When it comes to fashion and shopping, things can get confusing. Mood, price, occasion, and trends, are just some of the factors that temper with our rational (and often irrational) decisions. As if things weren't complicated enough, the boom in the online fashion space has led to countless choices that can drive even the most fashion-conscious person crazy. And to top it off, brands can run wild in terms of fit; each company has its own set of sizing rules that seem to change every other season.
With the huge amount of sartorial offerings online and the inconsistencies in how they fit, it's not surprising that an increasing number of fashion-tech companies are getting more 'personal' with the consumer. By trying to solve the puzzles of context, fit, and more importantly, one's definition of style, personal recommendation websites are working on two problems: the customer's pain in selecting what's right and the retailer's hassle of dealing with returns.
Dressipi, co-founded by Sarah McVittie and Donna Kelly, is one online company that seems to be getting it right; the service launched in public beta during 2011, and it now boasts over 200,000 members (mostly by word of mouth) with an email open rate of 45% (whereas the industry average is around 22%). The average percentage of garments that are returned at retailers (after being recommended by Dressipi) is 10%, in comparison to the industry average, which is around 30%*. Moreover, the company was recently included in Vogue's top 100 places to shop online, while Sir Stuart Rose (of Marks & Spencer) has joined the company as its chairman.The Dressipi way
Dressipi members engage in a range of activities such as tagging, liking, and disliking items, while at the same time the system observes and collects data on their implicit behaviour. In addition, by asking for details regarding a user's favourite item and identifying the garment’s multiple features (material, necklines, etc.), it can determine specifics such as the user's style. The data gathered are filtered through a set of rules dictated by Dressipi's stylists, and a unique profile called a Fashion Fingerprint is created for each member. This probabilistic model of what the user may like offers personalised garment and outfit recommendations. The more the user interacts with their profile, the smarter the system becomes concerning the recommendations it makes.
Furthermore, Dressipi has a Size Finder application, which predicts a member’s clothes size in over 500 brands and is specific to each garment type. Other fine touches include an option called 'Reveal & Conceal,' where women state which parts of their body they would prefer to hide or emphasise, and these preferences are then taken into account.
How much more sophisticated can shopping get? As Donna Kelly mentioned on Wired: “The aim is, you walk into a retailer, scan a barcode and the app tells you the dress will go well with that jacket you bought two weeks ago. The millions of data points that Dressipi collates could offer a far better experience than human stylists.”
Could you tell us what Dressipi does for consumers and retailers alike as a service? What is it about?
Sarah: It’s about understanding how women shop and how they behave. We focus on getting what really matters to women when they go shopping and what is good for them through a number of characteristics: their body type, skin colour, the occasion for which they are shopping, and so on. As a person, I may love what Donna wears, but I might not be able to wear it because it's not right for my body type. So I like to receive advice from the experts on what is suitable for me, then look at people who are similar and what they like, gain knowledge on new items, styles, and brands, as well as how to create a perfect outfit for my type and needs. Eventually, this will help me build up my fashion confidence overall.
Donna: And when it comes to retailers, we can look at the data we have and the range of clothes of a retailer and say, “You know, those didn't sell very well because they’ve all got high round necks, and the majority of your audience doesn’t like that.” So because we understand the details of a garment, we have a different view on it. We can calculate and do all of the analysis on the data that accompanies a specific garment and pull something out of it that the retailer cannot.
Do you describe each garment in every possible way, from a basic description to categorising it as 'elegant' for example or whether it belongs to a certain age group?
Sarah: Absolutely. We have occasions, and then there’s also an inferred way of recommending an item, based on who likes it.
Donna: We tag the trend, occasion, and various other things as well.Sarah: However, you have to be a bit careful with things like age; just because somebody is in their 30s or 40s, you can’t assume that you know their style.
You mentioned a person's 'fashion confidence', how do you measure that?
Sarah: Through a number of different questions and data points on the site. It’s one of the most important things to understand what our user's goals and motivations are. If someone’s got a low confidence, they want as much handholding as possible, as they basically haven't learned how to interact with fashion, they haven’t thought about themselves for quite some time; they’ve been wearing the same jeans and t-shirt for ages. They will then buy something that we recommend, get positive comments from their friends, come back to us, and continue from there.
Also, we are all about helping people to discover new brands or discover new ways to wear things. As soon as they become confident, they stop caring what our stylists say, they know exactly what they want, and for them the process goes: “you know me, find me this, save me time.”
Stylists are an integral part of your team. Can a user contact a stylist directly?
Donna: Yes, they can send anything. And they do send anything. [laughs]
Sarah: We get all sorts of feedback and comments. We had an email from someone’s husband saying that his wife stopped worrying about her weight, she started wearing things that fit her, and their marriage is now amazing...
So you are also saving marriages! [laughs]
Sarah: Haha, yes. Dressipi has changed my life as well. I used to wear pretty revolting clothes — I was just a jeans and t-shirt person — but I now have a completely new wardrobe. I understand how to use clothes and it’s amazing having that confidence. So we see that very linear journey in all of our customers.
What was the biggest problem that you had to solve when launching Dressipi?
Donna: Capturing the emotion of taste and fashion in an algorithm. We had technical people sit with us from day one and talk; having a developer discuss fashion with a stylist was quite an interesting conversation.
Sarah: It felt like “Beauty and the Geek,” and it is extraordinary how these two teams worked together. Another thing we had to deal with was the language that the industry used around style, which was rather meaningless. Even though somebody might say, “I am a rock chick,” it is not very helpful. You have to look at a person's shopping behaviour as a whole when it comes to tastes.
The Daily Mail had an article some time ago about virtual stylists, and the writer felt that they were not really helpful. How do you respond to that claim?
Donna: We actually got to meet the writer after that piece and had breakfast with her. The thing is that we had to launch the product, get it up and running, grow the base in order to get in more data and make it work.
Even at the risk of receiving negative feedback?
Sarah: Yes. When we started we weren’t pushing for any press. Then we got a three page spread, and the next day we got this piece in the Daily Mail. What the editor said was really accurate, but we are dealing with a very complex problem; we are a small team that is growing constantly and we needed time to get things sorted and relevant.
So what has changed since that article?
Sarah: Everything. [laughs]
Donna: The algorithms that we’re building are very complex, and we now do single product recommendations really, really well. However, if you are a very confident fashionista, we won't service you particularly well as this person is not our primary focus at the moment. So we do great single product recommendations, and our whole focus is extending out to outfits.
How do you keep up to date with the consumer's preferences? People's tastes do change over time.
Donna: They definitely change, but our users are putting the data in themselves. On the Fashion Fingerprint there is a thing called “Features” where for example it says, “With all of your activity so far we think you don’t really like red,” or “We really don’t think that you like fitted shirts,” and then you validate that. We won't change things without saying, and also there is some inferred things as we learn about you.
Sarah: We are also adding things over time, like life stages (a pregnancy for example) or you could send an email saying, “I just lost stated weight, where do I start from?” for example.
Could you tell us about your average customer?
Sarah: They tend to be busy working women who need to save time. The average age is 34. Lots of moms who just had kids, and have very limited time or don’t live in London, and they have to have access to all the brands they like to have over the internet. About 30% of our members are in London, and there’s about a 50% split between metropolitan areas and rural areas.
How does social media influence the way people interact?
Donna: When you ask people about shopping and social media they always say, “You know, Facebook helps me find out what my friends like,” which is something that we’ve all been doing forever.
So people are using a different medium to do the same conversation.
Donna: Yes. However you have a more expedient, wider conversation. Where it becomes interesting is when brands and retailers want to reward people for that. It makes the conversation murkier when suddenly you as a friend are telling me that you like Acne. Is it because they are paying you? Does it therefore make your recommendation less valid? So we have to be careful about it. It’s a great form of conversation that’s already happening, and there is definitely a lot more interesting things that you can possibly do with it, but…
Sarah: It’s about who you trust, isn't it?
Indeed. The fashion industry was a bit slow to catch up with the new trends in technology, but we are now seeing a lot of start-ups trying to get a piece of the industry. How do you feel about this?
Donna: It’s exciting that are a lot of people trying to improve the fashion industry or the consumer experience.
Have you seen any disruptions?
Present company excluded! [laughs]
Sarah: I do think that we approach the retail model differently. If you look at the retail sector generally and how it operated with all the unwanted stock for example, a big part of what we want to do is to help make the industry more efficient. They have average sell-through rates of 60% and average return rates of 30%, and they have average conversion rates of about 2%, maybe 3% if they are doing really well, and the numbers can definitely improve significantly with the use of technology. But a lot of retailers don't necessarily have the skill sets required to process all of the data. The more we meet retailers, the more we understand that only now are they beginning to understand how important data is, so there is so much opportunity to make the industry better for everyone involved. So there are a bunch of sectors in the fashion segment that will deliver value and will be disruptive, but I think that where we will see the most disruption is in the personalisation space.
Donna: Exactly, as it’s one of the hardest spaces to get right.
Sarah: And with fashion, emotion and context is as important as the brand itself. A lot of the time it doesn't matter what the brand of a dress is, as long as it makes me feel sexy and it is appropriate for the occasion that I wanted it for. How do you understand that emotion, that context in respect to that dress?
What do you think is online fashion shopping going to be like in the near future?
Donna: It’ll definitely be personalised, more in context. What is exciting is what retailers do online and how that fits with what’s going on in store and mobile, and connecting these different approaches to shopping but at the same offering a unique experience to the user in each medium. Ultimately, brands just have to listen to their consumer, as a lot of companies make the mistake of a) not listening or b) protecting themselves twenty years in the future. So they might have the right idea but slightly at the wrong time, and the consumer will tell you that, they definitely will. Consumer behaviour doesn't change overnight, and it never will. So, it’s about using the tools at the right time to deliver the right consumer experience.
Sarah: The good thing about online businesses is that you can tell if you’re making a mistake or not. You can see your conversions increase if you get it right, and if you don't the data will tell you, so you can learn quickly. Also, that optimised shopping will be about getting a brand experience, like going to the Apple store. You don’t go to do the convenience purchase, you go there because it’s a great shopping experience; everything the brand stands for is there.
*All data provided by Dressipi.