In this exclusive guest piece, Mangal Anandan, Vice President, Products and Solutions for First Insight, Inc., explores some specific applications of machine learning in relation to PLM systems. First Insight empowers retailers and brands to design, select, price and market new products with confidence.
Digital transformation will be the key to survival for many retailers and brands. We recently saw Mickey Drexler, who just announced his resignation from J.Crew, admit in a Wall Street Journal article that “he missed it,” referring to his underestimation of the speed of change that technology would bring to retail. With the advent of big data, machine learning (ML) and other technologies are introducing new ways to manage everything from customer engagement, to design, to supply chain, to product lifecycle management and beyond.
As more and more retailers and brands consider how best to leverage the latest technologies, it is time to look at a few best practices being embraced by early adopters in retail, and how the outcomes are driving the evolution of PLM processes and systems into the future. But first, let’s take a look at why the idea of machine learning makes sense for retail today.
Machine learning is actually not a new idea; it was popular back in 1960s and again in the 80s within the theoretical realm of computer science. But the lack of available data, coupled with computing systems incapable of handling large volumes of information, kept the idea from becoming tangible reality. Now, in 2017, the combination of big data and advanced, inexpensive computing systems has made machine learning practical in real work application, and it is driving positive change as part of the product development lifecycle.
In the context of PLM software, it is no different. PLM systems are collecting vast amounts of input data, with most of the product attributes housed in a PLM system, from seasonality, materials, specifications and more. And the power of this data can now be harnessed through machine learning models.
Here are a few machine learning applications taking hold within the retail community which demonstrate the impact of this advanced technology on PLM. First, however, it is worth noting that all of the advanced technology in the world will not be effective if the solution being used is not proven, and the input data has not been generated by asking the right questions.
1. Decision Support Systems
While some define ML-powered decision support processes as the ability for machines to make decisions, the truth is that the decision is ultimately in the hands of the user. By utilizing data and overlaying machine learning, companies are empowered to make the best decision through a semi-automated process that delivers outcomes in the form of a dashboard, which can often be manipulated to forecast outcomes based on altered products, timing and more. These kinds of artificial intelligence solutions are becoming ubiquitous in many industries including retail, as they add a layer of intelligence that brings confidence to the decision- making process. Two examples of machine learning for retail decision support include product recommendation engines, which can better predict and align to how a customer’s needs change over time, and dynamic pricing, which provides retailers with the ability to consider a multitude of pricing variables including seasonality, supply and demand to optimize their pricing with increasing frequency.
2. Optimized Line Planning
This artificial intelligence-based application provides actionable insights to retailers earlier than previously possible in the product decision cycle. The application factors in performance of products that have sold in the past in combination with how they performed at the transactional level, capturing real-time consumer input and attributes of that consumer, and the price at which they are buying. The system then recommends different design options to designers to ensure they create designs with the greatest chance for success in the following season. For example, a recommendation could suggest a button-down collar and front pocket to ensure it appeals to a specific segment of customers. The Optimized Line Planning solution helps retailers and brands understand key characteristics of their customer base and how they impact revenue and margin, inform the assortment strategy based on the understanding of the consumer, and create optimal line plans and design products guided by specific attribute recommendations from the tool.
3. Supplier Management
Supplier Management in PLM is often integrated with supply-side systems like Purchase Order Management, specifically with factories in the Far East. The process of selecting and sourcing vendors has remained largely manual and time consuming. Through machine learning, artificial intelligence systems can gather real-time data on vendors based on on-time delivery, quality of products and raw materials, and more, continuously learning and tracking performance of vendors and their deliverables. The system can then automate the process, offer suggestions to the retailer or brand on vendor selection, and issue a PO based on a risk-adjusted model.
Additionally, while it may feel like blockchain is still in its infancy, the fact that it has already started to move the needle for retailers and manufacturers makes it worth noting. Blockchain is essentially a shared ledger where all transactions are recorded—eliminating the errors that can occur when each party participating in a transaction maintains its own data set for that transaction. In retail, blockchain allows all parties—supplier, manufacturer, retailer and end consumer—to trace and verify a product’s journey. Blockchain will be able to ensure, for example, that a new Chanel handbag is not counterfeit, and enable a consumer to track the steak he buys from farm to plate. Not only will this create greater efficiencies in the supply chain, eventually ‘approvals’ will be automated with ‘pre-approvals’ given that specified criteria has been satisfied with no need for manual verification, increasing speed-to-market significantly. A recent HBR article by Casey and Wong offers a peek into some exciting examples of this technology applied to the supply chain.
However, one thing to consider is that various brands and retailers must address the reality of less-than-perfect data, typically inherited from siloed systems of past decades. For example, they often give unique names to the same product attributes like color values or fabric types. An effective machine learning process requires high quality data on the products they sell in their own stores, as well as across the industry. Standards need to be adopted more widely within the retail industry. One way this could work is to maintain a standardized data model. As and when innovations in product design happen, driving the need for new terminology, the industry’s trade organizations can lead the way in helping them maintain a standard.
Right now, PLM systems do not have machine learning capabilities built in, but many retailers and brands have begun partnering with outside technology providers that can integrate the machine learning that enhances the value of PLM. Retailers and brands need to consider whether their PLM system is enabling them to keep pace with the needs of today’s retail environment, and consider the outcome of letting the industry evolve without them.