Background: Chicisimo built the interfaces and infrastructure to help people find what outfit to wear, and what garment to buy. We focused on (i) creating input interfaces that allow people to express a need; on (ii) building the data system that understands and responds to that need, and; on (iii) creating an environment where users are incentivized to provide data, specially descriptors of outfits and garments, to keep the learning loop going. 

Process: We approached the above opportunity by building a small team of product people and engineers with a common understanding of the what-to-wear problem, seeking to build interfaces with the following characteristics:

1. The interfaces we build must allow people to express a what-to-wear need, from objective (type of garment, color, fabric, print…) and subjective (styles, occasions, age…) points of view; 
2. The development of interfaces and data infrastructure influence each other;
3. For 1 and 2 to happen, we have to uncover the organic data structures that arise from said interface;
4. While users provide input and respond to output, the entire experience encourages them to provide further input. Input comes in the form of (i) expressed taste, such as expressing what type of style they like, expressing what occasions are relevant to them in terms of clothing, what type of clothing they like or own. Input also comes in the form of (ii) categorizing content, such as grouping outfits into described albums, tagging outfits, establishing correlations among items, providing relevant queries, etc. This input supplies our team with feedback to build more effective interfaces and accumulates data so we can respond to more complex inputs with it;
5. The data structures are refined through unsupervised learning and collaborative filtering, by allowing us to find new data structures and ways to aggregate the data which aids in the consequent interface decisions. A learning loop is created.

As a result of the above, our product experience leads people to  create content for us, search for it and interact with it. It is this community who, by creating content, categorizing it, and searching for it, allows improving our understanding of the data and creating new algorithms and experiences for the community. 

Automatic curation. Our system encourages content creators differently based on the content they create: As not all users are as good as tagging or describing their content, we’ve defined an approach to easily classify users based on their quality content, and we give a different weight to the information provided by different users. Some users are featured in prominent places through our consumer apps, creating a positive feedback loop where they get rewarded for posting more and better content. We also monitor outfits with high engagement and add them to this list of outfits if they meet certain criteria.

Future iterations: There are several ways in which our process can be improved, and we have other sources of data we can use to improve or create new experiences:

• We are in the early days of allowing people to describe/categorize the clothes in their closets. We are in the early days because we have not needed to be at a later stage. This will change;
• We are not using the style of outfits people like, when returning results to their queries; 
• We are not using the closet data to help people shop for new products. From the point of view of our data structure, a closet is equal to a catalogue, so obtaining shoppable clothes that correlate with items in a user closet something we do easily; 
• We also know we can have the Graph attach further descriptors to certain outfits. We’ll expand the collaborative filtering by adding the descriptors users are querying for when opening an outfit or product. For example, if an outfit shows up in a query including “red pants”, and is constantly being opened, there is a high probability that the look is about “red pants” too. Albums names could provide additional descriptors to outfits, and it has not been a priority for us; 
• We are not using contextual data to uncover new possibilities, or to influence the output, like geolocation, time of the year, weather.

The post Our unsupervised learning model appeared first on Chicisimo.

A Taste Profile summarizes the taste of an individual shopper, what clothes she has in her closet, and what are the drivers behind her purchases.

OUR SERVICES

We help you build a Taste Profile for each of your shoppers, as part of your Taste Graph. Taste Profiles include:

• 1. We help you summarize taste. We register taste-related data of each shopper, and assign clean, structured and correlated descriptors to her profile. These descriptors are generated by our interpretation of their activity in your site. We use our ontology, graph and Smart Virtual Closet Technology.
• 2. We connect the Taste Profiles to our own interpretation of your products and to all the elements of your Taste Graph.
• 3. We help you build a Dashboard. Our Dashboard helps business teams to fully understand Taste Profiles and have a sense of control. With the Dashboard, business teams end up working hand-by-hand with tech, with the same understanding.
• 4. Taste KPIs. Whether you are capturing taste data or not, each shopper is telling you what motivates her purchases. KPIs tell you the % of shoppers for which you capture clean data, broken down into types of data.

SPOTIFY AND NETFLIX AS EXAMPLES

Spotify and Netflix also build Taste Profiles for each of their users.

• Spotify registers the music you listen to, and builds a Taste Profile with their interpretation of what each song means. As you listen to more music, your Taste Profile reflects pretty accurately what are the true motivations of you listening to a song. On top of this data, they automate several internal processes, detect early adopters of future trends, and offer personalized content.
• Netflix follows a similar approach. Netflix ontology plays a crucial role in their interpretation of what a movie means, and how that defines the person watching it. Your fashion ontology, although very different in structure, plays a similar role.
• In fashion, building the infrastructure to create Taste Profiles is way more complex that in music or movies. But once it is done, the richness and cleanness of the data is extremely eye-opening.

5 DRIVERS OF PURCHASES

Taste Profiles mostly contain data related to Products, Occasions, Influencers, Brands and Trends:

• Products. Products contain properties. One shopper might find relevant one or more properties of a product, or even none of the product properties.
• Occasions. A shopper will most likely have different purchase drivers for each occasion they dress for, special moment or everyday occasions.
• Influencers. Sometimes purchases are driven by an influencer. At the Fashion Taste API, we care about this data and register it.
• Brands. The brands someone has in her closet are not always easy to capture, but there are approaches to obtain this data depending on your needs. Brands are definitely an element that might predict what someone is going to buy in the future.
• Trends. The Fashion Taste API pays attention to the likelihood of someone buying a trend. We want to know if a given shopper is a heavy shopper of trends. The information provided by this cohort of shoppers is interesting in terms to revenue generation, but also in terms of data collection.

The post Taste Profiles of each shopper appeared first on Chicisimo.

To explain it easily, this system is like a brain that understands clothes and outfits, and allows you to organize and display products at your convenience, or your shopper’s convenience. It was constructed after analyzing millions of perfectly described outfits and fashion products uploaded to our system by different subsets of users of Chicisimo, and after analyzing how people interact with them.

The system converts fashion products into meta-products, which are abstractions of specific products of any catalog or closet. A fashion product is ephemeral, but its descriptors are not, so the system retains the value.

A meta-product is the most basic yet relevant description of a product, and one of the first tasks of our infrastructure is to convert any incoming fashion product into a meta-product. While a person might see a given garment, our system reads a set of descriptors, for example: burgundy + sweater + v-neck + comfy + casual + for school + size 42 + cashmere, etc.

For any given retailer, this system can automatically digest its catalogue and then, automatically: (i) Understand each product; (ii) Identify missing information; (iii) Identify similar products, defining similarity in a number of ways; (iv) Build complete looks mixing and matching the clothes in the collection; (v) Identify products that make sense to display together; (vi) Recreate any outfit with garments of the catalogue; (vii) display the correct products for each shopper, or for the current interest of each shopper; (viii) If the system detects a product that it cannot understand, it isolates the descriptor and incorporates it into the fashion ontology if the team so wishes.

The post Clothes classifier appeared first on Chicisimo.

How did we build our Fashion Taste Graph for Omnichannel Personalization

Through Chicisimo’s consumer apps, we’ve received millions of described outfits, the described clothes in millions of closets, and hundreds of millions of what-to-wear queries from people trying to decide what to wear.

With so much consumer data, we started capturing the correlations provided in the data, and built a fashion ontology to understand incoming input.

• Described outfits: We capture this type of data from millions of described outfits and we also capture how other people interact with those outfit creating more relations among items and allowing the Taste Graph to assign further descriptors to the outfits and the Taste Profiles;
• Described clothes in millions of closets: We also receive the described clothes in millions of closets from where we can extract a lot of meaning in terms of correlations and in terms of our own interpretation of the products;
• Queries: We also receive hundreds of millions of what-to-wear queries from people trying to decide what to wear and what clothes to buy, for any occasion you can think of. Each of these queries informs the Ontology and the Taste Graph.

This is the origin of our unsupervised learning model. The Taste Graph Technology receives data and correlations. It also assigns the correct descriptors to clothes, outfits and to people. It is the tool on top of which teams can effectively work in personalization. Delivered via a website, in a physical store, or delivered via Alexa in a bedroom or closet. This model allows for omnichannel personalization with a Fashion Taste Graph.

A Fashion Taste Graph is more than a Shoppers Graph

Our Fashion Taste Graph Technology for omnichannel is a personalization engine that allows us to assign descriptors to shoppers. It allows us to understand the taste of each shopper, and provide them with personalized omnichannel recommendations.

The Omnichannel Personalization with a Fashion Taste Graph is the infrastructure and data that contains, manages and understands the taste of each individual shopper, what clothes she has in her closet, and what are the drivers behind her purchases. The behaviour of each individual, the aggregated behaviour of everyone, and the behaviour of any subset we’d want to target.

Should you build a Fashion Taste Graph?

We can help you build your taste graph.

Your Taste Graph contains the intelligence generated by your shoppers and their interactions with your products. It’s formed by your shoppers Taste Profiles, your fashion ontology, and your evolving catalogue of products.

Your Taste Graph is trained to assign descriptors to your products and your shoppers. It is your unique Omnichannel Personalization Engine.

By building your Taste Graph, you are retaining that intelligence and building the competitive moat around your business that will help you delight your shoppers, with true omnichannel personalization.

What can you build with your Fashion Taste Graph?

Your Taste Graph is allows you do work around automation of catalogue, personalization of content, building unique shopper experiences:

• Deliver your collection based on the Taste Profiles of each shopper.
• Automate processes related to the management of your catalogue.
• Identify shoppers who can help you identify trends.
• Build WOWing experiences: Digital Closets with experiences they’ll love.

HOW DOES IT WORK

Our Taste Graph Technology executes a number of processes to register the taste of each individual shopper, what clothes she has in her closet, and what are the drivers behind her purchases.

Imagine that someone buys a “red and black oversized sweater“. In our opinion, this is not necessarily a predictor of her future purchases, not a descriptor of her taste. We focus on looking for the true motivations of a purchase. That’s the role of the Taste Graph and it’s underlying backbone: the ontology.

We patented in 2013, and we’ve perfected it since then. Your Taste Graph captures all the taste-related activity and data generated in your properties. It captures all the relations among your shoppers, your products and your descriptors.

A strategic asset that improves exponentially

The Taste Graph forms a strategic and tangible body of data that provides you with the intelligence generated in your properties.

The more digitally powered experiences you offer, the more taste data you will capture. The more data you capture, the better you can automatically serve your shoppers, and the more data you will capture. It’s the the virtuous cycle of data.

Your personalization engine

In terms of personalization, there is nothing more powerful that the data asset that understands the taste of each individual shopper, what clothes she has in her closet, and what are the drivers behind her purchases.

Any personalization attempt gets exponentially better, as you scale your Taste Graph. Any new sale, any new interaction, any new query or event, and your Taste Graph gets better, your personalization potential gets exponentially better.

Even the activity of your editorial stylists, at a smaller scale, contains an intelligence that can become an asset.

This is strategic. If you are here to stay in the next years, you need this dataset. Also, any sustainable personalization strategy will require the intelligence behind data and correlations that you are most likely letting go today.

The post Omnichannel Personalization with a Fashion Taste Graph appeared first on Chicisimo.

Why we built our fashion ontology?

Fashion lacks a standard to classify clothes or to refer to the variety of concepts that describe products, styles, and personal fashion preferences.

At some point, we found ourselves receiving a lot of data via our consumer apps. We were receiving millions of described outfits, the described clothes in millions of closets, and hundreds of millions of what-to-wear queries from people trying to decide what to wear

Looking at this data, we only saw unorganized data, so chaotic that it was impossible to manage or build on top of it.

Our users could not really express their needs precisely in a ways that we could understand them, we couldn’t even describe the content displayed in our apps, in a way that could be found by those in need of it.

The backbone of our Taste Graph

Today, our fashion ontology is the backbone of our Taste Graph technology, and it is core to our development process.

We divide our ontology into two parts:

• Products ontology. It is a 5-level ontology that describes products and subjective characteristics of products;
• Outfits ontology. It is a 2-level ontology that describes outfits, mostly with subjective descriptors.

Our fashion ontology today

Our fashion ontology is our tool to understand products and people: it goes well beyond assigning descriptors to clothing. It has a very strong emphasis on describing people.

Our ontology today is the classification of descriptors needed to define a fashion product and a shopper’s need, in terms that are relevant for shoppers, for retailers and for data scientists. It contains the vocabulary used by people and by fashion retailers, with an understanding of how these descriptors are used together. Our ontology covers all aspects related to fashion purchase drivers: product properties, occasions, brands or influencers, or trends: It is formed by +2500 unique concepts and a few hundreds of thousands of descriptors.

What we’ve built on top of the ontology?

All our Chicisimo products have been built on top of our Fashion Taste Graph, and the ontology is it’s backbone. This is what the ontology has allowed us to build:

• Smart Virtual Closet Technology
• Taste Graph
• Search tech
• Recommendations is all forms and personalization
• Clothes classifier
• The ontology has also provided a unique common language for the team to communicate, and work, more efficiently.

How can I fashion ontology help your business?

Well, that’s a question you’ll have to answer prior to building one. You might already have a fashion taxonomy, and you’ll expand it little by little as your product team demands more tech strength.

In general, your fashion ontology will help you describe products and shoppers in a way that allows you to:

• Understand products and shoppers;
• Manage products and shoppers.

Your ontology will help you manage your catalogue with new browsing methods, humanized categories adapted to different types of shoppers or situations. It will reflect how your organization wants to talk to shoppers, always retaining total control over your voice while understanding the voice of your shoppers.

It will guide your data-attaching mechanisms to attach the right descriptors to products and shoppers. Data-attaching mechanisms include your taste graph, your editors, your deep learning algorithms, and other sources of product and shopper descriptors.

Characteristics of our ontology

Again, it will depend on the state you are at. Your ontology or fashion taxonomy is a tool that contains the organized and structured descriptors that define your products and your shoppers, in the terms that are relevant for your organization and your shoppers.

• Your ontology includes clothing and non-clothing descriptors;
• The ontology includes descriptors used by your organization, and by your shoppers;
• Descriptors are classified in categories, types, properties and more;
• We capture the correlations among your descriptors as defined by your products and your shoppers.

How can you use your ontology?

The Fashion Taste API main focus has been about the search for structure, in the unstructured world of clothing classification.

Once we learnt about clothing classification, we learnt that purchase drivers go well beyond clothing taste, and we gave a structure to those as well. So, believe us, we love clean and structured data that our algorithms can work with.

Despite of the above (or because of the above), we have found amazing value in shoppers’ unstructured input. This input are descriptors that you would not expect your shoppers to be using when interacting with you.

We’ve learnt that people refer to their fashion needs in ways that are very different to how retailers do. Also, the type of taxonomies that fashion retailers tend to have cover mostly product descriptions. But drivers of purchases are related to other factors: products, occasions, influencers, brands and trends. You can read about it here.

Need help?

Contact us

The post Fashion ontology appeared first on Chicisimo.

OUR SERVICES

There are tons of providers of product recommendations out there, and if you want a traditional recommender, we suggest you get in touch with any of them.

Our interest goes well beyond that scope. We’d be happy to help you take control of your data via helping you create your own ontology, and then on top of that, we’d help you define algorithms for different use cases.

The approach above will make sense for those teams fighting for the long term, seeking substantial increases in conversion and not just quick wins – which you’ll rarely find.

If you like the approach, get in touch. Here’s what we’ve built in the past.

COMPLETE THE LOOK SUGGESTIONS

• Shopper is looking at a product:
  • A product she just bought, added to favorites, or a Product Detail Page;
  • We display other products in your catalogue that match well with that first product, personalized to the shopper;
  • We display clothes in the shopper closet that match well with what she’s about to buy.
• Shopper is looking at a garment in her personal closet:
  • We tell her how that garment matches with her other garments;
  • We tell the shopper what clothes in your catalogue match with that garment.

SIMILARITY SUGGESTIONS

• Shopper accesses your homepage or any of your categories:
  • We provide a list of products she’ll find most interesting.
  • Suggestions are grouped by similarity, “complete the look” or other criteria to be defined.

FACTORS WE CONSIDER

Our approach takes into account three factors:

• 1. The interpretation of your products by the fashion ontology.
• 2. The connections contained in the intelligence of your fashion taste graph – or ours;
• 3. And are filtered considering your shoppers taste profiles, if any.

The post Online discovery – Products & Outfits appeared first on Chicisimo.

The In-Store Outfit Recommender helps shoppers understand how a product they are about to buy, matches the clothes in their closet at home.

THE PROCESS

In the video above, you can see our colleague Maria looking at clothes at a physical store, and how the In-Store Outfit Recommender recommends what outfits she can create with the clothes in her closet and the garment she is about to buy.

The device reads the QR, extracts the images, and sends the selected image to our system where a Deep Learning algorithm extracts the descriptors of the garment. These descriptors are then sent to our Taste Graph, in charge of identifying how to best combine the new garment with the clothes in the shopper’s closet. (Note: We do not develop our own image recognition algorithms, it is not our focus. We use 3rd party algorithms). Read about our Smart Virtual Closet Technology.

Smart Fitting Rooms

Our Smart Fitting Room helps shoppers make purchase decisions. When they are about to purchase a product, the Smart Fitting Room tells them how they can combine that product, with the clothes they have in their closet at home. You might want to offer this service together with the In-Store Outfit Recommender.

The post Outfit Recommender for Physical-Stores appeared first on Chicisimo.

The In-Bedroom Fashion Stylist helps your shoppers decide how to wear their clothes. Right in their bedrooms!

HOW IT WORKS

In the video above, you can see our colleague Maria asking the In-Bedroom Fashion Stylist what to wear with her black jacket, and the In-Bedroom Fashion Stylist returning that same black jacket, together with other garments that she has in her closet. 

The In-Bedroom Fashion Stylist is powered by our Taste Graph and the Alexa Platform, understands the user taste, and knows what clothes the user has in her closet. It can be programmed to be connected to your app and your catalogue, and offer the shopper unique services with her closet and your catalogue.

THE 2006 IPHONE

Omnichannel in fashion retail is like the iPhone in 2006: we are about to see a game-changing consumer experience.

Omnichannel goes way further than physical stores and apps. It is going to reach bedrooms, and this will change everything.

Without any doubt, this experience will be built on top of a Taste Graph, and on top of shoppers Taste Profiles.

The post In-Bedroom Fashion Stylist appeared first on Chicisimo.

We have built and shipped to 5M users a smart virtual closet that allows women to easily digitize most of their clothes in minutes. We have learnt that with this approach, the shopper is absolutely retained.

Our Smart Virtual Closet Technology allows your shoppers to digitally store all their physical clothes, without any friction, in their own smart closet.

Our Smart Virtual Closet Technology increases your shopper engagement. With zero friction, it allows you to:

• 1. Offer outfit advice on how to wear her existing clothes;
• 2. Help shop for new clothes that match her existing needs;
• 3. Help her plan outfits.

Smart Virtual Closet Technology in action

These services can be provided via your traditional website or your mobile apps, or via new omnichannel retail experiences such as an In-Bedroom Fashion Stylist or an In-Store Outfit Recommender.

CURIOUS ABOUT OUR TECHNOLOGY?

Continue reading if you’d like to have some details on how the Smart Virtual Closet Technology works. Our Smart Virtual Closet Technology focuses on 3 components:

• Interfaces to capture input;
• Fashion ontology to interpret input;
• Taste Graph to capture and store the intelligence generated by shoppers’ closets, and to deliver personalized content.

ALL STEPS IN THE PROCESS INFORM EACH OTHER

We are fascinated by the efforts required to build the Smart Virtual Closet Technology. Of all the interconnected efforts involved, there are two aspects that we particularly love:

• Interfaces + Data infrastructure. How the interfaces to capture input are so strongly related to the data infrastructure interpreting that input. One cannot evolve without the other;
• Incentives to provide data + Requirements to provide value. How the incentives offered to shoppers to provide their personal data are strongly related to the consequences of having lots of closets created. People offer you their closet data because you give them something in return (normally, outfit ideas). But you can only provide ideas if you have a large dataset of closets.

ONBOARDING USERS TO A SMART VIRTUAL CLOSET

A relevant area of our Smart Virtual Closet Technology has been conceptualized and delivered to the market because of the need to onboard new users to a virtual closet, without friction.

When a new user opens an app (or any other interface) for the first time, she needs to be onboarded in a way that she sees value in seconds. Under no circumstances someone is going to upload her closet if the system is complicated. With that objective, we looked very carefully at the query data of our own consumer apps, and shipped lots of incremental iterations of our capturing interfaces and our data infrastructure.

The result is a very effective Smart Virtual Closet Technology interface and a very sophisticated data infrastructure to onboard users to a Digital Closet.

We define onboarding as the process by which new users find the value of the Virtual Closet as soon as possible, and before they abandon.

DIGITALLY-POWERED PHYSICAL INNOVATION

This Smart Closet Technology, together with our In-Bedroom Fashion Stylist and the In-Store Outfit Recommender, are the current consumer-facing expressions of our data solutions, our omnichannel retail solutions. All of them are built on top of our Taste Graph.

INTELLIGENCE TO DETECT TRENDS

Depending on your business needs and how you deploy your smart virtual closet, using the Taste Graph might help you to isolate early adopters of third party trends.

We built a similar system for songs, back in 2006, which was later acquired by Apple for Apple Music. It was an anomaly detection system that identified popular songs and traced them back to their early beginnings. There, it identified who were its first listeners. The behaviour of “first listeners” was used to inform of potentially future trends.

The post Smart Virtual Closet Technology appeared first on Chicisimo.

With 5M users, we’ve built a method to learn fast and iterate, and just on iOS we’ve shipped 204 public releases and 919 pre-production releases.

Read about our tech, taste graph and ontology here, and about how we reached our first 4M users, here.

In Q3 2019, Chicisimo opened its API to 3rd parties. You can read about our technology in the Fashion Taste API website.

Read about Alexa SEO, our blog, No Simp September or discover the Best Zapps – Apps for Zoom.

The post Real fashion on real people ® appeared first on Chicisimo.