【#CBK】SEQUEL：AIを使ってスナップ写真からアイテムレコメンドや需要を予測

The series [SEQUEL] tells you about the start-up that was introduced on the pilot boat. This time, we asked about “afterwards” of New Rope Co., Ltd., which develops a fashion engine “#CBK scnnr” (Kabuki Scanner), which raised funds in October 2019 and searches for similar items from photographs.

(Interview: October 2019)

Image analysis of fashion snaps with Kabuki Scanner. Recommend recommended items

Recognize items just by uploading photos

First of all, I would like to look back on the previous article about New Rope.

New Rope has several businesses. First is the fashion media “#CBK”. Realistic fashion snaps posted by Instagrammers and reader models are curated and published under the consent of the person.

(image: #CBK)

#CBK tags items for each post image. In the lower right of the above image, “Autumn”, “Beige Parker”, “Maxi Skirt Bordeaux Gather” etc. By simply carrying out this tagging, New Rope has obtained tens of thousands of snapshots and tag combinations. “#CBK scnnr (Kabuki Scanner)” combines these with AI as teacher data.

#CBK scnnr analyzes snapshots (preferably whole body) and instantly determines fashion items on the photos.

(image: New Rope)

For example, in the above image, when you upload your own photo to a system that includes #CBK scnnr, items such as “denim jacket”, “sneakers / high cut / white” are recognized.

In October 2019, Newrope announced that it would raise funds through a third-party allotment. Procurement amount is 100 million yen, and suppliers include VC, as well as a large mail-order business company called Dinos Cecile (hereinafter “Dinos”). It is going to work on "promoting industry issues".

To fashion-vertical SaaS

The purpose of the above financing is mainly product development. Many companies that deal with AI are dealing with fashion for the first time, but they are also expanding their domains into surrounding areas and creating general-purpose AI. In that respect, Newrope will focus on fashion and create vertical products.

In terms of general-purpose AI, I don't know when Google or Amazon will come in, and the fashion industry has unique challenges. So Newrope steered in the direction of creating vertical products. (Mr. Sakai, the same applies below)

In fact, for example in the beauty field​ ​Amazon​ ​Is EC page,​ ​YouTube​ ​Has implemented a virtual make-up function using AR in the video, and it should have some impact on startups that are developing this function. It is not strange that similar cases occur in apparel.

Therefore, Newrope will proceed with the vertical SaaS service. In developing #CBK scnnr, there are many requests from user companies that they can do this. Among them, more and more general functions are added, and #CBK scnnr is converted to SaaS.

For example, a recommendation function on EC. When users upload photos of their favorite Instagrammers on EC, they can search for products similar to the items used in the photos on EC. For example, if you upload a photo wearing a gray hoodie, an item similar to that item will be presented on EC.

(image: New Rope)

The use of #CBK scnnr is not limited to EC. Introduce to fashion media to lead to media commerce, personalize catalog based on purchase history, support customer service with digital signage, analyze visitor with surveillance camera and use it for marketing, AI also It is also possible to generate an original pattern. In the future, we will promote vertical SaaS by expanding the range of functions not only for sales but also for design.

Demand forecast by analyzing SNS photos

New Rope not only performs fashion analysis using #CBK scnnr, but also conducts demand forecasting, trend analysis and forecasting. That is “#CBK forecast”. Travel through various social media such as Instagram and analyze fashion trends. Quantify information and provide information to companies. As the information population, information on materials, features, patterns, etc. is collected through “SNS”, “(large) EC”, and “collection”.

Of course, even the apparel industry has conducted user interviews, analyzed trend books with collection information, and made future predictions. However, user interviews can be biased in nature and are not comprehensive. Trend books are too general and require further analysis for their own use.

On the other hand, #CBK forecast analyzes trends using data obtained from SNS, etc., so it is highly comprehensive. The analysis results can be customized, so it is possible to report in more detail according to the brand, such as “If this brand, this is popular”.

(image: New Rope)

For example, there is a company that manufactures and wholesales shoe care, an actual client of #CBK forecast. The company uses #CBK forecast to analyze the trend of shoes in the world. If the analysis that “the leather shoes are increasing in the world” comes out, it can be inferred that “the cream for leather shoes is necessary”. In this way, demand is predicted and utilized for manufacturing and sales to retail stores.

In #CBK forecas, you can also find out how the clothes you sell are matched with other brands. I thought that there were many stylings of “leather pants in leather pants”, but they were actually dressed with jeans. So, for example, try to increase the MD of jeans and sets.

User interviews and collection data tell the future. However, what is known from SNS information is “now”. You can see what items are distributed and how much, so in that sense you can get data that you couldn't get.

#CBK forecast analyzes not only your company's information but the whole society, so you can analyze items that you didn't make. Information that is not made in-house but is actually worn or not, such as “sale of red items but not so much, so let's discount early” It can also be used for demand forecasting.

Recognizing EC measures with many variables through AI

As I mentioned earlier, we are trying to create a new rope fashion vertical SaaS, and we are developing additional functions. One example of this is our efforts with Dinos, whose mainstay is the mail-order business, which is the source of investment for this fundraising.

(image: New Rope)

New Rope has implemented DM measures in the past with Dinos. A personalized DM was sent to a customer who purchased at Dinos, saying, “This item goes well with this styling. The results were excellent, and this connection also led to this funding.

There are two major initiatives in the future. First, use AI to promote personalization measures such as DM. Another is demand forecasting.

Although it has not yet been commercialized neatly, we are focusing on demand forecasting. Sales information is provided by the company, and product information is image-analyzed. I think the results of the analysis will be used in future MDs and orders.

There is no time to mention that if a photo of a product that cannot be sold at Apparel EC is re-taken after changing the model, sales will suddenly improve. However, there are too many variables such as model selection, photo composition, and explanatory text, so it is difficult for humans to guess the effect, but AI can quantify it.

Since #CBK is characterized by tagging all items in the system, you can grasp the effect quantitatively. Therefore, even if you do not perform A / B testing, if you know a certain amount of product information, you can estimate what kind of measures will have an impact.

However, a large amount of data is required for demand forecasting. In that respect, Dinos has received a fund this time, and has established a system where people can sit down and meet each other. Sakai says that a simple test has already been carried out at another apparel company, and “a positive number has come out, so I want to improve it.”

To services that save the entire industry, such as design and MD

Finally, we asked about the future development of the #CBK series.

I would like to be able to enjoy various values easily by specializing in the fashion field such as apparel, accessories, bags, shoes and hats. By simply putting a tag on EC, you can predict demand and make recommendations.

In addition to EC, customer support using digital signage at stores and marketing support by analyzing customers with surveillance cameras are also conceivable. If the analysis results are converted into data, it may be reflected in demand forecasts.

In addition, AI wants to support design, and to handle natural language processing such as "You can match these tops with these bottoms."

Of course I am happy to use the service as usual, and since it is still a developing service, I would like to ask you to "I want you to make this service" and "Would you like to work with you?"

(From left: CTO Arai, CEO Sakai, VPoE: Ishii, image: New Rope)

A new rope that tries to solve issues in the fashion industry using AI and other technologies. The industry has many problems, such as population decline, labor shortage, and inventory disposal, but there are things that can be solved using these technologies. If you want to find clues to tackle your own or industry issues, please contact us.

(eyecatch image: New rope)

The contents of the interview are also distributed by podcast

We are delivering interviews at the podcast interview.

Production team