Behind the tech: Retail Onboarding for home & decor

Care to share?

Technological changes in recent years have shown how important it is to be able to react quickly. It requires solutions to be flexible enough to allow new services to be implemented immediately. You don’t have time to build everything from scratch.

Some areas have highlighted the need for these rapid responses more than others. We can see this in natural language processing, the development of machine learning (in particular transfer learning), or recent changes in augmented reality technology.

This perfectly fits in the composable eCommerce approach, which is a natural consequence of the microservice-based architecture that has been popular for a long time.

See the full presentation about the Retail Onboarding concept.

Monolithic vs. composable approach 

Without going into detail, this approach is easiest to illustrate with the comparison of choosing a computer. 

The monolithic approach is represented here by the Mac. Its components were created in order to cooperate with each other. This allows for effective work in specific tasks, but any hardware changes generate high costs. 

As an alternative, we have a PC computer. It allows for relatively easy changes to the components. In our metaphor, it corresponds to the composable eCommerce architecture, and its main advantage is flexibility.

Key architectural assumptions

The solutions proposed below for the home and decor industry sector, intended for customer onboarding, are based on the above-mentioned approach. The end result is WebComponents. They allow you to become independent of the platform you’re using while still allowing for deeper integration.

Looking at the entry point, in line with the composable eCommerce idea, we place our solution in the product life cycle, connecting to existing systems, like PIM, and/or DAM. This allows us to analyze the product data on our own and use it in our process.

Connected services


An important element here is the service provided by It allows you to extract features from digital assets, such as color, brightness, pattern, material, or product category, with the help of pre-trained models. This service has an additional role here. We can analyze a photo provided by a user. It may be their inspiration, a Pinterest board with their favorite designs, a link with an interesting product, or a photo of the entire room.

Using this information, we can suggest products that best suit your customers. With the information collected from the user during the process, we can prioritize specific product features. The scoring base database used in our architecture allows us to rank the proposals according to what the user appreciates the most. 

An additional feature of this solution is that, with access to product photos, you can avoid creating the full classification of products with a list of tags and features.

Product discovery platform Syte uses visual AI to identify and tag features and visual attributes within your catalog images. This includes elements like color, pattern, style, product category, embellishments, and more with the help of pre-trained models.

This software as a service can then apply the same algorithm to images provided by shoppers through its Camera Search solution. For example, customers may upload an inspirational image from Pinterest or a photo of an entire room they liked. Syte is then able to analyze the image and suggest products within your inventory that are visually similar to the items pictured. With the help of merchandising rules and hyper-personalization technology, the retailer can then automatically rank the visual search results according to what a specific shopper is most likely to buy. 

Furthermore, the product tags that Syte’s visual AI assigns to each image within your inventory create a rich database for text search as well as a wealth of unique visual data that enables you to provide the most relevant product recommendations for each shopper. 

As a whole, Syte’s three suites of visual AI-powered solutions, Visual Discovery, Searchandising, and Hyper-Personalization, ensure that your shoppers are able to quickly and easily find products they want, regardless of how they prefer to navigate through your website. 

Azure Communication Services

An interesting component that allows you to cross another barrier between the customer and the product is Azure Communication Services, which allows you to make video calls. In this way, we can arrange for the customer to talk to the designer, or even connect directly with the seller in a local store, to see the product closely.


Threekit lets brands create customizable and configurable product visuals at a massive scale with interactive 3D, virtual photography, and augmented reality. Threekit works by marrying a brand’s product catalog with 3D artistry and technology. 

Product data, such as colors, materials, or options for each product, come from the brand’s eCommerce platform, their ERP, or their Configure-Price-Quote (CPQ) engine. The visual component is based on 3D or CAD models, materials, and stages with each product option in the product catalog mapping to a mesh on the 3D object. 

Discerning brands, like Lovesac, a furniture company, experienced a 15% increase in add-on purchases with Threekit’s 3D configurator. Crate & Barrel created three million hyper-realistic images in a month with Threekit’s Virtual Photographer at a fraction of the cost of traditional photography.

Optional Services


An additional noteworthy element is the ONEBOT technology developed by us. It allows for the sharing of content using Messenger, Telegram, or Slack channels, making it easy to cross another barrier of communication with the client. The models of natural language understanding used here can reduce the customer's abstract statement to precise product filters which, enriched with information from other channels, allow for even better matching of product proposals and remain on the same channel as the user.

Azure Personalizer

Azure Personalizer, in turn, is an element that can be used to further automate the onboarding process

User interactions create a behavior path. These paths are classified automatically. After selecting a group of products, the service learns to indicate the most appropriate of them. It does this to maximize the expected effect, which is making a purchase and/or some other defined action.

Leveraging composable eCommerce

Reading about the above set of services, one can notice that the idea of ​​composable eCommerce is inscribed not only as a place in the ecosystem for our solution but also in its interior. Each of the services mentioned above may be changed and further services may be added. At a time when technological development seems to only be accelerating, the possibility of dynamic adaptation is of particular importance in this context.

It’s just the beginning. If you’re up for eCommerce innovations and would like to see what retail onboarding can look like at your online store, simply contact us.

Innovations should always fit your KPIs. Let’s talk about how to customize your onboarding tool and tailor it to your business.

Published June 9, 2021