AI and Fashion Forecasting
Among all the industries destined to be disrupted by AI, fashion forecasting is one of the less obvious ones people would think of. The main reason is that the current way forecasting firms conduct these analysis of future fashion trends is a mix of data-driven methodologies and also subjective interpretations by their staff frequently visiting bars, shops, and catwalks around the world. A specialized forecaster would spend time walking around cities like Milan, visiting local shops and bars that would otherwise have no information on the internet, gather intel on the latest trends during these site visits, and summarize their findings. On the other hand, a lot of data is used, including macroeconomic trends such as political sentiment and economic indices.
An early use of AI in fashion industry is sales demand forecasting. Since retailers’ sales data is now mostly digitized, past demand and sales history could be leveraged to forecast short-term demand, which is used by EDITED, a retail analytics firm (1). The accuracy of such forecast is yet to be determined, but the additional insight provided to retails in inventory planning could be significant from a supply chain management perspective. The benefits of more and more accurate forecasting in sales demand would definitely help with improving inventory turnover, better managing ordering cycles by taking into consideration product ordering lead time and popularity. The side effect of an over-reliance on AI in this type of trend/demand forecasting is discrimination against more up-and-coming designers yet to prove themselves in the market. Historical data would favor incumbents in forecasting future demand, while relatively young, small players would miss out on potential sales due to lack of market proven demand. Since the level of unpredictability is very high in fashion, and new brands usually become popular in unexpected ways (social media, online influencers, advertisements, trunk shows, etc.), the heavy use of AI in demand forecasting and therefore inventory management for retailers could be a boon to large incumbents but a curse for young players hoping to gain a solid foothold in the industry.
Large tech firms have now entered the fashion trend analytics arena. Google has set up a “Trendspotting” division specialized in summarizing and publishing annual fashion trends using its large database of search data. So far, the findings through this department haven’t been “groundbreaking”. Basic, sometimes obvious trends are being uncovered through their analytics division. Maybe this is a sign that there is still a place for more traditional fashion forecasting firms that send human agents to brick-and-mortar stores trying to scope out the latest trends and tell stories of future fashion trends.
Anyone working in the fashion industry would probably agree that the rules of the trade are less data-centric than they are human-centric. Past data can serve as a basic guide to what the industry would favor tomorrow, but the products & trends making waves now and tomorrow are sometimes dictated by carefully orchestrated, human-driven decisions made by a select few of people at the top of the fashion food chain. It would be very exciting to incorporate the use of AI in the industry, but there would still be a need for human “expert voices” that could influence the latest trends by some simple Tweets, Instagram posts, or comments on fashion magazines.
References:
- https://www.economist.com/news/business/21725599-technology-may-be-disrupting-peculiar-business-can-data-predict-fashion-trends
Users who have LIKED this post:
5 comments on “AI and Fashion Forecasting”
Comments are closed.
Nice Post Andrea!
I find your comments on the use of predictive AI in the area of fashion to be really insightful. I would agree with you that fashion is human-centric in nature, hence it is arguable how accurate can models trained based on historical data predict fashion trends. Unlink a simple recommender system, predicting fashion trends needs to take many factors (or features in machine learning terms), can determining such factors before training the model can be hard. Afterall, predictive AI is not a magic, and whether or not it will work depends on the specific field it is being implemented.
Zonglin Li
Student from MS&E 238A
Very interesting! I am excited to see how the predictive AI will have influence on the fashion designers when reading an annual fashion trend for their shows. I guess either super disruptive or no influence at all since fashion is human-centric, as you mentioned; fashion is more like art than business. So yes, there may not be any impact of using AI in fashion industry… I agree with you that there will be a need of human expert even if the use of AI becomes bigger in fashion industry.
I believe AI is more useful in some area but less helpful in predicting future. Since future is so complicated. AI could help people designing, and weather forecasting. While fashion history is kind of repeat and repeat. If we can build such successful model to predicate fashion trend, I think at that time we have found the rule of fashion trend.
Andrea,
Have you seen Amazon’s new device called Echo Look? Essentially, it’s a personal AI fashion assistance. This device is a great combination of natural language processing, computer vision, machine learning, and big data to provide personalized style advice. It also helps you discover new brands and I’m sure Amazon will use this data to target ads to you. It’s not out yet, but it’s definitely an interesting application for AI.
Great Post Andrea!
I totally agree that fashion trends do not depend on data but more of the similarities of the top designers’ design and how much the market accepts them. Fashion forecasting is crucial for a fashion-related company. Although trends cannot be predicted by the technology, utilizing AI technology can help a company being a good player in the fashion industry. The San Francisco based company Everyone did a really good job. It focus more on the basic style and combines it with the silhouette details that what is mostly trendy and accepted by the market right now(which can be concluded from the data gathered).Marketing is more human centric, as long as the brand’s image and philosophy can be distinguished by the costumers, even though the styles are not unique, the brand can standout from the crowd.