BYSON - Bring Your Store Online

Platform that allows shop owners to bring their store online super-easily and connect with customers

Comment 2

BYSON - Bring Your Store ONline

Local shop owners are being hit hard by the current crisis. While setting up a webshop offers them an opportunity to keep selling supplying their neighborhood, many of these stores struggle to shift their business online. Our goal is to make it as easy as possible for shop owners to bring their store online and start offering their products to customers in the neighbourhood. We want to connect them to local customers by offering a customised online experience that captures some of the flair of these unique stores and handle ordering, payment and facilitate the delivery of their products. By focusing on small stores and local business we want to make the experience of starting and running an online business as painless as possible.

The problem that our project solves

Many stores are trying to shift their businesses online because the Coronavirus crisis forced them to stay closed or operate at reduced hours. While setting up a webshop offers them an opportunity to keep selling in their neighborhood, many of these stores struggle to shift their businesses online. The main issues local shop owners face are:

  1. It takes to a lot of time to enter a product catalog
  2. It is expensive to recreate the look and feel of your store online
  3. It is hard to attract customers to your online store

The solution we bring to the table

We provide an easy-to-use AI powered platform for local stores

Our goal is to leverage state-of-the-art artificial intelligence (AI) systems to help shop-owners stay close to their customers. Our main focus is to address the difficulty of bringing a personalized store online and attracting customers to it using AI powered solutions, instead of just developing yet another webshop platform. We are addressing the issues listed above by:

  1. Easy onboarding: Minimizing the effort required catalog the products using smart import tools
  2. Automatic personalization: Automatically create a personalized online appearance based on pictures of the store
  3. Attracting customers: Making shops easy to be found and helping store owner optimize their product offering

1. Easy onboarding

To make it as easy as possible for store owners to build an online catalog of their products, we offer two smart tools to help them with this task -- avoiding the need to manually enter all the product details for every product. The first tool allows store owners to take a picture of one or multiple products and detects products in the image. Based on the image and optionally a detected barcode, the tool provides suggestions for the product image, product type, product category and more. The second tool allows store owners to upload any Excel files they currently use to manage their local stores and automatically extracts product information from these.

2. Automatic personalization

To convey the look and feel of a store, we automatically create an online template based on images of the store. We automatically determine a matching color palette to reflect the shop’s visual identity. To allow further personalization to the shop-owner, we also suggest spatial configurations of different components that can be used to create a webshop layout.

3. Attracting customers

To attract customers to the online versions of local stores, we offer two main features. First, we allow stores to be easily discovered through a unified search tool that finds stores in your neighborhood selling a certain product. Second, we offer sophisticated analytics tools that provide useful insights to shop owners as well as concrete suggestions to optimize product presentations, pricing strategies or to strategically expand their product offering. These analytics can be based on data from a single store or, on an opt-in basis, across stores on our platform.

Next to these main feature categories, we offer various smart services to improve the experience for both the store owners as their customers such as: Intelligent scheduling of pick-up times for customers to avoid long queues and allow customers to pick up orders at multiple stores in one trip; and automatically improving product images.

What we have done during the weekend

We started this idea from scratch and used this hackathon to seriously kickstart our project. Concretely we:

  1. Created an extended slide deck on Google Slides: link
  2. Drafted a business plan and platform architecture on Mural: link
  3. Wrote an extensive white paper to detail how to implement the various technical components: link
  4. Developed a mockup of many key interfaces on Google Slides: link
  5. Crafted an extended story-board user interface for store owners: link
  6. Set up a Django project to mockup parts of our user interface and developed started to develop proof of concepts for some technical components
  7. Created a demo pitch deck on Google Slides: link
  8. Pitched our project to a lot of mentors (thank you all for your feedback!)
  9. Recorded a 2 minute video using of our demo pitch deck

All in all it was pretty intense, a lot of work and a lot of fun, and amazing opportunities to learn and meet people.

Byson's impact to the crisis

Byson has the potential to have the following impacts:

  1. Help local stores shift to online webshops easily and operate them successfully, allowing them to reduce their financial burden and staying connected to their customers. This way they can keep their heads above water in times of lockdowns and careful reopenings and don’t see all their business disappear to large online retailers. On a larger scale, we hope such shifts can help in the fight against an oncoming recession.
  2. Spread sales over a larger number of smaller stores, while avoiding overcrowding of single stores through scheduled pick-ups; playing its part to flatten the curve, while avoiding unnecessary trips by combining pick-ups across stores.
  3. Increasing product availability to buyers by connecting them to local stores

The necessities in order to continue the project

  1. Outreach to local shop-owners
  2. Front-end UI / UX design
  3. Legal advice
  4. Extensive market research and interviews with local shop-owners
  5. Contacts to trade unions and local entrepreneurial associations
  6. Data storage and compute infrastructure to cope with large scale sales data
  7. Why not partner with Google Maps? <3

The value of Byson after the crisis

Our main strengths, AI powered solutions for easily creating and successfully running a webshop, are uniquely important during the crisis, but retain their value after the crisis. Helping local stores capture a larger part of the online retail market will hopefully strengthen their position in the long term and help them compete with big online retailers. Additionally, a shift to local stores and pick-ups in walking distance or local deliveries will help in tackling the next crisis that is already brewing: climate change.

Technical details

We broke down our workflow and identified various components, which can be enhanced using AI, and developed independently of other components. They are divided in three categories: importing products, personalising and improving layout, and connecting to customers.

We briefly describe all the components now:

Product Finder

This component links new products to the web in order to quickly retrieve suitable information. It is capable of:

  1. Reading multiple barcodes in any orientation from a single image
  2. Reading the name of the product from its label
  3. Finding product information online by using a combination of search engines and Natural Language Processing
  4. Detecting products by images by either using Computer Vision (a pre-trained neural network) or by segmenting out the product from the image and relying upon third party reverse image search engines
Information Extractor

In the case when the Product Finder fails to find structured information of the product available, this component would extract relevant information from the relevant web pages by using free text and image search.

Product Matcher

In the case when the shop-owner uploads all product descriptions using text or spreadsheet, he/she would still want to upload images of all the products in the catalog. If this is done naively, the shop-owner would have to first click on each product and then take a picture. Product Matcher allows the shop-owner to upload all the images at once. It then tries to automatically match a product or product type with an image. It might still not be perfect but it would still save much hassle for the shop-owner.

Color Detector

Based on the images of the shop’s interior, facade or/and logo, this component identifies the dominant colors in the images and generates a color palette reflecting the shop’s visual identity. Using a machine learning model, the color palettes can be ranked before being presented to the user, who can later fine tune these. Finally, the colors are assigned to the different color scheme roles like background color, text color, accent colors.

Layout Generator

It suggests spatial configurations of different components that can be used to create a webshop layout.

Category Generator

Hierarchical product categories are used by the Layout Generator for suggesting layouts. This component generates a hierarchical structure based on the specific set of products and their types. The idea here is to automatically keep a balance of the number of products in a category. A category will automatically get broken down into multiple categories if a lot of new products are added by the shop-owner. Similarly, 2 related categories might be merged if there are not a lot of products in those.

Image Improver

This component improves the image quality of the products by:

  1. Either, automatically optimizing the brightness, contrast, saturation or white balance of an existing image
  2. Or, suggesting a better (higher rated) image from our existing global product database
Pickup Scheduler

This component is responsible for suggesting time slots when customers can pick-up their orders from the different stores. It has multiple objectives:

  1. To avoid long queues and overcrowding in the stores
  2. To consider and uphold the opening hours and pick-up time preferences of the customer and shop-owners
  3. To combine multiple pick ups from different stores
Smart suggester

We provide a sales dashboard to the shop owners where this component could provide sales suggestions. This would involve the following kinds of recommendations:

  1. Changing prices of existing products or offering discounts
  2. Adding more product types that customers may buy with existing products
  3. Uploading better pictures of products

Architecture and tools used

Front-end

We plan to roll out a mobile app and a web interface powered by React and React Native. The app will be available for both Android and iOS platforms.

Back-end

The back-end of the web platform runs on Django in Python and can easily interface with our custom implementations used to power the AI components of Byson. These implementations will also use existing techniques and implementations from a variety of AI domains:

  1. Computer Vision (Image Segmentation, Object Recognition, Color Detection, Character Recognition)
  2. Natural Language Processing
  3. Machine Learning (Clustering, Pattern Mining)
  4. Automated Reasoning (Logistics Planning)

Built With

Share this project:

Updates

Submission history