We are entering a new age of production. From multi-national enterprises to small-to-medium businesses, consumer expectations for speed, quality, and omnichannel service are prompting brands and retailers to try and better understand and control their manufacturing processes to a degree that hasn’t been seen since offshoring became the norm several decades ago. And in the process, the long-standing “push” supply chain model – where production levels are set based on historical trends and purchasing patterns, and in-season responsiveness is extremely limited – to a “pull” value chain, where real product demand and machine-learning-derived insights actually drive production.
The problems with the traditional push-based chain are easy to see. Disconnected design, development, procurement, manufacturing, logistics, warehousing, and retailing processes mean overstocking or understocking are both common. And bottlenecks and delays mean replenishment often comes too late, or obsolete products are left to clog shelves and stock rooms, before eventually finding their way to landfill sites.
A pull-based chain, on the other hand, can be driven by real demand through a fully connected, all-digital workflow that begins at the point of sale. Or alternatively algorithms that can accurately forecast demand through an introduction, rise, plateau, decline, and obsolescence curve can help shape the timeline for production. The benefits are also obvious:
To be clear: I don’t believe that the pull chain can replace the push chain completely. Not every product needs to be demand driven, for one thing; basics that are never out of stock can continue to be produced in the traditional way. Although I should note that production for these types of products should still be demand-driven in the sense that manufacturing should be triggered by minimum stock levels being flagged up by Enterprise Resource Planning (ERP) for true automation of essentials.
For trend and fashion products with a limited window of success, though, the pull chain offers several clear advantages.
But how does a pull chain actually work?
Intelligent inputs
The traditional retail world has been under constant pressure for several decades now – fighting off giant online sellers like Amazon, JD.com, Alibaba, eBay, Rakuten, Zalando, and a constant influx of start-ups that all have one thing in common: machine learning. Each of these businesses operates on a analytical, data-driven model that allows them to identify market opportunities, understand (and even shape) buying behaviour, and then leverage their lean, hyper-fast processes to deliver what customers want at a much lower cost to themselves than any traditional retailer can match.
Unlike other technologies, where a workaround or a different approach can sometimes be just as good, the AI and machine learning arms race will leave non-adopters in the dust. To match the levels of customer satisfaction and retention that the eCommerce retail leaders have achieved, other retailers will need to incorporate at least some level of machine learning into their design, production, and sales cycles. Brick and mortar business can supplement this with the customer relationships they have been able to build in the real world, but unless that in-person data includes complex variables like purchase intent, basket size, and social status, the chances of a physical retailer being able to know their customer better – or even as well as – than their online counterparts are slim.
But retailers need to understand that integrating elements of machine learning into their eCommerce storefronts to create smarter recommendations and cross-selling opportunities is just the beginning. A true pull-based chain will have to be smart both up and downstream. Knowing your customer means better service at the point of sale, sure, but that knowledge can also flow upstream and allow for the development of new and better-fitting products that respond, in-season or even more quickly, to that customer’s particular needs!
AI is not something that can be adopted on a piecemeal basis. From automating repetitive tasks to delivering completely new insights and transforming job roles, AI and machine learning can only properly feed a new model of production when they are adopted organisation-wide. And to take measurable steps in this direction, I believe retailers and technology vendors need to be working on proof-of-concepts that bridge hardware and software to map out the inputs and outputs that will drive a new, continuous loop of design, development, production and retail that evolves based on real-time demand.
Planning and pre-production
A pull chain will require a new standard of real-time planning. From merchandising to design, processes will need to be directly informed by demand as well as balancing the market’s needs with the capabilities, machinery, capacity, and availability of manufacturing partners.
At the same time, new design solutions – from 2D CAD to mixed reality – and digital material libraries are offering bi-directional data connectivity with PLM and ERP, allowing brands to visualise and control the impacts of their design and material choices on production, and improving the odds of products striking the right price point and quality level.
And when the time does come to send an order to production, Blockchain technologies potentially offer multiple ways to ensure that the speed and efficiency gains made in design and development aren’t lost in negotiation with suppliers. From smart contracts that can automatically issue letters of credit to public blockchains that provide an open book on transparency and product provenance, Blockchain could offer serious benefits to businesses looking to create a new, more accountable model of production.
To the factory floor
Everything I’ve discussed so far falls under the umbrella of pre-production. Trends, demand planning, style and concept development, contracts, and raw material sourcing – all of these things need to be in place before actual manufacturing can commence.
When production does begin, we start to see just how different the factory of the future might look. Material and component suppliers will receive their orders electronically, and data embedded in those orders will then help set-up and drive knitting, weaving, printing, dyeing, and inspection hardware linked to measurement data that can be shared with others partners along the value chain. And as those machines work, predicted throughput times and estimated delivery details will be visible to both the factory management team and the brand or retailer who commissioned the order.
Armed with those insights and that level of connectivity, manufacturers will be able to plan their own operations to a more granular level – taking a weeks-long process down to days or even hours. And the same information can also be used to coordinate the factory’s output – finished products – with freight schedules planned in advance based upon accurate call-offs , whether they’re delivered to distribution centres for more evergreen products, or going straight into retail channels or direct to the consumer.
More specifically, the data generated by each piece of connected manufacturing machinery – whether it’s a digital knitting machine or digital thread dyeing hardware – for detailed production planning. Throughput times, machine types, operation breakdowns and standard minute values can all be taken account of to allow factories to provide customers with accurate labour rates and turnaround times – reducing the fashion industry’s reliance on factory overtime and inequality. This data will also provide new tools for buyers and sellers to negotiate fair pricing.
In the cutting room, physical materials can be batched ahead of time based upon data coming from the mills – by shade, length, width, or quality – with the resulting information being shared with an automated marker-making solution that can optimise material utilisation and yield linked to the order breakdown . Today, the actual batching and handling of materials is done by people, but in the near future I expect to see greater use of material handling robotics supporting humans, that will not only batch and categorise materials, but also transfer components between hardware stations.
We have already seen digital printers and cutting machines having their speeds synchronised so that one directly feeds the other, but greater data sharing between spreading and numerically-controlled cutting machines, among others, will open the door to robotic and automated overhead rail systems that will go on to transfer parts between operators linked to factory planning software solutions. And this is on top of the detailed maintenance information and telematics that connecting manufacturing hardware has shared for servicing purposes for several years – and is what we refer to as The Internet of Things (IoT).
Finally, sewing (or machining operations in factory lingo) will be able to rely on dyed-to-match thread that’s produced in real-time, rather than relying on cones of specifically-dyed thread to reach the factory from elsewhere. And when we consider that thread for smart sewing operations can be produced in precise quantities linked to line balancing requirements – reducing the wastage caused by the current model of having to order thread in bulk, with uncertain colour matching.
A new view on the value chain.
As production progresses, brands, retailers and even consumers will want to be appraised of changes. By using smart tagged bundles – with QR, NFC, RFID or other labelling and tracking solutions – every order can undergo work-in-progress monitoring, helping to identify issues early and prevent delays.
But while this level of insight can, in theory, bridge the gap that has long existed between the push model’s “out of sight, out of mind” attitude to production and the goal of a truly connected, transparent factor, it’s important to remember that the scale of investment required to deliver the latter is large. So brands and retailers may wish to jump straight to using all these different technologies to create their own in-country production facilities, but the cost and time involved in jumping from push to pull this way will be prohibitive.
Instead, reversing the flow of production will need to be a multi-stage process – and one that connects, over time, each of the value-chain partners, upstream and downstream.
As I wrote last week, factories in China – where investment in technology is high – may have some of these technology pieces in place already. Replicating their setups elsewhere will not be an easy feat technologically speaking, and that’s before we consider the challenge of training and onboarding the kind of resources who both understand production and are also willing to learn new technologies.
In my mind, there is little question that the push method of production has a limited amount of time left as the dominant model. As a mindset it fails to take account of just how rapidly the fashion market – and consumer expectations – is evolving. As a technology and investment strategy, the efficiencies it can realise will be incremental compared to the huge potential offered by moving instead to a demand-driven pull chain.
Making it work at home or abroad will not be an easy task, but I believe that the factory of the future is now firmly in reach. And while there will always be a place for offshore, high-volume production, even that paradigm is changing as a result of automation, risk mitigation, and the knowledge that making products destined for discount or landfill is no sustainable – in any sense – in a world where real-time data and demand can truly inform production.
