Mar 18, 2021

Every year, big brands have to deliver between 6 and 12 collections. Can you imagine the stress for fashion designers? Not to talk about the amount of waste produced just to create the collections.

Can technology come in the help of humans? Yes, of course! There are already different tools and graphic design programmes that are being used daily by fashion designers to create and modify collections, test and communicate ideas in a quick and sustainable way.

Time frame: 9 days

Team: 3 teammates

With two colleagues, I collaborated with the company Yoona Technology. Started out as a fashion label and transformed into a full-fledged tech company, their aim today is to build an AI-based B2B software solution to shorten design processes to just a few simple clicks.

We started out by creating our roadmap to be sure to have enough time for both the UX and UI part to be done by the deadline.

We started by reaching out the stakeholder to plan the kick-off interview. Then, before starting any quantitative survey, we needed to know more about this topic, which was quite new to the three of us. Documentation was crucial at the beginning and during the whole project to deeply understand the business, the market, our users and the latest trends about AI and machine learning in general — and, more specifically, in the fashion industry.

In the fashion industry the “problem” is that the design process is manual. This may sound crazy as we all have this romantic idea of a stylist drawing his pieces by hand. But we have to deal with reality and know that nowadays, because of the fast pace of fashion cycles, this would be impossible.

In most of the cases, especially in the fast fashion industry, the design process is not creative, but based on performance analysis, and existing designs of a company are combined with new trends or, as in this picture, simply minimally changed in cut and colour.

Design creation processes in the fashion industry are inefficient in terms of both cost and time. Due to the absence of holistic automation solution, fashion designers still have to proceed trend analyses, data collection, mood board creation and other processes manually.

Inefficiency of design process has led to major issues: inefficient resources’ allocation, not focused on the companies clients, subjective designing, overproduction and also important designers burn out. Designers need 1700 hours for one collection and only for colour concepts 400 hours. As a result of not accurate data and trend analysis, as well as incorrect managerial decisions, costs are immense, billions of never-to-be-sold clothing are produced together with tons of contaminating waste.

What is Yoona’s vision? Replacing this process with an Artificial Intelligence-based B2B software solution that shortens the value chain of the fashion industry to one click instead of 720 h.

Yoona’s solution automates the design process, makes it faster, more efficient and sustainable.

And what are the customers’ benefits? Customers can reduce their expenses on collection production up to 50% which could make up to 3 Million Euro a year. It can release designers from manual tasks and give the company the opportunity to operate resources efficiently.

In addition to efficiency, productivity increase and cost reduction, Yoona also supports the companies to work more sustainable.

Our challenge was to create a cloud-based web application for fashion designers of large-to-medium fashion companies who are looking to automate aspects of the design creation process. This tool will allow to:

generate design images with one simple click,

save the preferred designs (and change them in cut and colour, if needed)

create and share collections

How should it work? As a simple tool, that allow users to:

select or upload a print and/or a design;

generate new design through AI;

choose and save the best ones

And what’s the aim behind it? Becoming the world leader in automation of product design and let the fashion industry save time and money, and be more environment-friendly. This means that the 70% of the process is automatic while the remaining 30% has to be done by humans with their creativity.

On the second day of our project we had our first meeting with the stakeholder. It ran a bit differently than expected as we needed to know much more about the company, their strategies, goals, competitors and clients than they were able to share. We also needed to understand exactly what they expected from this project and which technical features they wanted us to develop. So we scheduled another meeting for the afternoon of the fourth day and this time we sent them our questions beforehand. In the meantime, we began the qualitative and competitive analysis.

As Yoona Tech is a B2B SaaS (Software as a service), we experienced some difficulties from the very beginning in finding the right interviewees. The people we needed to hear from were experienced fashion designers, so we sent parsimoniously our Google Form to only this specific audience.

The quantitative research showed us that:

clothes are mainly designed for all genders;

designers use a lot of different programmes not only to communicate (like e-mails, Slack, WhatsApp, Google Hangouts, Zoom) within the team, with their buyers and stakeholders — but also to provide them a visual representation of information (Adobe Acrobat, Figma, Instagram, WeTransfer);

the design cycle for a piece can vary from one day to two weeks

Afterwards, we ran six in-depth interviews with fashion designers, not only working for the heavy fashion industry but also small and independent ones, and university students. Our goal was to validate the data in the survey and better understand the role of fashion designers, their creative process, the tools they are using and what they would need in the near future.

Synthesizing all the ideas into a simple visual framework was crucial. We collected all the information gathered during the interviews under categories and had a clearer view of the problem.

Competitors’ Analysis

Starting with the tools named by our interviewers, we analysed the most used ones and found some other powerful fashion design software programmes. At the same time, we run an extensive feature analysis.

Competitors’ Analysis

On the fourth day of our project, the second Zoom meeting with the stakeholder took place: we had the chance to better understand their strategy (and where they want to place their product on the market) and the role of AI and machine learning in this tool. We also explained what we did until that point and decided that we could create our MVP in total freedom and based on our research and intuitions.

Market Positioning Chart

Inspired by our interviews, it was at this stage that we understood the potential of the product we were about to design: why just keep it for the big labels? Why do not expand the potential audience to independent fashion designers, small brands, fashion design students — not only from Berlin (or Germany) but from all over the world? As inclusivity is also a topic we are particularly concerned about, we understood that it would be a great idea to create a product which can be indistinctively used by anyone working (or learning) as a fashion designer regardless of their purchasing power. So we planned a shift in the target audience to allow anyone to collaborate and express the ideas visually and remote. It was clear to us that this decision would have implicated the need to find a solution for the high cost of the tool. Even if this is not our job, we thought about a few options, as offering different plans (S, M, L) or discounts to universities and designers’ collectives. This would ultimately give more visibility to the tool and the company.

And now meet our 3 personas: Emilia, working for a big brand, João, an independent Fashion Designer and Mia, a university student. The three of them experience the same pain points: they have an unbearable workload when a deadline approaches, are frequently not able to express ideas visually and wish a tool that helps to generate prototypes quickly and waste-free.

We started the third stage of the Design Thinking process by articulating our design problem and a clear-cut objective to work towards. The problem statement was followed by the How Might We (HMW) questions, the best way to open up to brainstorming.

At this stage defining the features that our tool must, should, could and won’t have was crucial and this method helped us to clear our minds and get unstuck.

Evolving from real people, the jobs stories are a powerful tool to look at the exact motivation of a user.

At this point, in spite of all the articles we read about AI and machine learning, there was still something not completely clear to us: how can a tool process the data that a human feeds it with? And which type of data should a person provide it exactly?

We had two additional talks with two external professionals, an AI Expert and a Data Analyst, to better understand the way data, AI and machine learning work. Thanks to their help, we understood that our intuitions were leading us in the right direction and that tool would be based on GANs (Generative Adversarial Networks). But what are GANs? The latest technology that can be used not only to generate images of any kind, but also to carry out text-to-image translation, to convert one type of image to another, and to enhance the resolution of images among other applications. And how do they work? To be able to give results, the AI needs to be fed with designs, sketchbooks, mood boards, and prints, which will be eventually transformed into new collections, individual designs, prints, materials or colour concepts.

At this point we were able to exactly define the characteristics of our tool: • a CAD tool (not only a web app) to aid in the fashion design creation, modification, analysis and sharing; • an inclusive tool not only for the heavy fashion industry but also for small and independent fashion designers, collectives and students; • a collaboration tool for the communication of ideas visually and remote; • an AI-based tool to generate, edit, and share design pieces and collections

The flow shows the creation of a new project, the upload of data (both for the design and the patterns), the generation of a specific article of clothing (a dress in our case — as we learned that fashion designers working for big brands are responsible for some specific items) through AI and the selection of a preferred piece among those produced by the AI. This is followed by some modifications, the saving and sharing of it with a team member.

A lot of sketches are being needed to come up with the latest ideas.

Before moving on to the prototyping of the mid-fi, we rejected the idea of conducting card sort — also because we were running out of time. The extensive competitive research run before revealed itself to be very useful once again. The categories in fashion are default and, in order to create a straightforward tool that people can quickly learn to use, we took inspiration from other tools.

3 options for our moodboard — the first one was our interviewees’ choice

We asked our interviewers to test our wire-frames and realized some changes needed to be brought: • add the possibility to select the prototype to be edited; • 3d human avatar is a must-have; • some steps were missing in the loading; • a second dashboard to work with the 3d human avatar — as all the main fashion 3d programs have

3 options for our moodboard — the first one was our interviewees’ choice

Google Forms results

At this point we focused on the design. As you can see in the results of the Google form, our 51 interviewees judged the first moodboard as the most techie, modern, futuristic and confident one.

Fonts, buttons, icons and images had been also implemented and the style tile was ready.

And now, after all this writing, you can test the MVP directly on Figma. Please follow the user flow described beforehand ;)

What have I learned during this project? A bunch of things, in the order:

research must be done during the entire process;

it’s important to release solutions quickly to gather continuous feedback;

if the competitive analysis is run thoroughly, the product is going to be intuitive;

the importance of creating a valuable solution for a potential employer or client;

the importance of learning more about a such important topic as AI, which is probably going to reshape our lives in the near future;

collaboration isn’t just good for a successful product, it’s also the best way to learn.

What is coming next:

presentation of our project to the company;

integrations with other software programmes;

view of the designs from the desktop app live across all the iOS devices;

gender-inclusive mannequin;

testing, testing, testing!