock and returns unnecessarily between channels – translating to a considerable sustainability benefit.
Retailers today are faced with a lot of questions – how to increase sales, improve sell-through, improve profitability, and insulate themselves against upheavals like the ones that have characterised 2020. But as we approach the end of a year like no other, forward-thinking retailers are beginning to understand that from here on out, no year will be like the one before. And as a result, the way to answer those questions will not be found just by extrapolating from the past, but rather by making decisions informed by current data, with delivery supported by unified systems, channels, and strategies.
This is the essence of the new retail revolution – the idea that simply blending channels into a cohesive whole is not enough, without also fuelling designers, buyers, merchandising teams, and channel managers with reliable, real-time insights that replace their prior reliance on past data and intuition.
Because if the last twelve months have demonstrated anything, it’s that the past can no longer provide a framework for the future now that the rules of fashion retail are being constantly rewritten.