Dressing Up with Data Science (Part 1)
The fashion industry is one of the most fast-changing industries, relying heavily on human creativity and the lasting effect of a trend. Put simply, this is a field where trends change almost every season and there is hardly any pattern that it follows.
“Since the life cycle of a fashion trend is tightening and condensing, the period of time between a trend’s appearance and disappearance is faster than ever. From introduction and adoption to peak and eventual rejection, a trend can be around for as little as a single season,” Grace Gordon, SavoirFlair’s Fashion and Beauty Director comments.
As fashion designers and companies compete vigorously to be at the forefront of fashion, there is no greater advantage than being able to respond swiftly to changes in preferences and predict the next season’s fashion trends. While we all know that big data is about analyzing past data to create useful insights and predict the future, it is interesting to note that various fashion companies and online fashion services have incorporated big data in predicting trends and consumers’ preferences to take their businesses to a higher notch. Global brands like Zara have utilized big data to strategize their seasonal offerings to create a sense of urgency, while H&M has employed artificial intelligence (AI) to relocate their inventories in alignment with consumer demand.
In fact, the impact of big data goes beyond just the market demand of fashion. In the fashion industry, there are generally five production stages: designing, prototyping, manufacturing, distributing, and marketing, in all of which we have seen trails of big data. This article will unveil the practice of data science in the first three production stages (designing, prototyping, and manufacturing), supporting each with case-studies from renowned fashion companies.
Stage 1: Designing
“So, who is the new designer? AI or Human?”
To begin with, the designing stage is the root of all fashion trends. It is where designers scratch their heads just to produce the one masterpiece after hundreds, or even thousands of editings. Designing is more than just sketching garments. More accurately, it encapsulates many components of the design process like researching, drawing, coloring, conceptualizing, pattern making, etc.
In recent years, famous designers and top fashion companies have started utilizing big data in creating product designs that align with the consumers’ preferences in terms of styles, patterns, and colors. With that, fashion companies are able to strategize their design planning to take place either ahead of the fashion trend or during the season.
[Ahead of the trend]
In this first case scenario, where we are predicting future fashion trends (“trend forecasting”), we require mass data collection and analysis on the past sales, as well as comparison of trends up to one to two years in advance. While most fashion retail companies have the financial capability and capacity to develop models to extract insights from their customer base, some of them also leverage external platforms to gauge market movement to quickly respond to the fluctuating fashion trends.
There are online big data tools and trend forecasting platforms that provide data insights ranging from the fashion industry at-large (such as fashion trends), competitive analysis and price fluctuations, to the nitty-gritty details such as color codes of a particular trend with the extensive use of big data analytics and machine learning. For example, EDITED is a data analytics tool that uses visual pattern recognition, optical character recognition, and text crawling to collect fashion product data and provide insights such as the product activity of certain design style and distribution of patterns. These platforms help businesses identify and develop strategies in terms of product designs and sales based on the key success factors of other companies, instead of relying on guesswork.
Apart from analyzing from a business perspective, trend forecasters typically identify trends from Fashion Weeks and social media using image recognition and hashtag detection. Images of fashion wear from runways and social media platforms are categorized and analyzed to project the popularity, as well as the spread of trends given a particular region at a fixed period of time. However, one shortcoming of image recognition technology is that it does not provide information about customers acceptance and perception towards a certain trend. It turns out that hashtags on social media platforms are able to make up for this as it can provide information about consumers sentiment, as well as the engagement between a fashion brand and its customers. As the forecasters closely monitor such data online, they are able to predict trends in the near future be it a comeback or new fashion, allowing fashion retailers to plan their design process ahead of time.
[During the season]
In the second case scenario, where garments are designed and produced during the season, fashion retailers usually conduct real-time analysis on product sales and consumer sentiment such that they are able to identify changes in consumer preference and purchasing behavior to refine product designs and alter production. Companies adopting such strategy usually produce relatively smaller batches of product for each given design with flexible and easily accessible distribution channels.
Get Winfrey Xian Ying Kong’s stories in your inbox
Join Medium for free to get updates from this writer.
For example, Zara has created a central management system that constantly analyzes and evaluates their customers’ fashion sensibility by consolidating all types of engagement with their customers. At offline stores, their personnel actively collect feedback from customers, ranging from the overall shopping satisfaction to the tiniest detail like color, zips, pattern and sizes of apparel. The feedback would then be reported to regional managers through the system on a daily basis, thus allowing the managers to further analyze customer preferences across the region. Zara operates in a way that they only send a small number of new products to the local stores bi-weekly with more than 350 in-house designers authorized to make product decisions. This allows Zara to continuously improve their product design, quickly adapt to changes and even predict their customers’ preferences in the near future.
Stage 2: Prototyping
“Can designers try out different colors and patterns just by a click?”
The prototyping stage is where the first sample will be made based on a set of designs. Once the first sample is produced, fashion designers will then revise the design elements like patterns and the type of fabrics. Traditionally, once a design has been finalized, the product development team would send it over to the manufacturing team, and the prototypes will be manufactured for each design.
With the advancement of technology, digital prototypes are introduced to the fashion industry. Artificial intelligence (AI) is used for visualization, enabling fashion designers to digitally comprehend their designs in 3D format with any digital device. By integrating with big data analytics and AI algorithms, designers are able to match and review designs with popular styles, as well as comparing to both the success and failure trends in the fashion industry. With that, designers are able to make more informed choices that aligns to the interests of their target audience, such as deciding whether microfiber should be chosen over cotton for sports wear, or whether vertical stripes pattern should be printed on dress to achieve a certain visual effect.
CLO is an online 3D fashion design platform that uses AI insights to help fashion companies immediately visualize a garment design in 3D based on a 2D design. At CLO, designers are able to perform real-time simulation for their designs, from there they can make instant changes and reviews, as well as unlimited exploration in terms of colors, patterns, and fabrics. A more interesting feature is its AI fashion models (so called Avatar), which designers get to select specifically based on their target audience — in terms of height, weight, body shape, skin color and even pose — and then test their digital prototypes onto the avatars with an exact 1:1 representation of the physical end product. CLO helps designers to make better and more well-informed choices too, by giving designers the access to data of various design materials through a cloud platform where designers and suppliers all over the world share information about fabrics in terms of colors and textures. It is also extremely convenient and time-saving as designers can directly “drag and drop” a variety of fabrics onto their creations in just a split second.
AI technology has effectively optimized the entire digital design process and brought efficiency to a whole new level. What’s greater than that is its contribution to environmental sustainability, as designers can now easily get a comprehensive overview of the final design deliverables digitally without having to produce a number of different “real” prototypes that potentially leads to wastage of resources, knowing the fact that the cycle of enhancing and prototyping could go over and over again, and especially inefficient when some companies have their product development team and manufacturing team based in different parts of the globe.
Stage 3: Manufacturing
“How can data science mitigate risks and reduce wastes?”
This is probably the one of the most exciting stages for fashion designers, seeing their creations turned into real-life products and mass-produced. While it may seem pretty straightforward to manufacture standardized goods repeatedly, there are actually lots of underlying factors and risks to be taken into account in the manufacturing process.
In fact, most fashion retailers have faced challenges in managing the manufacturing stage due to seasonal and demand fluctuations. Data science has come in handy to help suppliers and manufacturers to be more agile and be able to swiftly respond to changes, by allowing them to track new changes in demand, as well as unexpected risks in real time.
Natural Language Processing (NLP) is used both internally and externally to monitor the potential disruption in the supply chain. Within a manufacturing plant, algorithms are built in and trained to keep a check on machine readings, as well as to detect abnormal readings in terms of runtime, measurements and machine temperatures. For any error detected, production will be halted and the team will be notified to schedule a repair immediately. On a bigger picture, NLP is used in benchmarking business practices, calculating and reviewing the risk score of supply chain networks. With web scraping and social media listening, manufacturers are able to identify keywords and track information related to their partnering suppliers, raw materials provider, some external events, as well as the impact on their target market. For example, when deciding on acquiring a key supplier, companies are able to track the financial performance, ethics, and reputation of their potential collaborators to gauge the reliability and feasibility of having such collaboration. Other information such as local news, government regulations, market fluctuations, weathers, impact of natural disasters and diseases such as COVID-19 can be extracted and used to estimate business performance, as well as to measure the negative impact on supply chains. Given such valuable insights, manufacturing companies are able to further develop tools that serve as an alert indicator to supply shock, seek for alternative solutions and implement preventive measures at an early stage.
In terms of market sizing, big data helps companies to estimate the number of stocks to be produced within a certain region by matching garment design to the potential market size, based on past sales records, trends, and population patterns. Combining with real-time sales analysis, it helps companies to spot potential demand shocks, allowing them to plan ahead and better coordinate with manufacturers. As the manufacturing process is optimized and streamlined with the use of AI technology, companies are now able to effectively reduce the cost of production and achieve more flexibility in its production stage. For instance, H&M has adopted an AI solution that analyzes the consumer demand to the influencer trend, with the hope to improve their inventory management. Soon, the team found out that consumer demand generally rises when the influencer trend starts moving downwards. With a better understanding of the movement of market demand, H&M was able to devise a healthier rotation of garment production and distribution, thus enabling them to cut down the number of unsold goods and resolve the problem of overproduction after years.
It is no doubt that data science has played a crucial role in each stage whether it has led to greater efficiency or convenience. Data science is the foundation of making a data-driven decision and we have seen how it can transform the entire industry. In the next article, we will talk more about the interaction between fashion companies and their customers, how big data is able to nudge the behavior of customers, as well as a special case-study, so don’t miss out!
