SFIX - Stitch Fix

[ajc]

So I'm long the stock, but I might pare down my position.  Namely the main reason I was long was because in addition to the GARPy nature of the stock, the fact that they have data on millions of transactions, which creates a virtuous flywheel and moat in terms of AI. 

However I recently came across
this article: www.forbes.com/sites/brookerobertsislam/2019/02/28/artificial-intelligence-software-outperforms-human-stylists-at-fashion-week/amp/
and the startup here used data scraped of Instagram etc to train their AI.  I dont know the quality of their data, but it seems like it's quite easy to get the amount of data sfix has at there fingertips by just scrapping it off Instagram, which puts a serious dent in their moat.  There seem to be quite a few fashion ai startups doing this sort of thing.  I'm curious what other people think about this in terms of the moat.

Thanks for the link. There's a few things that come to mind after reading this.

One, is that apparel on both the label and retailer side is far from a winner-take-all industry. In fact, it's highly fragmented (see image below).
In other words, if you own one of the top 5 retailers that millennials and others eventually gravitate strongly towards, then your chances of losing your principal are less than if you invested in the ride-sharing company that went up against Uber and Lyft.
Looking through that list below, you can pick out numerous businesses that were well-run, had good models, and ended up delivering very reasonable returns for many years if you invested in them shortly after the IPO.
So, I'd say keep in mind the nature of the industry and its participants.

The second thing relates to something SHDL said. Artificial intelligence right now is great at doing certain things, but not so great at other stuff.
One of the things it's bad at is to be "able to read a newspaper or do all sorts of things that will still be the provenance of human beings. Basically, anything physical, linguistic, or design-related is still going to be the provenance of humanity for the foreseeable future" (paragraph 26, https://www.theringer.com/tech/2018/11/8/18069092/chess-alphazero-alphago-go-stockfish-artificial-intelligence-future).
As SHDL noted, Stitch Fix is a business where you can "tell them to get me, say, a sport jacket that I can wear to a concert at Carnegie Hall along with a dress shirt and a pair of shoes that go well with them."
The AI demonstrated in the Forbes article might be good for a narrow range of run-of-the-mill, standardized outfits, but one of the main functions fulfilled by the human stylists at Stitch Fix is that they perform those tasks that AI can't.
Put differently, Stitch Fix stylists can interpret, gather context, and make judgement calls, all of which machines are unable to figure out with any real accuracy.
While I can see an Intelistyle option potentially being suitable for some people out there, I think it's just not going to cut it overall for the majority of online shoppers who might have specific requests, exacting standards, or complaints about the fit, texture, and so on.
For that, Intelistyle would need to hire a team of human interpreters anyway and since Stitch Fix already has that, plus the warehouses and distribution, plus the relationships with top labels, it's difficult to see how Intelistyle might end up destroying Stitch Fix by copycatting it.

This kind of leads to my last point, which is that Stitch Fix was really built around the idea of "a truly personalized shopping experience (https://newsroom.stitchfix.com/wp-content/uploads/2016/08/Stitch_Fix_AllBios.pdf)." To my mind, the stylist relationship plays an important part in that.
For me anyway I mostly shop on a functional basis, but for many people out there (women in particular, if I can make a general - and hopefully not overly offensive - estimate), clothes shopping is often a somewhat social activity and in a cold, algorithm-run world that we speedily seem to be headed towards, there's something about that which could turn out to be quite valuable even if it's not put-downable right now on a balance sheet or income statement.

Another thing worth emphasizing is that Lake has hired an "executive team from major companies such as lululemon, Netflix, Nike, Sephora and Starbucks."
If you read about her history, I think you'll likely conclude that's no accident. One thing most of those companies have in common is that they've essentially taken a commodity product and created a customer experience and aura around themselves that has allowed them, if I can use a Buffettism, to get 'share of mind.'
Making coffee, lycra, or make-up, isn't rocket science. However, I think one thing that Stitch Fix does really well is marry the technology with the personal psychology of apparel shopping in a way that really pleases clients (https://www.today.com/parents/it-was-kind-how-stitch-fix-fashions-fans-true-acts-t100208) and labels to a far-higher-than-normal degree.
To my mind that's not something to be written off easily, nor is it something where someone else can necessarily just step in and replicate it, just because they've got a great AI-related discovery.
Building a business and a brand in the way that Stitch Fix has is hard, and so while I won't write off any competitors sight-unseen, I will say that until they start posting some really impressive numbers and glowing reviews from customers and labels alike, in the way that Stitch Fix has almost continually done since 2011, I'd say the onus is more on them to prove that they're going to take everyone else's share at the press of a button, rather than the other way round.

Basically, that'd be my 2-cents worth of a rebuttal since I'm making one. Then again, I've been tracking the company, somewhat admiringly, since about 2013 so if anyone here is biased and worth ignoring, it's definitely me. Oh well, so it goes.

[cameronfen]

^ comment on the above. 

1.  While I think that retail is fragmented business model, I think delivery kit businesses like hello fresh are less that way.   This probably benefits SFIX as they are the market leader though. 

2.  Mainly regarding the AI.  I have some expertise in the area and in my opinion, lots of things that are "distinctly human" machines can do quite well.  Basically if you can quantify the cost of making a decision (ie can define a good loss function) and have very structured data (i.e. x is good, x2 is bad) machines can learn these things very well.  For example translation used to be thought of as distinctly human, but we have lots of data from the UN and so state of the art neural networks can now translate almost as good as the top humans in the languages we have lots of data in.  There is a soundcloud interview further upthread I think, where Lake admits that restricting the top 30 styles for stylists to pick instead of top 50 leads to more purchase as when stylists are confident there pick is better than the machine they are usually wrong. 

I think stylist do add something for example need a outfit for a concert which machines cant yet implement, but wouldn't be surprised if machines do more and more "distinctly human" tasks

Additionally thinking about the Forbes article a little more, it is very likely they are using a generative algorithm which can generate good styles, but if you gave it data like the person is 5'6" and likes floral patterns pick an outfit it wouldn't be able to do that, mainly because it doesnt have that data.  Would extracting the more supervised component be difficult?  I think so, but I didnt think how one could do this and there are a lot of smart people and they could think of something I'm not thinking of. 

[SHDL]

On AI, my understanding is that the current generation of algorithms are limited by the fact that they still have trouble learning something in one context (e.g., learn what type of place Carnegie Hall is by reading a NYC guidebook or something) and then applying the knowledge acquired in another (e.g., pick a clothing item that’s appropriate for someone who’s planning to visit the place).  So that would be one area where human stylists should still have an edge. 

Now whether things stay that way remains to be seen, but a good thing about this business is that by having a system where human stylists and AI algorithms work together,  they are set up in a way that should allow them to adapt flexibly to any changes in technology.  For example, if 5 years from now their AI algorithms still don’t have the “common sense” to go to a concert hall’s website to check if there are any dress codes when it needs to, then they can just have their human stylists continue to take care of that.  Or if their technology ever reaches a point where everything can be automated, they can adapt to that as well.  I like that much better than, say, someone else who might be betting their company’s future on an unproven technology that may or may not work.

[cameronfen]

Right as of right now, if you can't clearly define the objective in mathematical terms like numerical reward the algo won't know what to.  Like right now algos, can't know the search google to find a picture of what the Eiffel tower is.  Because it's unlikely you can devise a bunch of punishments and rewards to quantify the general concept of what search engines can do. 

However once you even have (lots of) data of experts doing something correctly, then even if you dont have a loss function you are golden.  But again, if I show the computer how to search for pictures of buildings by expert demonstration, it probably still can't learn how to search for flights on google flights as progress in these fields (transfer learning, domain adaptation) are not even close to that sophisticated. 

So stylists do play some role still that machines cannot.  But computers can already do many sophisticated tasks like for example they can put a box around frisbie in a picture when you give a sentence "the man through a frisbee to a dog" to the machine with the word frisbee underlined. 

All this is a moot point though in some way as stylists arent sfix competitive advantage.  Anyone can hire a bunch of stylists to pick styles.  It may build customer retention but the act of stylizing is basically a commodity input.

[ajc]

Stitch Fix reported their Q2 results yesterday evening.

- Active clients increased by 18% YoY
- Net revenue increased by 25% YoY
- Gross margin grew to 44%
- Revenue and EBITDA guidance for Q3 and FY2019 were both raised
- Free cash flow grew by 79% YoY

https://investors.stitchfix.com/static-files/9de744a5-5584-4e0d-8828-5587d1b1518f

[Castanza]

Almost pulled the trigger on this two months ago. I just can't stomach the business for some reason. Not something I feel comfortable buying into. It's expensive, not completely unique, can easily be replicated (see blue apron clones) which will eventually bring razor thin margins. And it seems like a fad that won't be sustainable if we see any type of economic slowdown. I know people who have used it and said it's not really worth it (anecdotal). I'd just rather park my money somewhere else.

That being said, I can see them being a buyout target in the future. SFIX is best in class at what they do and as mentioned above they have the most data scientists etc. Now, that doesn't necessarily mean anything.

If you had a position, congrats!

[SHDL]

Apparently Amazon is testing a me-too service:

https://www.barrons.com/articles/stitch-fix-stock-falls-amazon-prime-wardrobe-stylist-51555343917

It should be interesting to watch how this goes. 

[SHDL]

… and Walmart too, lol:

https://www.cnbc.com/2019/04/15/walmart-launches-its-first-subscription-box-for-apparel-with-kidbox.html

[Jurgis]

… and Walmart too, lol:

https://www.cnbc.com/2019/04/15/walmart-launches-its-first-subscription-box-for-apparel-with-kidbox.html

You too now can own perfect Britney Spears wardrobe.

[ajc]

Apparently Amazon is testing a me-too service:

https://www.barrons.com/articles/stitch-fix-stock-falls-amazon-prime-wardrobe-stylist-51555343917

It should be interesting to watch how this goes.

I saw that. Without wanting to sound complacent, I guess I see this as vindication more than anything.
Almost a year after they debuted Prime Wardrobe and 8 years after Stitch Fix was founded, probably the world's best e-tailer has finally admitted the Stitch Fix model is the one they need to be copying.
This after Stitch Fix has already established its dominance over other players (Trunk Club, etc) in the personalized apparel styling market.

I wrote either in my original thesis or the following pages, how I thought it would play out once Amazon got in this game.
The issues I see for Amazon are that designers and labels would much rather work with little, friendly Stitch Fix instead of the 800-pound gorilla that has a reputation for copying your products once they know how, slitting your throat, then pushing your lifeless body into the Puget Sound.

Another issue for Amazon is they have a reputation (deserved or not) as fast and cheapish. For most products that's a positive, but for fashion that's not necessarily the case. Stitch Fix has positioned its brand as more upmarket and so I think it's more likely to be thought highly of in terms of consumer options than Amazon's offering. In other words, there's a kind of snob appeal that matters. McDonalds coffee has improved a lot in recent years, but people still prefer to pay a little more for Starbucks and walk around with that cup because it has greater cache and the customer experience is higher quality.

Related to that, I kind of wonder how good the Amazon service will actually be experientially. I worked at Amazon for a few years and part of their culture is to make things as bare-bones and functional as possible. It keeps them lean and efficient. On the other hand, the place is also very functional and cheap as opposed to being classy and fashionable. I suppose I have my doubts that a place with such an ingrained culture of frugality, directness, and robustness, has the subtlety, suaveness, and aesthetic appreciation, to really beat Stitch Fix at their own game?
That's not to say Amazon won't execute very well and design something that works properly, but as I said above there's more to the success of a Starbucks than great execution. There's a certain perception of their brand they create in consumer minds, and in how the company and its people relate to you, that allows them to charge a premium and Amazon's culture has always blatantly and loudly been about the exact opposite in many ways.

The final thing about Amazon is that Stitch Fix has a massive head start. Eight years to be exact and so they have had a ton of time to make the big mistakes they needed to make, understand the strengths and weaknesses of a new industry model, etc, etc. I won't deny that Amazon is a ferocious competitor, but Lake and company are the best in their industry (which they've proven repeatedly) and very smart about their decisions and execution.
So while there's no way I'd write Amazon off, I do think that even if you assume they execute 2x as fast and well as anyone else could, that still gives Stitch Fix a four year lead.

On Walmart, I've got a ton of respect for Marc Lore. He competed head on with Amazon during his Quidsi days and if he's their next CEO then they'd have got a bargain with their Jet.com purchase. However, I'm not too phased by them for a few reasons. One is that Walmart has an even worse reputation for classiness than Amazon. The idea that some housewife from a wealthy neighborhood is going to shop via Walmart for high quality, personalized clothing is hard to believe. On the other hand, she may well be willing to go for Stitch Fix's Premium Brands offering. Also, as far as I know Lore has zero experience in apparel retail so Walmart would have a long way to go learning-wise if they wanted to have an upmarket offering there.

As I said in the beginning I think this is more a confirmation of how good the Stitch Fix model is than anything else. I'd also add that there are some very good reasons why just having a frugal, efficient executing mindset is not nearly enough in this industry and that being approached by an unfriendly megacap is not something designers are necessarily going to feel good about. I think all of that adds up to a lot of opportunity for Stitch Fix to position itself as the smarter, more experienced, and classier, alternative channel on both the customer and fashion label side. I'd be pretty happy if Stitch Fix leaned more towards the high end, in reality and in consumer's minds, and others could then still do okay on the low end.