Wouldn’t it be great to have a crystal ball to tell you what to design and how much to manufacture every season? Wouldn’t it be fantastic to already know what product categories are going to be trending next season? How about being able run a program to help you predict profitability patterns by revenue stream? We’re not there yet, but there is some fascinating new technology which may transform the way we gather, use, and analyze data.
Data analysis has been around for a long time. Gathering data from various sources and systems to slice, dice, and analyze predates the 20th century, if you can believe. Henry Ford may be the first well-documented proponent of analytics when he measured the speed of his assembly lines. The use of computers to analyze data goes back to the 1960’s, when they were utilized as decision support tools.
Today, we have a number of different types of tools currently, or soon to be, available to aid in forecasting and decision making. What are they, and how do they work? All data analysis starts with data models. What is the data you want to gather and analyze? Data can be pulled from various sources and is generally contained in some sort of data warehouse or data cube. As an example, a manufacturer may want to pull product information from multiple systems such as PLM, ERP, retail POS, ecommerce, and/or forecasting tools. Data mining delivers vast quantities of data, often unstructured. If your data isn’t clean, clearly defined, or attributed correctly, it can make it next to impossible to utilize it, however. The goal is to amass data in order to inform an end result, search for patterns, and find relationships within the data. Data models are built to narrow down the data required for analysis. Business analysts make assumptions and query the data to locate potential answers. Utilizing business intelligence tools, the data can be compiled and presented in a readable and understandable format.
Data mining analytic tools aggregate data in various models and review historical data to help determine patterns e.g. “How did style 12345 in red for Holiday 2018 sell across wholesale, retail, and ecommerce channels?”
Predictive analytic tools review historical data to make assumptions and predict forward looking “what if” trends and patterns e.g. “Based on how style 12345 in red sold for Holiday 2018, should I increase or decrease production of red sweaters for Holiday 2019?”
Machine learning (artificial intelligence) goes beyond predictive analytics to review historical data, automate data models, analyze patterns, and make decisions e.g. “You should increase production of red sweaters for Holiday 2019.” Taking that one step beyond, machine learning gets smarter as we adjust our inputs, outputs and decisions. It makes assumptions, reassesses data models, and reevaluates the data, all without the intervention of a human, which is a game-changer. Just as machine learning means that a programmer or business analyst does not need to anticipate every possible action/reaction, machine learning is able to test and retest data to predict every possible result, at speeds no person could achieve.
All of these use statistics and data models as a base for analysis. Machine learning takes it a step further by analyzing data and trends and develop answers to questions you may not have thought of yet e.g. “Sweaters sales are trending in the Fall season and slowing down for Holiday.” Retailers are already using machine learning to analyze individual purchases, responses to promotions and buying history to make predictive suggestions to consumers to personalize shopping experiences, develop marketing strategies, and optimize inventory positions.
As our industry moves toward optimizing the supply chain, minimizing lead times, and creating more efficiency in product development, design, and forecasting, and shifting away from waste it’s easy to see the benefits of predictive analytics. The need to track and manage large numbers of products across various categories, track consumers’ buying habits, and maintain an exciting brand that keeps consumers coming back is reason enough to invest in future technology. Predictive analytics and machine learning may also be used for demand planning, forecasting, and streamlining the supply chain and help make sourcing decisions easier.
Today, these types of predictive analytics work best in stable and basic product categories, where the same or similar products are produced season after season. At present, it’s almost impossible to predict the trends and whims of fast fashion, high fashion, and streetwear. Not to say it can’t be done in the future, but we will need to have the technology and manufacturing capability along with speed to market to get those products to the consumer before they fall out of favor. The future may not be that far off, it’s time to start thinking about what predictive analytics can do for your business.
Business Process and Technology Consultant