Jade Huang, Co-founder & CEO of analytical ‘virtual assistant’ platform, StyleSage, took some time out in early August to chat to us and share her story thus far, and her views for the future. Jade’s experience in design, technology, and business shines through in this exclusive piece – as does the depth and rapid development of the platform.
StyleSage have become a frequent contributor to WhichPLM in recent months, and you can find Elizabeth Shobert’s (Director of Marketing & Digital Strategy) exclusive content here.
Name: Jade Huang
Occupation: Co-founder & CEO, StyleSage
Likes: Spicy foods, fluffy animals, and archaeological ruins/historical sites
Dislikes: Long lines, and people who don’t pick up after their dogs
Words to live by: Well-behaved women seldom make history
WhichPLM: We usually begin these conversations by jumping straight into the business specifics, but I’d actually like to begin by getting a quick intro into you. I understand you’re an ex-designer, with some impressive accolades under your belt?
Jade Huang: Sure. My experience in fashion is actually pretty short-lived. I did study Fashion Design at Parsons School of Design and, to be completely honest, I prized partying over going to school, which meant I lost my scholarship and flunked out – much to my parents’ shame – and accidentally found myself involved in technology. I became a digital designer, then user experience, and taught myself programming. So I was actually in tech for about 10 years, before going away to business school at INSEAD where I was inspired, by a business case on Zara, to start this company. I suppose I found my way back to fashion after being in love with technology for a long time, and found a way to bring them together.
I knew enough about the fashion industry to know some of its challenges. It was then we had the idea, and we went out and talked to actual veterans of the industry to really understand the perspective of things that really keep them up at night, things difficult for their teams, things distracting them and diluting their focus. And that’s how this idea came about: to automate a lot of processes that their teams were having to do manually, so they could focus on what they do best which is creating really beautiful garments that people want to wear.
WhichPLM: So you’ve had experience as a designer, as a technologist, and as a businesswoman. In your own words, how would you describe StyleSage? You’re a startup, delivering insights and analytics to the world of fashion?
Jade Huang: I think of it as a fashion business’ virtual assistant. And the reason I say that is because an assistant has one of the hardest jobs: he or she has to do multiple different things for multiple different teams, and that’s kind of what our platform does. In our journey working with clients, our platform is one analytics platform that works with the strategy teams, the consumer insights teams, the pricing teams, the trend teams and design and product development teams, all the way to the marketing and e-commerce teams. Each of these teams is very, very different in the nature of what they do and their own disciplines, so it’s one platform that works for all of them. In most, specifically, we address four key areas for them: pricing, assortment, promotions, and trend.
WhichPLM: Tell us more about the technology behind it – we’re sticklers for how things work, and the majority of our readers are focused on the technological aspects of businesses.
Jade Huang: To get into the nerdy side of things, there are a couple of different layers that build the platform up from the ground.
Number one is a data collection process. In collecting data, we build something very similar to what Google search technology is. People often know that as either scrapers, or crawlers, or spiders. And essentially what these scrapers do is they go onto e-commerce sites and social media sites and collect data. That’s the basic layer of how we actually capture that data. But our spiders actually go beyond collecting information that is on a page – we dive deep to collect information on sizes, availability, color options – all the critical information to evaluate the market options available to the customer.
The second layer is, once we have all this data, how do we clean it up? For example, in the States at least, retailers refer to jeans in two different ways – jeans or denim – so everyone uses a different term even within the English language. If you multiply that complexity by the different languages spoken around the world (if you’re scraping sites from China, or Japan, or Germany), even within the same locations, you can see the challenge. So, the whole process of how to clean that data takes all of that into account; we translate all those different languages into English and we standardise into one universal taxonomy, so that when you’re comparing jeans you’re really looking at apples versus apples, and not at jeans versus chinos versus linen pants.
Then the third part is making all this data understandable – the technical term for which is descriptive statistics. It’s describing things around what it looks like, for a brand, around an assortment for women’s tops for example. It’s 20 percent of their overall apparel assortment, but within women’s tops there are different sub-categories of tunics, button-down blouses, tank tops etcetera. And that’s not rocket science, but it takes really diligent cleanup. So we describe that for them, which is the third layer.
The fourth layer is the really interesting part, which is taking all of that and understanding that it’s not just a two-dimensional piece of information on what assortments look like, but there are actually multiple dimensions on top of that. How do you look at, for example, the assortment not only by styles but also by colour, by materials, by pricing strategies, and also by sleeve length, neckline or hemline? All of these different elements that drive consumer-purchasing decisions are difficult to extract when looking from a numbers perspective, so we use image recognition to extract those attributes, and create a multi-dimensional view of what the market actually looks like.
WhichPLM: That’s a very detailed yet clear view, thank you. We’re seeing more and more video coming into play – as well as images – incorporating augmented data over the top. Where do see that? And how do you see yourselves plugging into that?
Jade Huang: In terms of video, there are two ways I’m thinking about it: one is the straight up advertising, and promotional videos on Youtube, for example, where a celebrity or influencer is wearing a certain item. The other part is around virtual reality video and what things are actually shared in those alternate reality worlds. That’s a very new, emerging field. VR is hot but it’s not yet widely adopted so, for us, we would prioritize video because it’s a medium that’s been around for a long time, but in the format that has wide adoption amongst consumers.
In our next phase, we will be looking at how we capture those videos. Video is just film, which is still frames put together in very rapid format to create the illusion of movement. What we would do is take a technology to create stills of a video at certain points in order to extract the image and analyse it that way. It’s the method of extraction that will be different, but the method of analysis and extraction will be the same – it will be familiar methods that we can apply.
WhichPLM: With the rise of video as promotional media there’s a big rise in using augmented data in combination with this – data coming over the top of a video around an item you see, it’s price and where to purchase it. Do you see any of that creeping into your platform?
Jade Huang: Not in the short term. And the reason for that is that when you think about the purchasing journey, if I’m a consumer looking at this video and I click through to a site to purchase something, ultimately the point where the actual product still sits is still in a website format or a mobile format. What you would be capturing is the journey from the point they click to actually purchase. Within our platform we do look at Instagram and what influencers are sharing from a visual content perspective to be able to see rising trends and we pair that against, let’s say, Google search trends so we can understand whether consumers from a certain location and influencers from a certain location are sharing similar styles. Then as a retailer, can we systematically de-risk our decision by allocating some of those styles to a particular location, or do our marketing segmentation target for that location with similar styles.
Video is just one method for which to capture consumers in that journey, and it’s more from a marketing channel perspective. From a marketing channel perspective we do capture what the different emails are that retailers send out to consumers, what kind of homepage promotions they’re running, what kind of design and creative tactics they’re executing. And certainly video will probably become the third format that we incorporate, but essentially what we will capture will be the types of video marketing activity different retailers are using, to provide a benchmark practice to see whether that’s a channel that marketing teams need to start allocating a budget for. In terms of actual content itself, we will probably analyse the length of the advertisement, what products are being shared within those advertisements, and what the consumer responses are. On Youtube, the simple thing to capture is how many likes and how many views. If you pair that against marketing data, and see that a type of product is selling out at full price, and you’re seeing popularity around what consumers are searching for and influencers are wearing, then you can make a really strong bet around whether a product is here to stay.
WhichPLM: That makes perfect sense. With regards to all of this data that we’re describing, is this data available for multiple clients? Can more than one retailer or brand, for example, access the same data set as their competitor? And if so, are there any challenges there?
Jade Huang: Yes, they can all subscribe to the same sets of data if they like. The only data that is not shared between clients is if a client elected to upload their own POS sales data to analyse it against, let’s say, the market product pricing movement. A client can do that, but their own data would never be shared with other clients. Other than that, if a client did subscribe to the same competitor set in the same country, then yes. The reality is that each organisation has a very different way of working and a different strategy. Two organisations can receive the same set of information, but make different tactical decisions based on different priorities.
WhichPLM: You mention that clients can elect to upload their own information to work with. So, not only do you supply data sets to help against the competition, but you also analyse a client’s own data. Is that right?
Jade Huang: We do have clients who, after using our platform, can opt to export that data so that they can bring it into their internal systems and can do additional analysis if they want. We do have some clients who do that, but typically they have pretty advanced data science teams, or statisticians or consumer insights teams. If it’s a client where we’re working with a more creatively-focused team (like a merchandising or design team) that almost never happens because they just need to know top line statistics, rather than deeper analytics around that, and get to what they do best – designing beautiful garments that consumers want.
WhichPLM: Of these clients we’re discussing, where are they spread on a global basis? Are you an international company, or predominantly North American?
Jade Huang: We’re pretty international. From day one, our company has been split between Europe and the US. We are a US-based company, but our second office is in Madrid, and we’re split between the two. We have clients from all over the world, in Asia, in Europe, in North America, as well as South America.
WhichPLM: Is there a particular use case or client (that’s public knowledge) that you can tell us about, in relation to their use of StyleSage, and the benefits they’ve seen?
Jade Huang: Sure, I can speak around the use cases from an anonymized standpoint. Most of our clients, except for just one I think, do over one billion [US dollars] in annual revenue, so for a lot of them they’re publically traded and we have been spoken about, in public shareholder calls, as a key ingredient to how they’re growing their business. It’s a sensitive topic to discuss exactly how from an individual business perspective, but I can speak more generally on use cases.
One of our clients, whose merchandise is sold in over 40 countries around the world, has a very particular strategy where they look at anchoring their prices to a set of aspirational competitors. This is a really interesting strategy if you think about it because their style will have a similar appeal to the brands they anchor to, but their prices are much more friendly. So, for them, when you’re sold in 40 different countries, there are two things: one is their aspirational brands in the major hubs within their home turf, and the second part is once you go into more localized markets (take China for example), there are not only those aspirational competitors on a global level, but also localized competitors on a local level – so how do you not only localize your assortment, but anchor the right pricing so you’re in a similar market position? They use us to very tactfully do that, and for them it’s been a winning strategy for their profit margin; they’ve been beating their quarter on quarter estimates. That’s a very specific strategy that they use, and in this case, from a global perspective, if I’m the CEO sitting in headquarters and I know that these 40 country markets are my key markets to expand into, and I want to make sure my teams are all executing on the right strategy, that platform has both the purpose of on the ground teams being able to execute on the anchoring strategy and the senior side being able to monitor that without having to wait for reports that bubble up from the ground. It’s aligning an entire organisation so they’re moving much faster.
WhichPLM: You mentioned earlier, when we were discussing the technology and the levels within the platform, a couple of areas that you excel in. If someone were to put you on the spot, like we are right now, and ask you how StyleSage differs from other analytical offerings out there, what would be your major points?
Jade Huang: I think the nerdy answer, which people who are very data driven will definitely understand, would be around how we were born as a company. My co-founder and myself are pretty quantitatively driven people, so coming out of business school, the first two years when we were building the foundation of the platform, we invested all the energy and money into building the infrastructure – the right database, scalable technology, the right QA pipeline. When you look at data quality, if your QA process and your pipeline aren’t built to be super efficient, you can’t scale to crunching to, say, over 500 million data points per hour and have that data be very clean in your output. So we invested a lot of money there.
The way that you can think about it is like building a house – you can build a house with a very solid foundation and when that design becomes outdated in 20 years it’s very easy to give it a facelift with some renovations, but you can’t go back and change the basement and the foundation material without ripping down the entire house. I think our competitors invest a lot of time in building something really beautiful, that creative teams are attracted to, but they fail to look at the infrastructure, and a lot of their customers have actually switched to us, because once they got comfortable with the analytics they saw that most of the analysis didn’t make sense because the data quality wasn’t there. I think that any analysis you make, if the data quality isn’t there, it really doesn’t inform you of anything. We invest a lot of time and energy into that, and a lot of proprietary cleanup technology that we developed goes into that process.
That’s the nerdy answer.
On the other side of it, because we’re a pretty young company we’re incredibly nimble. Last year alone we did over 100 releases into our platform with an engineering team of seven. So, we move really fast. It’s mainly because engineers are motivated by working with cool stuff. For all the clients we work with, the reason these billion dollar clients trust a small business like us, is that if they want something …it’s in the platform next month.
I think those are two of our main competitive advantages. The platform has now moved to a point where one platform services four different teams within and organisation, so clients really get value out of it. They don’t need to purchase four different systems for four different teams that have completely different needs.
WhichPLM: Your point about how quickly you work is all, I suppose, dependent on passion. Great businesses are built on passion. Speaking of how quickly you’ve been churning out releases, are you able to give us a glimpse into any upcoming developments? Or any news in the pipeline for the coming months you can share?
Jade Huang: [Laughing] We’re actually in the middle of working on a press release [Editor’s note: this conversation took place early August 2017], but we do have two major releases we’re pushing out in the platform. They officially went live in the platform yesterday, but we will be announcing that not only have we made a correlation between search data versus influencer data, but we’re also going to pair that with pricing data in what we call a trend dashboard. That’s the first thing we’re releasing; the second is a dedicated global promotional tracker. What’s important about this particular feature is looking at promotional activity and tactical marketing activity together in one single dashboard that’s really helpful for e-commerce and marketing teams. That one is built from feedback from our clients, who all have really different types of business – from activewear, to department stores, to fast fashion – and we’ve taken all their different teams and the way they work and built it into one single dashboard that’s really flexible for them to use.
That’s something we’re really excited about rolling out very soon, and there’ll be a press release imminently.
WhichPLM: We’ll let you go shortly, we promise, but just one final question away from your specific sector: what would you like to see coming from other applications and software in the not to distant future, for fashion as a whole?
Jade Huang: I have two particular things I’m really excited about. The first is the buzz topic of the moment; VR and AR are very interesting, and this is the new generation that wants very different experiences. This is a topic that’s been beaten to death recently, so it’s nothing new, but I am really excited about it. I think bringing that type of AR and VR into the masses really changes the way we function as a society – not just the way we shop, but the way we interact with brands and learn about causes that are important to us. We’re quickly seeing that fashion doesn’t happen in a vacuum; it’s a lifestyle and something that people want to be a part of and emulate. I think that retailers in the last few decades have sort of lost touch with that, and it’s a great chance for them to re-engage their customers and get to know this new generation of consumers, and how they behave and what they think, and be able to service that better. I’m really excited about that.
On the non-sexy side, that consumers don’t really think about, I’m excited about this new importance that’s being put on supply chain improvement. That’s actually the backbone of what makes a business successful, but I think that within the industry itself, in my cursory understanding, there hasn’t been that much disruption from a startup perspective. We’re starting to see little startups popping up here and there that look at logistics and supply chain efficiency, and I think that’s really exciting. If you think about why Zara is number one in the world – and has beaten everyone to the punch – it’s because they’ve understood that and they eat the higher costs of manufacturing by having their manufacturing facilities very close to the distribution centre and very close to the creative team so they can bang out new collections at the right price for their customers in six weeks. That’s amazing. A company with a supply chain in China can’t really move that fast – but maybe with the right software, you can close that gap. That is really exciting to me, but it’s the ‘less sexy’ part that consumers don’t really care about.
WhichPLM: We want to re-enforce that point; we agree completely. We’ve been having conversations, with the likes of the Inditex group like you mention, dating back to the turn of the millenium around that ‘non-sexy’ neglected space of the supply chain – that connected element.
A large proportion of today’s European and North American ‘sexy’ retailers and brands – and active sportswear brands and so on – are actually practicing PDM as opposed to product lifecycle management. What you’ve said about those supply chain improvements is hugely under-supported and under-managed; it’s still disconnected and it’s still on Excel, all these years later.
Jade Huang: Indeed. And I’m excited to see how things do progress in the next ten years. Even if you think about the way that we behave today, versus 2007, versus 1997, 10 years brings exponential growth.