Bringing Shavatar’s statistical model to the cloud for accurate clothing size advice
Belgian fashion start-up Shavatar – a spin-off of Antwerp University and strategic research center imec – created a statistical model that helps shoppers choose the right clothing size when ordering online. Based on the user’s bust, waist and hip measurements, the tool builds a personal 3D avatar that has a margin of error of only 7mm.
Online clothes shopping has skyrocketed in recent years – over one-third of all clothing purchases now take place online, representing around $403 billion globally.
However, variations in sizing across brands have fuelled the rise of the ‘serial returner’: 41% of customers order multiple sizes with the intention of sending back items that don’t fit, resulting in around 70% of all purchases being returned. Retailers bemoan a ‘phantom economy’ of lost revenue due to returns, and the practice also has an environmental impact.
The team behind Shavatar was convinced of the huge potential in helping customers choose the right size the first time. They developed a statistical shape model that could create an accurate 3D ‘shavatar’ (a portmanteau of ‘shape’ and ‘avatar’) with just three body measurements and a few other details – equipped with a personalised shavatar, a user could then access sizing advice for over 50 popular brands.
The tool wasn’t yet accessible to anyone other than its creators. So they challenged November Five to create a Minimum Viable Product that would bring Shavatar to the outside world, while keeping a close eye on costs.
Once the Shavatar team had explained their product vision and ambitions to us, we carried out a time-boxed technical ‘spike’ to explore the feasibility of running their statistical model in the cloud. When we saw this was a viable approach, we organised a series of workshops to define the user story backlog and functionalities together. These then served as the basis for the User Experience and User Interface designs.
As scalability and cost were key considerations for the start-up, we opted for serverless technology. This also helped us reduce development time – using our scrum methodology and weekly check-ins with Shavatar, we were able to deliver a working MVP in just five weeks.
Within eight weeks of the launch, a total of 11,000 shavatars had been created, with 3,000 users registering an account. Eventually, Shavatar hopes to integrate the model into webshops so that customers can check their sizes live while shopping.
created in less than eight weeks
render time for your personal 3D model
“We were delighted with the in-house expertise and collaborative approach of November Five to bring our model to the cloud in just five weeks.”
CEO Shavatar
Follow us on LinkedIn for insights, learnings, use cases and more.
To explore whether it was technically possible to run Shavatar’s statistical model in the cloud, and carry out a feasibility check in a short time frame, we conducted a technical ‘spike’.
As part of this, we started rudimentary development to check for performance and scalability – as well as making the tool available to the general public in the short term, Shavatar also has big plans for future B2B iterations that could transform the online shopping experience for users and help retailers reduce the ‘phantom economy’ of lost revenue from serial returns.
Once feasibility was established, we organised a series of workshops to define the user stories, and subsequently, the functionalities. With these defined, we established the User Experience (UX) flows and created the User Interface (UI) designs for the various pages that would contribute to the creation of a 3D avatar.
Shavatar’s MVP communicates a user’s request via an API and then taps into the statistical shape model which turns the measurements into a 3D avatar.
Traditionally, this would all have been hosted on a dedicated server, with starting and infrastructure costs of around $1000.
As part of our efforts to optimise the budget for the start-up, we used serverless technology to eliminate the infrastructure costs and bring the model and user profiles to the cloud at a fraction of the price – serverless only charges for the computing power you actually use, so starting costs are low.
The flexibility of serverless infrastructure also means we can expand capacity according to Shavatar’s business needs as they develop.
With its easy-to-build 3D avatar and integrated data from over 50 popular brands such as Adidas, Superdry and Zara, Shavatar’s MVP already has clear real-world appeal.
Eight weeks after launch, the tool had already attracted nearly 3,000 registered users – by registering, users can save their Shavatar and can return each time they need sizing advice for a brand. Including non-registered users, a total of 11,000 avatars were created in the first two months.
Shavatar is now looking towards their next B2B phase, using feedback generated from users. Their goal is to integrate the model into e-commerce websites so that customers can check their sizes live while shopping, and make size-related returns a thing of the past!
Good news for shoppers, good news for retailers, and good news for the environment.
