Insiders’ concerns + existing solutions + data science
In the past 5 years, everyone was trying to use Artificial Intelligence or Machine Learning to make quantitative prediction, from stock price down to which ads a Facebook user will be interested. However, as for fashion, something so personal, cultural and abstract, is it predictable?
What is Fashion Trend Prediction?
First, let’s define the above term.
Fashion prediction forecasts the upcoming popularity of a certain attribute, such as color, fabrics, textures, prints, materials, etc.
In reality, fashion trend prediction is looking at a far, far, far… away future. For a fashion brand like Gucci, she would predict/lead the next season fashion, thus design and manufacture those clothes in advance. Due to lead time (manufacturing cycle), 3 months prediction is the least duration needed. Often in times, prediction can be up to 6–12 months. So, imagine someone predicting what you are wearing today before COVID-19. Such lengthy predictions gave me a headache whenever my boss asked me to his room.
Second, it can be abstract or specific. For instance, designers in Gucci said the trend is going to be 70s style in 2021. 70s lead to a bunch of fashion attributes trending in e-commerce.
As for specific prediction, pleated skirt (a manufacturing technique) by Issey Miyake got incredibly popular in 2018 because a new manufacturing technique was invented.
70s is very abstract while pleat is a highly specific texture.
Third, there are different audiences. Designer, Buyer and Merchandizer all play various roles in the fashion games and require different range and depth of prediction.
Designers pay attention to abstract ideas in the furthest future.
Whereas buyers and merchandisers keep an eye on nearer future. They decide what items suit the store the best and drive the most sales for the company. It involves regional and business consideration. Muji is one of the bestsellers in Asia, but imagine it selling the same clothes in LA. Probably some new tactics are wanted. (Beyond The Pandemic: Why Muji Failed To Survive In The U.S., by Forbes dated on June 2020)
So in real life, fashion trend prediction is foretelling an abstract but specific concept that meets short-mid-long term needs. Suddenly sound like a granny looking at your Tarot card and reading your fortune right?
Maybe I will change my career to fortuneteller one day. But before I do, let’s keep on solving this complex yet interesting question.
What was the industry doing in the past?
Look at the top of the fashion pyramid — fashion shows and high-end brands. Gucci, Louis Vuitton, Versace not only predict fashion but also lead the fashion. Fashion magazines Vogue, Business of Fashion, Women Wear Daily keep an close eye on top brands and comment on fashion trend.
Most of the brands follow the fashion trend by looking at high-end brands’ designers. So luxurious brands are always one step ahead. However, in the era of individualism, when everyone creates their own style, companies start to exploit data and become a game changer.
Combining the power of data and design experts’ experience, WGSN is the biggest fashion trend forecaster in the industry. Every designer has been looking at their fashion trend reports since student life. Numerous reports were introduced every month in every aspect to make sure you follow the trend.
Zara is one perfect business example. Apart from its full RFID-monitored supply chain, Zara extracted trending fashion attributes ( silhouette, style, etc) from fashion shows at crazy speed and pushed the details into their designs. (Zara & Big Data: A 5-Minute Case Study)
Data was gradually attracting attention in the past few years but compared to other industries, it is just in the beginning steps and there are many possibilities to be explored.
How can we take advantage of data in this game?
With e-commerce data, we can do the followings:
For designers, there are various types of print and pattern ( flower, abstract, animal, paisley, etc). Which one should you choose or what is the proportion of SKU should you take on those prints?
Floral print is the dominant print. Among all types of floral print, all over floral print occupied more than 49% of all floral prints.
For short-mid term forecast, we can use new-stock items, discount products and product YoY to conduct data analysis.
Percentage changes give buyers and merchandisers a sense about what product they should include in their stores, both online and offline.
These are the some solid conclusions we can acquire from e-commerce data.
Is fashion predictable? With data science , there are many means for us to foresee the future in a quantifiable way.
1) Demand forecasting on sales of SKU with time-series models
2) Natural Language Processing on magazine articles to summarize editors’ comment
3) Computer vision on Fashion Shows and extract the latest fashion attributes (Tagwalk, Visenze)
Let’s discuss the above topics at other times. Leave some comments if you have any thoughts!
About the Author
Gary Leung is a Data Scientist in Fashion Industry.
DeArt is a platform targeting Sustainability and E-commerce Trend. We believe fashion, AI and Big Data can drive sustainable fashion and keep an eye on the latest fashion trend. Click to know more about out platform.
Follow our Instagram to know more about sustainable & fashion news.