Data Science is not just a fashionable word in IT, it is the name of the section of computer science about working with data, which today finds its application even in the fashion industry. One thing is clear: the future has already arrived. Moreover, fashion and technology make our lives easier and strive to elevate responsible consumption to a higher value. On the eve of the upcoming Fashion Tech Summit, the FWD editorial team talked to one of the best experts in the field of data analysis and data mining (in other words, Data Science) and asked him about why it is important for the fashion industry, how to use technology for fashion houses and brands, and whether fashion can exist without technology.
Why is Data Science important to the fashion industry?
Data Science and business analytics in general will and always has been important for any industry. The fashion industry is no exception.
Thanks to the wave of Big Data craze, data has become mainstream and popular. Thanks to artificial intelligence and machine learning applications, we are on the cusp of new discoveries. Nevertheless, this is a natural evolutionary process of what has always interested humanity: to understand a situation better and deeper in order to make faster decisions. Hammurabi ran his kingdom with simple analytics a couple thousand years ago, my father ran predictive models to optimize parts accounting in the 60s. Today, we have more data, more tools and algorithms at our disposal.
I would rephrase the problem and relate it to the issues we have in the fashion industry: in what areas is there a lack of knowledge, understanding or efficiency? In what areas can't the human brain alone comprehend things well enough to make an informed decision? In any of these areas, you can get much better results thanks to Data Science.
What are the new approaches to working with data and the specifics of using it in the fashion industry?
There are an infinite number of approaches and they depend on the task of each specific business and only they know the specifics of their needs. Wherever they feel they need a deeper or broader approach, there will be an application of data analytics.
There are areas where working with data is simply essential:
- Customer Overview - 360 to understand everything about customers;
- Modeling the buying process to predict cross-sell opportunities and raise the sale amount;
- Logistics optimization to ensure fast delivery of goods;
- Marketing analytics to increase sales;
- Social media analytics to influence social media.
How can high technology be used in the fashion industry?
High technology is finding new and interesting applications.
- Image analysis can allow automatic recognition of clothing items in order to improve marketing or to advise on the creation of images;
- Virtual reality or augmented reality can show you how you look in certain clothes;
- Artificial intelligence models can learn what colors or styles will suit you best;
- Advanced statistics can identify patterns, clusters and trends in your customer base to best serve them.
Again, there are thousands of ideas for your business questions.
Have you had experience working with fashion companies? How can a Data Science professional help them?
I advised a very interesting startup called Savitude, which uses AI to customize clothes to fit. Its goal is to make it easier for you to choose and to keep you from buying things that won't fit you well for your build. This technology allows you to first browse the entire range of products and then, using your nine key parameters and your lifestyle information, makes recommendations on what fits you best.
I've also been working on analyzing customer behavior: how they walk around the store, where they stop and what holds their attention. You need to know this in order to understand how to better organize merchandising.
From an analytics perspective, any business intelligence can be useful if backed up with action.
Neural networks are already able to paint pictures in the style of famous artists - it is enough to recall, for example, the "New Rembrandt" project realized with the support of Microsoft. When will neural networks be able to create independent works of art and not only?
Art starts with the soul, emotions, feelings: machines will never be able to do that. What they produce are products, not works of art in which the artist has put his whole soul.
Sure, machines can create beautiful things that are pleasing to the eye. But it will never be art. Can machines produce cool clothes? Absolutely. But I will always feel better in things that a human has put their soul into.
As part of Fashion Tech Summit, there will be a startup competition. And how do startups store data?
It depends on the volume, speed and type of data you store. Also, given that the best analytics are those that are useful to the business and can be applied, I would definitely choose solutions that involve flexible evolution and easy access for the business, because ultimately it is the business that will be the primary user of that data.
How can the average person apply Data Science knowledge in real life?
Applying algorithms and data for advanced visualization has become quite simple. After minimal training, we can all use logistic regressions, random forest or the k-means method.
Nothing super complicated there. We only need to consider a few aspects:
- Have we defined exactly what we are looking for? Without a clear question, statistical analysis is moot.
- Do we have access to relevant data? Data sets are not given readily available. They must be collected before any analysis.
- Do we have good quality data? We need to remember that incomplete data, unreliable information, use of irrelevant variables will ultimately lead to inaccurate results.
- Do we understand the essence of algorithms correctly? If we don't know the meaning of the "letters" in the results, we get confusing questions like "Is the value of P, Pr (> | t |) 0.278657 good or bad?" which further confuses us about the reliability of the model.
- Are we choosing the right model for prediction?
- How robust is the process so that we can perform this analysis on a regular basis without wasting time?
What is fashion without technology?
Fashion doesn't need technology, but there are complex issues that require technology. With more interactions, rapid globalization, and less tolerance for intuitive solutions, fashion, like any industry, needs to be equipped with auxiliary intelligence. Our brains are not enough to see, understand and process it all: this is an area where technology helps.
Fashion Tech Summit 2018 (31.08-2.09 2018) is a new format of educational event that aims to create a quality platform for interaction between fashion and tech industries. One of the best Data Science experts in the world, who teaches at Stanford University, will be a speaker at the summit. Gauthier will talk about new approaches to working with data and the peculiarities of their use in the fashion industry.
