Conversion up – Returns down: Sizekick wants to make webshops more profitable using AI sizing advice

Munich-based startup Sizekick wants to help fashion and sports retailers reduce returns in online commerce through precise sizing advice. Deep learning models and the Hohenstein Institute’s huge body database ensure accuracy.

The topic of returns is as old as online retail itself. Clothing brands in particular have to (learn to) live with the fact that consumers return many products due to fit issues. Return rates of up to 50 percent are not uncommon in fashion retail, which of course has a negative effect on the margin and profitability of the store, but also on the environment.

Founded in 2022, the Munich startup Sizekick wants to solve the problem of size-related returns with the help of a self-developed, AI-based software that offers customers individual sizing advice in the webshop. The idea is actually not new, but only the current technical possibilities with deep learning and AI create the accuracy required for real sizing advice. The two founders of Sizekick, David Oldeen and Jake Lydon, know what they are talking about. They have long known the business with 3D body data and digital sizing advice from their time together at Presize, which also offered digital sizing advice and was acquired by Facebook parent Meta in 2020.

This is why Sizekick has already managed to count a whole range of renowned online shops among its customers after such a short time – from Marc Cain to Outletcity.com and small streetwear brands such as Vicinity. The success factors include not only a simple implementation of the software on the shop side and good user-friendliness on the user side. Because in fact, more than 95 percent of shoppers who click on the Sizekick button go through the advice to the end. The sizing expertise and the extensive database with real body measurements of the Hohenstein Institute, which Sizekick managed to secure as a strategic partner and investor, also played an important role. We spoke to David Oldeen, co-founder and CEO of Sizekick, about the success of Sizekick and the further plans.

Sizekick promises to reduce returns. To what extent have you succeeded so far? Do you have a good example?

One of the best cases is indeed from one of our first customers. We have been live there for more than a year now. There, returns could already be reduced from 32 to 25 percent – ​​across the entire store. And the great thing is that this improvement was constant and not a snapshot, but that we have achieved this in a truly sustainable way. In principle, however, it always depends on where you start. For example, if you have stores with a return rate of 50 percent or more, you can achieve even faster and even greater effects.

The important thing is that we continue to reduce returns. With Sizekick, we didn’t just want to build a conversion engine that only increases conversion on the front end and doesn’t improve anything on the back end. In our case, it does both.

You mean that such a size advisor could simply increase conversion, without actually reducing the number of returns?

Yes, according to studies, about a third of purchases are not made because of uncertainty about the size. Those who are uncertain and do not feel like ordering multiple sizes simply do not buy anything at all. Therefore, ideally there is a combined effect: an increase in conversion and a reduction in the number of returns. That is the expectation we have.

How exactly does Sizekick work in the webshop? How and where do customers encounter Sizekick?

Sizekick is integrated into the webshop via a button. Each webshop gets a custom-made button that seamlessly connects to the product page. That is, at some point the shopper comes to a product that he or she wants to buy, but is unsure about the size. For this purpose, there is our Sizekick button in the immediate vicinity of this size choice. If you click on it, our app opens in the partner’s store and each user receives personal size advice, based on his or her own body measurements.

What information do you need from consumers to determine the correct size?

We work with deep learning models. We don’t use AI as a buzzword, but AI is really the heart of Sizekick. All users go through a journey, behind which is a deep learning model that my co-founder Jake Lydon developed. At the beginning of the user journey there are four questions: What is the biological sex, what is the age, what is the weight, what is the height? Now of course it is an exciting question how much you should ask and how much you can leave out. There are also solutions on the market that only ask for gender, age, height and weight. But with that alone you can’t be accurate, because with this input there are still far too many different body shapes. That’s why we go one step further. Based on the first input, we suggest body shapes, from which the users choose the body shape that most resembles their own. And for that we use real body shapes from the Hohenstein Institute, our strategic partner. Through Hohenstein we have very good knowledge of what human bodies look like.

Sizekick suggests body shapes from a pool of thousands of possibilities. In total, users have to select a body shape four times. Finally, our deep learning model generates the individual body measurements from the entire input. The size proposal is based on this, and the user also sees where he or she is between the sizes and can then individually decide which fit is desired.

How accurate are your body measurements?

With our technology, we achieve a precision that was previously unmatched on the market with smartphone input. For example, if you have an average deviation in body size of five or ten centimeters, which is really common with other solutions, then you quickly jump two sizes up or down in a size chart. With our technology, we achieve an average deviation for all body sizes of 1.0 to 1.5 centimeters. This allows you to give accurate size advice. And that within 35 seconds – our survey does not take longer.

There is also a second way via a video recording. How does that work?

Exactly, we offer every user the possibility to make a video-based body scan. That is, you can choose an alternative user journey in the app and instead of selecting suggested body shapes, you turn once in front of the smartphone camera, let Sizekick do the work and then you get the size advice. No one has to undress for the body scan, the model is trained with data in such a way that it works if you are wearing jeans and a T-shirt.

We achieve almost identical precision with both solutions, but offer users two ways to achieve it.

Why does it make sense to enable both ways?

That way you maximize the use. Maybe one person wants to quickly click through the bodyfinder and someone else says: no, you have to see me, you have to make a video scan of me. We are the only ones who can offer both solutions.

What are the usage rates of the respective variant?

There are more users who use the bodyfinder solution with the choice of body shapes, because it is simply the solution where you do not have to get off the couch and you can also use it in the office or on the train. But that is fine with us. It is important that consumers use our tool and do not drop out, for whatever reason.

Do ethnicities and different body shapes also play a role in finding the right size? Can you be as accurate in the US?

We discussed this extensively with Hohenstein. Hohenstein works internationally and does body measurements worldwide. My most important lesson was that body shapes worldwide do not differ that much from each other. What does differ is the distribution of the individual body shapes. That is to say, in Europe a different body shape predominates than in the US, but there are still the same body shapes everywhere. So we are not limited to the German-speaking area or even to Europe.

So that was the consumer side. What do the brands need to deliver so that you can bring the individual body shape of the customer and the sizes of the product together?

Each brand receives a custom-made sizing system from us. This sizing system is placed in the backend at product level. Behind each product is a dynamic body size chart that is continuously adjusted based on user interactions with the products in the store. In this way, users receive product-specific sizing advice, for example for product A size S and for another product B size M, because the products are different.

How complicated is it for brands or stores to work with Sizekick?

We have made sure from the beginning that the effort on the shop side is extremely low. The level of internal knowledge and available data on the topic of ‘sizing’ on the brand side is often very different. Some young brands have not yet built up any internal knowledge, because the cuts are determined entirely by the producer.

We want and can help all brands and retailers and work with what is available on the brand side. We can do that because, thanks to our background with Hohenstein, we know how to classify this information from ready-to-wear sizes to body measurements and how to create customized sizing systems from that. So we can also help small streetwear brands that don’t make their own cuts. We give them a digital way to calibrate and validate their sizing systems via app.

So a brand doesn’t have to have fully digitalized its product development to be able to collaborate with you?

That’s right.

How long does such a process take?

On the technical side, integration only takes a few clicks, which is done within minutes. And then there is the possibility to calibrate the sizing system before it goes live. In total, you can go live with us within two to three weeks. With other solutions, that was and is often a project of several months. That is why it was a key point for us during development that we set up the system in such a way that Sizekick can be easily integrated by brands and retailers.

You have very renowned clients now. How many are there and which stores are they?

We have been live for a year now and will soon be going live with our twentieth store. We have a really diverse portfolio of customers, from more traditional companies like Joop and Marc Cain to extremely hip streetwear brands with a very strong community. These include brands like Olakala or Vicinity. In a few days, Sizekick will go live in the Sanvt store, which recently won an award for its ‘perfect white T-shirt’. In the sports sector, for example, the American outdoor clothing brand Black Diamond was one of our first partner stores. In the multi-brand category, renowned retailers such as Grube-Versand, the shopping club OutletCity, Waschbär or Rrrevolve from Switzerland are among our customers.

How are the first experiences after a year of Sizekick? Are there any other advantages for stores that result from the collaboration?

E-commerce is generally an optimization game and Sizekick’s analytics play a big role for our partners. For example, they gain new insights into user demographics and buying behavior. On the product side, you get exclusive insights into who buys certain products, where the most uncertainty about size is or which styles and materials are preferred by certain buyer groups.

Every variable that e-commerce stores can control is crucial. For example, in many stores, a third or more of all customer queries are about size. If you don’t have to answer them manually and can send them to Sizekick, that’s obviously handy.

For me, it is also about how many users use Sizekick every day and how satisfied they are with it. For example, we see that more than 90 percent of users who click on our tool also receive size advice. We are very proud of that value, and for us it means that the application is user-friendly and simple.

Does Sizekick also influence the shopping cart?

This statistic is very interesting, because stores that sell a lot of multiple sizes, because customers order multiple sizes of an item and then return some of it, naturally want to reduce their shopping carts first. The goal here is that this customer orders fewer sizes. We were indeed able to achieve that. At the same time, we were also able to measure that more items from other categories were added. This means that with greater certainty about the size through Sizekick, we can also increase the shopping cart, and that in an economically very interesting way.

How do you finance yourselves, what does Sizekick make money with?

The stores pay a monthly SaaS fee [Software-as-a-Service, ed.]. Users pay nothing for the Sizekick size advice in the webshop. The common goal with our partner stores is that as many users as possible use Sizekick to maximize the positive effect in the store.

How many users visiting a webshop use Sizekick on average?

It varies a bit from store to store. If we reach between ten and twenty percent of the users, we are satisfied. Because if, for example, in a store only three to five percent of the users use our tool, you can have the most innovative technology, but if almost nobody uses it, you automatically have only a small impact. According to the feedback from our customers, we achieve absolute top values ​​with the use.

A vision for using individual 3D body data has always been to try on products virtually using avatars. How far away are we from that?

From a technical point of view, we can already display bodies in 3D with our technology. You can relatively easily overlay 2D product photos on top of that, as if the avatar were wearing those clothes. In terms of gamification, we already see something like that in some stores. I would call that ‘virtual try-on’. What I don’t think has been solved commercially yet is ‘virtual fitting’. In this case, the bodies have to be accurately captured and visualized, and then visually matched with 3D product data. I think brands would like to be further along in this, but at the moment it is still very labor-intensive to display every style in all sizes in high-quality 3D data. Only then can you combine virtual try-on with virtual fitting.

Hohenstein is also intensively involved with this theme, but the market is still far from a commercial and especially accurate application in e-commerce. Another theme is ‘mass customization’, where you produce clothing based on the generated body measurements. We are already working on specific projects there. Some suppliers may want to offer a larger size range with more intermediate sizes instead of just S to XL, others want to produce based on body data.

In principle, ‘mass customization’ or ‘production on demand’ would already be possible today, you would ‘only’ have to place a body scanner in every store – which unfortunately costs hundreds of thousands of euros. But that is of course not feasible and also not location-independent. So solutions are needed that work via mobile phone.

What is your goal with Sizekick?

We want to become the standard in e-commerce in the field of sizing advice and ultimately number one worldwide. So there is still a lot to achieve.

This article was originally published on FashionUnited.DE, translated and edited to English.