Despite the progression of technology, Savitude founder Camilla Olson perceived that fit remained a lingering issue for many women, designing an AI solution that merges styling with machine learning to match all bodies with brands ideal for them

Sizing issues are no stranger to the fashion industry, and many retailers have invested in expanding their size ranges to accommodate a greater diversity of body types. However, even within these extended runs, a glaring problem remains: fit. In fact, the top reason women return merchandise is not necessarily that they had the wrong size, but that their theoretical best size still left more to be desired.

This is something Camilla Olson, founder of AI fashion personalization solution Savitude, noticed after working with clients on bespoke styling projects. Olson combined her background in ecommerce and technology with her first-hand experience working with unsatisfactory fit, seeking to create an AI tool that could go beyond identifying a correct size and actually recommend merchandise to women based on what fit their particular body would need.

PSFK spoke to Olson to learn about Savitude's cloud-based, machine-learning technology that offers customers visual discovery, curates tailored outfit recommendations and yields crucial data for retailers interested in personalized re-marketing and localization. Olson touches on building plug-ins for retailers like Shoppable, enabling them to better match apparel with customers based on fit, as well as creating custom hashtags for clients once their fit profile is established that then yield shoppable feeds on platforms like Instagram.

PSFK: Could you tell us about your background and what led you to found Savitude?

Camilla: I started off in the pharmaceutical industry doing electron microscopy, regulatory affairs, financial analysis and product management. I was a venture capitalist for four years. Then I moved to the other side of the table as a serial entrepreneur. I founded five companies in the biomedical space.

I thought I was retiring. I spent a lot of time with my children then, and went back to something I had been doing a lot since I was a kid—sewing. I earned a master's in fashion and textile design, did a runway show at Lincoln Center to launch my ecommerce label and ran that for five years. During my time at the ecommerce label, I recognized a problem with fit, except that I thought of it in a personal sense. I was embarrassed, thinking, “I'm the only one that's not doing a good job with fit.” It took me a long time to realize that everybody's got the same problem.

I spent a good three years researching fit and even developed an iOS app. Then I realized we could solve the problem using technological approaches that I had experience with in earlier companies (that I founded).

Could you expand upon some of the specific aspects of fit that you realized needed improving and how you went about doing so?

We had a label with collections that we shipped to boutiques, that we offered online, and we custom work. The Franco family lives in our town. In our first year, James Franco was hosting the Oscars, and Betsy Franco needed some clothes. We dressed her for four events around the week of the Oscars. That was my first real custom and celebrity experience, but we ended up doing gowns for the Met Gala, White House and LACMA Gala events. Soon it expanded to other socialites in the Bay Area.

Working with real people with real bodies, I developed a real understanding that it wasn't only size that has an impact on fit. I got a really clear understanding of matching of body shape and silhouettes of clothing. Working with clients I would suggest that they try a dress on, and clients would say, “I don't really want to. I don't usually wear that.” I'd say, “Really, trust me on this and just try this one on. They would put it on and say, “Hey, this is actually pretty good.”

I developed a good sense of what would actually work for people by just working one‑on‑one with my clients. We developed an understanding from doing research on our own that there were nine body shapes and four key proportions. We drew 750 illustrations to show people how to dress for their body shape. From there, we set up an educational program to try and help women figure out how to look in the mirror and say, “I can recognize that this is a good look for me.”

Those illustrations are on our site now and can be found after taking our quiz. We started to link those illustrations to items in our shop, which led us to thinking, “Well, wait a minute, we turn this into an AI tool.” We now take the images that are on the ecommerce site and classify them, based on their design details. Customers answer a very easy quiz, and we'll then select from the assortment, the items that will look best on each customer. We can present the best items to this shopper in the context of outfits that are fully accessorized for their body shape and proportion.

The shopping experience is as if a stylist has curated everything for the customer personally, as if it were styled for her.

 

View this post on Instagram

 

A post shared by Savitude (@savitude) on

Could you tell me what the user experience is like for customers coming to Savitude?

We knew that if we wanted to sell our product, we couldn't require our retailers to do anything—we didn't want to make any demands of them, and knew that we had to be full-service. For example, we don't require any information for installation. We get their permission, of course, but the only thing we need them to do is for shoppers to take a quiz, so we consider ourselves 100% automated. We have an application in for approval on Shopify as a plugin. It's pretty much one‑touch and we're installed, on the retailer side.

We have four products. First is Flattery. It makes recommendations, that matches a body to which products work well for the individual body shape. We map silhouette, and body shape and proportion. Second, we form complete outfits that also utilizes available accessories. That can be done with or without flattering a body shape. Third, is our Find Similar product pieces. That's really fun. It reimagines a dressing room and you say, “I love this dress except it's got a round neck, and I really need to wear V‑neck.” Find Similar will do that. Then the fourth product is our loyalty product, using branded hashtags. That's at the end of the quiz. We can give customers an identity. It's a way for us to recognize them electronically.

Do you have any concerns about data and privacy when it comes to personalization? How does Savitude navigate that territory?

We are very concerned about privacy and data security. We thought about it when we put our initial spec together. We knew we couldn’t ask retailers to do the work of installing our software; we knew retailers wouldn’t give up their customer data. So we keep a digital wall between us and the customer. We know the body shape, the type of electronic device and her city. But we know nothing else.

What do you think is next for apparel retail and tech as we move forward? Do you think we will see more personalization, more AI, visual search—or something else? 

Yes, AI will succeed where it is helpful. It will die where it is based on fiction or regurgitating historical flaws. Recommendations based on past purchases perpetuates flawed decision making; it works only for those who know how to buy fashion. Companies where data is organized, linked to outcomes and can be repurposed will succeed. With the above recommendations, visual search and outfitting can be great helps to shoppers, gift givers and retailers.

Future innovations that we look forward to are using AI to predict trend and to design. We see this in three embodiments: from runway to streetwear, from inspiration to runway, and to perpetuate designs in the spirit of iconic designers.

Is there anything you could share that you're planning to be one to two years out? Any features you want to enable, or more social media integrations, or ways to further engage consumers?

We did a complete analysis of the fall 2019 season. It's 120 pages of where fashion designers are and who they're designing for. So many women are being underserved. Our long term goal is to get designers to design for all women, so that we can easily find clothes to fit our body that looks good on us. That's really what it's all about.

Pictured: Camilla Olson.

Savitude

For more from services like Savitude improving retailer and customer experiences, see PSFK’s reports and newsletters


Lead image: stock photos from Jacob Lund/Shutterstock

Sizing issues are no stranger to the fashion industry, and many retailers have invested in expanding their size ranges to accommodate a greater diversity of body types. However, even within these extended runs, a glaring problem remains: fit. In fact, the top reason women return merchandise is not necessarily that they had the wrong size, but that their theoretical best size still left more to be desired.

This is something Camilla Olson, founder of AI fashion personalization solution Savitude, noticed after working with clients on bespoke styling projects. Olson combined her background in ecommerce and technology with her first-hand experience working with unsatisfactory fit, seeking to create an AI tool that could go beyond identifying a correct size and actually recommend merchandise to women based on what fit their particular body would need.