Professional Documents
Culture Documents
About Choosy
• CEO and Co-Founder Jessie Zeng, CTO Sharon Qian and Chief
Operating Officer- Mo Zhou
• A technology start-up that uses machine learning based algorithms to
identify fashion styles trending on Instagram
• Value proposition: to offer unique, eye catching designs inspired by
those worn by celebrities which retailed at much lower prices as
compared to the original pieces ( $100s or $1000s priced between
$50 and $100)
Fast fashion
• Fast Fashion:
• traditional retail apparel cycle of 4-6 times per year
• in 2000s H&M, Zara and Forever 21 brought it down to once a week.
• After debut on runways, fast fashion retailers designed similar, sometimes
identical, designs to be sold much cheaper
• Production and shipping cycles of less than 6 weeks
• China a major manufacturing hub from 2000-2015
Shop the Creative’s original model
• Developed out of a personal blog by Zeng
• Took style inspirations from streets of New York or Instagram
• Ordered similar samples from a wholesaler in China through an app
• Modeled for the same.
• Based on feedback, list designs for sale on her website and submit
orders for production
• Brand building:
• Partnered with 3 influencers each with 500K to 1 million Insta followers for
photo shoot and promotional video
• Subway marketing campaign
• Advertisement in trains and stations
How was AI used at Choosy?
• Machine Learning: Supervised : Image Classification
• Recognise trending fashion
• Identify patterns
• Divide the influencers into categories
• 8L design
• Image classification models to identify between celebrity face or clothing
• CNN – deep learning
• Sentiment analysis: NLP to analyze positive and negative comments
2. Analytics and Machine Learning at Choosy
• Search and Classify relevant images:
• Image classification algorithm using 7 million parameters
• CNNs to recognize objects in images
• Trained to detect clothing : Training data – 800000 fashion images annotated with 1000
descriptive attributes
• CNN would review all photos posted by a curated group of high-impact Insta users over
a period of time and extract that had relevant clothing images
• Classified the influencers
• Each photo was classified .. Given a score between 0-100 based on number of likes,
sentiment analysis using NLP
• Use data about top selling product… get images with styles similar to top
selling
• Also expand to accessories and home decor
3. Rapid Manufacturing and Supply chain
• In 2018, Choosy launched with 10 styles per week in standard and plus sizes
• Each day, team reviews top insta photos of previous week, to decide which to create
• Design is finalized
• Development estimates the product’s measurement to create a computer rendering
• Product team submits rendering to Choosy’s team in China, which creates pattern for
each design, purchases the right fabric and sews 2 samples
• One sample sent back to US for photo shoots and other to manufacturer
• 20 manufacturer
• Pictures of Models photo shoot posted on Instagram and Website as biweekly
“drops”
3. Rapid Manufacturing and Supply chain (contd…)
• Inventory:
• No excess inventory
• First run made to order
• If style sold well, more were manufactured
• Production cycle of 7-10 days, total 2 weeks
• Third part logistics facility
• Return rate 15%
Challenges
• Legal issues
• Control by influencers
• Website design
• Finding the right product mix
• To drop ship?
• Plus sizes