Markets by Grant | Personalized Fashion Ecommerce

11 min readAug 9, 2021

Hello and welcome to another issue of Markets by Grant. With great boldness, I have selected a space which seemingly everyone has an intimate take on. We’ve all spent countless hours online looking for the perfect pair of what-have-you’s, and we’ve all developed our own well-worn paths to purchase — our own systems and shortcuts. Fashion ecommerce players are working to keep pace with the growing sophistication of ‘us’, but it’s a lot of work.

Before I complicate things, let me simplify. The job of the e-tailer is to optimize three things:

What’s behind the screen
Who’s in front of the screen
The screen itself

…Behind the screen we have product design, supply chain, and other core business ops. And it’s up to marketing, advertising, and branding to get the right people in front of the screen. The screen itself, of course, is the critical interface where it all comes together: it’s what unites the customer and the product. It’s up to the screen to convert products from ideas into sales, and to convert people from leads into customers. This is the fundamental competency of ecommerce today, and it’s what I’m going to zoom in on in this analysis.

Now, time to complicate things. Today’s fashion ecommerce players have a lot to think about. Here are some of the more salient challenges and trends “on the screen” today:

Pricing: How can I optimize pricing to match demand and position my products versus my competition?
Promotions: How should I think about drops, flash sales, and special discounts?
Shipping: How should I handle fulfillment/distribution? Should I charge customers for shipping?
Returns: How can I reduce customer return rates?
Subscriptions: Should I be offering Amazon-style, Costco-style, Nike-style, or Stitch Fix-style subscription/membership programs?
Channels: Which channels should I sell my products on? What is each channel’s role/strategy for my brand?
Search: How can I optimize search and recommendations to maximize sell-through?
Sizing: How can I ensure customers pick the right size online?
UI: How can I optimize my user interface to maximize conversion?
Frontier Tech: Should I be using voice, video, AR, chatbots, etc to enhance UX?

It’s a long list — and it’s very tough for today’s retailers to tackle. Given the relative fragmentation of fashion ecommerce (besides the ‘zon, of course), there’s not much centralization of expertise. Most brands are left to try to answer these questions for themselves — and they’ve got a lot of more fundamental operational efficiencies to achieve first. Meanwhile, the market is huge and has historically been growing amply (I’ll get to that later), which means that many players don’t feel a strong urgency to answer these questions right now.

In my opinion, this has led to an overall stagnation of innovation in the space. It’s a hot take, but most of the trends we’ve seen haven’t been all that groundbreaking (APIs for social media? Video shopping appointments? Modest UX feature improvements?). In other words, if I time traveled my ten-years-ago self from his Urban Outfitters phase to my living room right now and sat him down to do some online shopping, he wouldn’t skip a beat (although he’d be disappointed that skinny jeans are out of style). That concludes my blanket criticism — now on to the solutioning.

If a brand wanted to answer the laundry list of questions above and eventually turn those answers into a strategy, they could do it with gut feel, kneejerk competitive response, or data. Surprise: I’m going to talk about the data approach.

Specifically, the data-to-personalization trend in fashion ecommerce is the most compelling. Why? Because it’s the closest to being able to truly transform the space, and its downstream effects touch on many of the key questions today’s e-tailers are looking to answer. Personalized online shopping experiences touch search, UI, promotions, sizing, returns, pricing, and more.

The key question:

How can fashion ecommerce players collect, organize, and action data on its products and customers to deliver personalized experiences to optimize conversion?

Let’s dive in:

The Market

I won’t spend a whole lot of time here, because we all know that it’s plenty large. There are a few key insights when we look at the global trends, however:

Like I said, it’s large: The US market is estimated at ~$100B in size. The global market’s at $4.3T.

It’s growing: Globally, we’re looking at a CAGR of 9–15%. In the US, it’s projected at just 3–4%. You’ll need to drill down to the category level to identify areas of promise (like athleisure) and stagnation (like shoes) — but I’ll keep us above that for now.

It’s log-curving: as foreshadowed by shrinking growth numbers, the US fashion ecommerce market is projected to start to plateau in the next 5 years, which has significant implications:

Consolidation: A saturating market means disruptive innovation or acquisition become the new necessary strategies to sustain growth. In other words, we’ll see more roll-ups, marketplaces, and honest attempts at fundamental disruption from the e-tailers.
…And this means that the software providers who target these brands will grow in prominence as well, because centralized expertise and automation will become more of a necessary means of optimizing operational performance.
Globalization: As you might expect, while the US market reaches saturation and starts to transform, other developing markets will be going gangbusters — a few years behind us in growth trajectory. This means that more fashion ecommerce players will attempt to go global, either through acquisition or new market entry.
…And this will introduce a new set of challenges, as brands will be faced with adapting to new consumer preferences in terms of taste, shopping experience, and service level.

So What?

Recap: we’ve got a market which is giant, with long tail fragmentation, and slowing growth due to saturation. Competition is going to get fiercer and real innovation is going to start to get more essential to achieve competitive advantage both here and abroad.

I believe these conditions reinforce my thesis: that collecting, organizing, and actioning product and customer data remains the answer “on the screen”. For US competitors, delivering personalized experiences will be critical to winning the battle for business. For global entrants, understanding new customers and products quickly will be critical to making a strong start.

The Competency

What does Grant mean with all those vague ten-dollar words about data? Below is what I’m talking about, with a few examples. In short, a company is mature in this competency if it 1) collects rich data on its products 2) collects rich data on its customers, and 3) Meaningfully actions this rich data to get rich/er.

As you read, flip back-and-forth between the infographic below for a sense of the big picture.

Product Data Attributes

Crawl | Basic Product Info: This is about tagging your product data with basic, pre-coded attributes so that users can search for desired items. Even this can get pretty granular. A bulk of retail players are still here, and have been here for the past decade.
Walk | Detailed Product Data: The next level of granularity isn’t quite as empirically observable, but is equally relevant to search queries. This means mapping some of the key build components, performance features, and stylistic characteristics of the product. The use cases for this are similar to the above, but once you get into this territory, you might need some humans (or algorithms) to do some tagging of the products beyond what’s already coded in the SKU data.
Run | Psychographic Product Metadata: Now is where stuff gets interesting. Here we might start to tag the product with attributes which are much more subjective and experiential. There often aren’t clear terms for this layer of data collection — it depends on the hypotheses of the company doing the data organization. Here we might start to see products categorized by loose style associations, use cases, emotions, or personae.

Customer Data Attributes

Before I jump in, I’ll quickly note that getting customer data isn’t nearly as simple as getting product data. You’ll need to collect active browsing data via recorded sessions, cookies, accounts/membership, purchase history, and more. People aren’t as easily (or objectively) categorized as products, so a lot of customer data work is hypothesis-driven — geared toward matching customers with personae.

Crawl | Basic Customer Info: The most basic customer data can be collected passively and without customer intervention or permission. Things like where you logged in from and on what device, what you did on the site, etc.
Walk | Detailed Customer Data: Things get more invasive (with our permission, of course). Many e-tailers choose to have customers create accounts to more directly gather the juicier data morsels from them: things like demographic and biometric info, wishlists, interests. In addition to voluntary membership-based data, there’s all the cookies data: what other sites do you look at? What kind of social media presence do you have?
Run | Psychographic Customer Metadata: e-tailers can do plenty with the above, but this is where things really get interesting. I’m talking about personality and psychographic metadata derived from shopping behavior. This is not only stuff like: Grant has preppy style, but Grant is a naturally anxious person based on his click patterns, or Grant is a bold and outgoing person based on his love for bright colors.

Doing Stuff with all this Data

If you’re collecting all this data, and using all this memory and computing power to store and analyze it, you’d better be putting it to good work. This is the big hurdle: turning data into actionable insights. The resources it takes to execute this work are huge — and often only afforded by well-established fashion players who have already achieved key core operational efficiencies. Here’s a quick infographic to explain the type of insights which yield actionable personalization.

The Products

So, now we know how data can be used to personalize the shopping experience as the critical path to CX transformation. What are some companies which are enabling this for e-tailers and how are they monetizing?

Two strategies in particular are compelling. In short, you can either do this as a service for e-tailers, or do it yourself as an e-tail business model:

Strategy 1: ML/AI SaaS to enable meaningful personalization

This is the less disruptive, more immediate strategy. SaaS players with huge tagged datasets and pre-trained algorithms can plug-and-play with e-tailers. With their data, they improve search and filtering, recommendations, product display in the UI, notifications and campaigns, and SEO/SEM. It’s classically scalable and defensible, with strong centralization of expertise. It’s one of those complex technical disciplines which just makes sense to outsource. In addition to the basic stuff, compelling startups in this space are working to develop canonical models for customer psychographic profiling.

Here’s a quick look at where I see today’s market. Most of these players are relatively early stage:

The baseline expectation for these businesses is that they use visual AI to tag products, and combine this with basic customer data for UX personalization. Lily AI and Vue AI are the only two I’ve seen which use their AI to tag metadata attributes to customers to build psychographic profiles. Lily AI seems to have a larger, more well-trained dataset, which gives it a slight edge.

Strategy 2: Spotify for online shopping

This is one which I haven’t seen fully realized in the market. The idea is to have a shopping platform/aggregator which scrapes the web (or works with brand partners) for products which each specific user might resonate with. On top of that basic function, the platform might enable users to create lists, moodboards, catalogue their closet — and notify them of drops, sales, and new arrivals for brands and items which suit their taste profile/s. The user experience is therefore nearly entirely personalized for each shopper — with the intention of being the definitive portal to someone’s online shopping across brands and use cases. Path to purchase aside, it’s perhaps equally lucrative to be the definitive fashion taste repository for an individual. It’s easy to imagine how this can get social as well.

Two notable players are going at this strategy, but from different angles:

Lyst: Lyst has a lot of the characteristics I’ve described above, but it’s more of a classic shopping platform than a personalization play (and its personalization features, while a selling point, are modest). One could argue that Lyst is set up for an easy transition to the vision above — which it seems to believe in. Its landing page proudly reads: “The Definitive Fashion Shopping App”. It also has the luxury of making a data-driven transition from an already successful starting point (rather than making a big bet, or god forbid, listening to me).
The Yes: The Yes brands itself as “The Spotify for Shopping” (although it’s also a lot like Hinge). It’s a full-on bet on personalized customer experience, akin to my description above. Notably, it’s geared exclusively toward women, with higher-end products. Also notably, it requires users to take a tastes quiz to start off their journey (more on that below). All-in-all, it’s a narrower, more niche start.

If I were to oversimplify my thoughts above with primitive shapes, it would be below. Lyst has started off broad and generic in its business scope, and may hope to gradually refine its business model, while The Yes has started narrow and may hope to expand to more brands and demographics. Two legitimate stabs at the same thesis — we’ll see how they play out.