Every week, The Interline rounds up the most vital talking points from across the landscape of fashion technology news. This roundup is also delivered to Interline Insiders by email.
COVID could catalyse the need for turnkey machine learning and accessible automation.
Machine learning and artificial intelligence are a complex umbrella of different technologies – from computer vision to the sort of speech synthesis that powers the audio version of The Interline’s articles. But one thing those different technologies have in common is that, over the last few years, they have gone from being reservoirs of potential to being proven avenues for cost-cutting, product iteration, market testing, and much more – all at a scale that would be impossible, or undesirable, for human beings to accomplish.
“Out of the hands of research scientists, ready to start delivering new revenue streams and optimising existing ones, AI is currently in a tricky transitional phase between promise and proven deployment.”
That was true in 2017, but it certainly isn’t true today. Entire circular fashion businesses have been built on the automation potential of machine learning, using carefully-trained models to take on the burden of scanning, cataloguing, and even pricing second-hand items. At the same time, machine learning models are already powering the product recommendations and eCommerce frontend experiences of many, if not most, online retailers, and everything from material scanning to trend analysis is beginning to benefit from the incorporation of some degree of AI-labelled technology.
And of course one of the primary proof points of AI was that Stitch Fix went public just a few short months after WhichPLM released its 2017 publication. Today Stitch Fix shares are trading 450% higher than they were at the point of that initial valuation, and much of that value comes from the company’s underlying technology, rather than its ability to deliver personalisation at scale.
So far none of this is news, but this subject caught The Interline’s attention again this week with the unveiling of no-code machine learning startup Levity, which was profiled in TechCrunch. Now, Levity is not targeting the fashion industry, but what it is targeting is the ability for essentially anyone to take the repetitive, rote portions of their job role, train a model on a past data set, and automate those activities – all without knowing anything about how machine learning actually works.
This might sound mundane, but it’s interesting precisely because it’s so unremarkable. Unlike a glitzy startup like FINESSE, which is entering the transgressive area of using artificial intelligence to assist with design, Levity and the other, similar platforms that will no doubt follow it is proposing to use machine learning as a way of lifting the burden of administration and non-value-added tasks from a range of different industries. And whether we recognise it readily or not, there remains quite a lot in the typical fashion design, development, and production workflow that could quite easily be automated.
(The Interline will have more on FINESSE and the implications of digital design that tries to sidestep the strict, human labour-intensive 3D to 2D pattern workflow next week.)
Fashion design, of course, is a creative discipline. And indeed a lot of the friction around technology implementation has historically come from creative professionals feeling as though the addition of new technologies comes with an increased administrative overhead – one that detracts from core creative activities. Consider the generation of technical specifications, the labelling of artboards, the categorisation of artwork assets, the updating of supplier audit information, the costing cycle, and so on.
All of these are opportunities for automation. And while some existing platforms are already providing tools for streamlining these, by incorporating machine learning into their existing data models, that is a different prospect to the idea of an off-the-shelf automation engine that a creative designer, a technical designer, a sourcing manager, or anyone else in the value chain could feed with past data and simply use.
To refer one last time to our collective archives, the same WhichPLM deep-dive on AI and machine learning included the following statement:
“One of the most prominent reasons that businesses turn to technology is to avoid increases in overhead and headcounts as consumer demand and market pressures place greater stress on processes and individuals.”
It’s undeniable at this point that “pressure” and “stress” have both increased exponentially for many businesses and individuals over the course of the pandemic, which is likely to be place an even greater emphasis on creative and value-added processes, turning rote and repetitive tasks into even more of a burden than they might have been before.
For this reason, The Interline expects to see cross-industry solutions like Levity adapted to the needs of fashion very soon – or for more fashion-focused solutions to begin incorporating even more in the way of similar turnkey solutions. Because as cross-industry options continue to mature, fashion businesses may find themselves looking over the garden wall and cherry-picking from platforms that are already proven elsewhere.
Channel blending emphasises the need for the integration of channel-specific technologies.
This week’s announcement that online powerhouse ASOS would be acquiring Topshop’s brand and its stockpiles, but giving the cold shoulder to the store workers, commercial landlords, and suppliers that relied on its physical retail presence was hardly a surprise – especially in light of the 124% year-on-year growth that UK fashion has seen from eCommerce from January 2020 to January 2021.
ASOS are, it seems, planning to use the Topshop name for international growth, but its disinterest in the company’s established retail network could symbolise something deeper: the difficulty of building truly cross-channel systems and processes. By retaining only the online portion of the Arcade Group business (which, to be fair, was growing), ASOS will be able to sidestep some of the pitfalls of integrating online and offline systems that, until very recently, have been pitched against one another.
This week, The Interline enjoyed listening to a Modern Retail podcast interview with Aaron Sanandres, of UNTUCKit, where he spotlighted precisely this problem:
“The fact is, if you’re on Shopify, you will have a very difficult time executing a very clean buy online pick up in store. Shopify doesn’t allow you to split carts. So if you add two products — one that’s in the store, one that’s not in the store — Shopify doesn’t allow you to transact that in one transaction. So that is a massive gap for you to be able to to execute a very seamless experience.“
As vaccination programmes continue apace, and as different countries begin to look at relaxing COVID restrictions with varying degrees of speed, the question of what to do with physical real estate is going to arise time and time again. But whether the answer is to turn them into temporary (or even permanent) fulfilment centres, the way UNTUCKit did, or to transform them into experience centres the way China’s “new retail” businesses are doing, the key to ensuring that they remain part of a cohesive whole is going to be integrating online and offline systems to a single backbone. And while that may be a challenge too far for this particular acquisition, in these particular circumstances, it’s a challenge that all of fashion retail should be preparing to face.