The Complexity Makin' Goods

During a pandemic?!?!?

This week is my gamedev conference so a bit busy with manually running that and not devoting not too much time to this week’s newsletter. For those curious about assorted gamedev topics, you can see the published talks on YouTube. There’s even one from the most popular visual novel engine developer, Ren’py, live releasing a new version and talking about how py2 support will linger around for a long while because some people take many years to finish their game.

That said, I welcome you to a new rabbit hole:

I think it’d be cool if I had some swag made. Right now just to kick the tires, 1 sticker design while I figure out the whole supply chain situation.

I’m also looking for recommendations for a graphic designer to make the logo/art for the swag. Anyone who has recommendations on folk who happen to be taking commissions, especially if they use Illustrator/SVG workflows for logos, please relay them to me so I can check them out! It’ll be paid work, of course, but due to the sad state of the internet art ecosystem, I have to explicitly make that clear that it is.

This week’s post will be about a bunch of the thinking and planning that is going on in my head to even prepare for such a move. It’s like “what’s it take to make goods, preplanning edition”.

A lot of my prior experience with this comes out of the gamedev side of my career. Every so often the publisher I work with has to make physical editions of a game and I occasionally get pulled in to help out.

Remember those nice boxed packages lined up on a store shelf with a pretty printed manual and discs inside? The collectible item boom of the past decade or so continues to be a force that creators have to deal with. Plus it does feel really cool to hold a physical object you’ve contributed to in your hands.

But thanks to that experience, and my brief stint at Primary where they managed their own clothing supply chain all the way down, I have very few illusions about the complexity of a production process. So join me on a brief ride over what’s going through my brain. I promise we’ll get to a data angle further down. There’s a lot of setup involved.

Making stuff starts out “simple”

Like with all things, the 100,000ft view of things is sorta simple:

  1. Make or obtain a nice logo/design

  2. Have it printed onto some stickers and mugs and stuff

  3. Sell it on a web store

  4. Profit. Literally.

The fun starts when you zoom into the details.

Obviously, the tl;dr is that the simplest (but least flexible) route would be to use one of the Print on Demand services that exist out there. The most complicated would be to procure all the goods directly, hold inventory, and then sell on my own storefront.

But it’d be a boring world if we just ended it at the easy answer. Because looking super closely at the hard answer is MUCH more fascinating. So let’s go!

Making a nice logo/design

It’s common knowledge I have disastrous design skills, the only saving grace is I’m self-aware of the fact enough to know when to seek professional help.

So just hire a designer, you say? What’s so hard about that?

Well, even if I go and find a good designer that makes designs that I like, I still have to express to them what I want. It takes a surprising amount of thinking and creativity to even give a vague boundaries of what you’d like, while also leaving room for the designer to exercise the creativity you’re specifically paying them for.

For example, here’s an abbreviated version of what I’d like to make:

“The words ‘Counting Is Hard’ in text. Overall a tech/computer motif, with the word ‘Hard’ looking a bit goofy like a machine or computer breaking down, or it’s broken down and has been patched back together. It’ll be used for stickers and merch, so it needs to be CYMK color space for printing, and roughly square to 4:3 aspect ratio.”

If there were some more context, like if I planned on using it with certain palettes, or in specific situations, it’d be good to include those too.

Fail do this and you wind up wasting everyone’s time when the rough ideas come back and you realize you need to send it back for a retake. Thus, putting some serious thought into things beforehand helps move things along.

Making the actual stuff

We’ll get into full-service merchandise services in a bit.

Vendors matter a lot, picking one takes expertise (that we don’t have)

The market for printing common customizable objects, t-shirts, etc. is extremely competitive. A simple search for ‘silk screen t-shirt printing’ will yield endless websites offering the service.

Every vendor differ massively on every parameter you can think of, price, minimum order quantity, quality of base material, quality of the printing, even the consistency of the printing between orders. Some are giant megacorps, some are local shops in your neighborhood. They all of course claim to be of amazing quality for the price, but I’ve found that you must hold actual samples and take a high powered magnifier to everything to determine for yourself.

Some t-shirt printers use base shirts that are scratchy, stiff material that bleeds ink versus soft, well-dyed material. Offset printers, even really big printing houses, need someone to check proofs with a detailed 10x loupe to make sure all the color layers register correctly (or else you get blurry color images, death for an art book). With so many parameters, unless you know exactly how to request what you want and know how to check for it, things can turn out badly.

To make things even more complicated, all those parameters differ based on the product. A company that’s really good at making t-shirts is likely to be garbage for printing books or mugs, the processes are wildly different. For example, here’s a bunch of t-shirt printing methods, each of which can have different tiers of machines that do the work at different qualities. Meanwhile, book printing is a separate world of massive choice and terminology.

Big orders bring lots of benefits, if you can afford it

The general rule of thumb is that if you’re able to hold (or sell off) the inventory, you get the best quality/price ratio by going with the larger factories that do bigger volume jobs. Those places have the turnover to run the bigger expensive machines, with more advanced processes and quality assurance checks, to put out good product. Even for super niche boutique shops that do limited runs of things, the bigger ones will often give better results for the average project.

Many of these giant production operations for goods are now based out of various countries in Asia, and the quality and volume discounts available can often completely offset even the cost of shipping/importing goods back into the country after production.

That said, thanks to COVID-19 wrecking the global supply chain, it’s become difficult to have smallish packages of stuff shipped around at reasonable rates and timeframes, if at all. Many factories themselves are either under lockdown, or their suppliers have been locked down (like the gem faceting machine I ordered last year =( ).

GAH this is too much, is there a shortcut?

Following other people and use their suppliers!

Artists and other people in the self-publishing business often share their reviews and experiences with various suppliers. If you happen to own something customized that’s good quality, you can often look at the label and find out where it came from. While it won’t guarantee you the absolute best objective quality, it’ll at least get you a pretty good and familiar level of quality.

Okay, so about those Print on Demand services?

Let’s face it, I’m not realistically going to be selling even 50 t-shirts, mugs, notebooks, or even stickers. There’s no real minimum order level that I can sustain. I also don’t want to hold inventory in my basement or do a bunch of fulfillment. Print on demand services look super compelling in this situation.

This would be your Zazzles, your Redbubbles, etc. These companies, and their multitudes of smaller competitors, take up the niche for small creators and designers who want to make merchandise at small volumes without holding inventory or doing fulfillment.

If working with direct production factories is the IaaS (Infrastruture as a Service) offering of goods production (bare metal being you literally do the manufacturing yourself), then this would be the PaaS (Platform as a Service) level of tech. These services often let you set up a storefront on their site, put your designs up onto widgets for purchase, and when a customer actually buys a widget, the thing is actually made and shipped.

These places do charge quite a bit for their services, and the quality of their base materials can be questionable at times since they’re incentivized to cut costs there (samples are important!!), but in exchange for giving them a large share of your potential profit margin is the fact that you can avoid the most difficult problem with working in retail: predicting inventory.

OK, now we’re talking data — inventory (z _z)

Inventory prediction, or, capacity planning, is an extremely hard problem. It’s also often not thought of very often as a data problem, at least by tech-focused data scientists. The problem has been around as a general business problem since people started selling things millennia ago, and you normally can’t just throw a neural network at it. But it really is a deeply fascinating space.

Right now, we’re being asked to predict the future with no data. We can’t even make the (often wrong) assumption that past performance equals future behavior, because we’ve never sold anything yet.

This problem on the surface sounds like one of doing market research (to guess at how much could potentially be sold), and ultimately a simple business decision of how much risk are we willing to take. Even when the business starts selling stuff and you have a few months of sales data, it’s all still very primitive and a simple heuristic would do. None of that sounds very data science-y. Yup, we’re done here, everyone pack up and go home.

But the on the inventory and accounting side, there’s SO MUCH data work that needs to be done.

Imagine we ordered 100 shirts. They’re shipping on a boat to us before we sell them. How do you keep track of how much the shirts cost to make, ship, pay import duties, then subsequently ship out to the end customer? Did you make money on the sale? How would you collect this data, which often comes in the form of emailed invoices? How would you plan your future orders to make sure you don’t run out of stock, how do you account for the cost of raw materials changing with every order?

Typical companies pay HUGE sums of cash for accounting software to manage a bunch of those sundry details, plus pay EVEN MORE sums of money for ERP (enterprise resource planning) software to manage all the shipping/order tracking stuff.

Except we’re a startup, it’s not worth the trouble to pay for all that stuff. So who’s going to do this work? The accountant/finance folk often do the initial work because they follow long-established procedures for this sort of work, but putting the data into use to make decisions is often left to the data folk.

This is the painful world of scraping PDF invoices, ugly-formatted Excel, bizarre data APIs from 3rd party accounting software that make no sense but cost you an arm to even access. This is the world you’re daring to poke your nose into if you decide “yeah, I think I can handle my own inventory.”


About this newsletter

I’m Randy Au, currently a quantitative UX researcher, former data analyst, and general-purpose data and tech nerd. The Counting Stuff newsletter is a weekly data/tech blog about the less-than-sexy aspects about data science, UX research and tech. With occasional excursions into other fun topics.

Comments and questions are always welcome, they often give me inspiration for new posts. Tweet me. Always feel free to share these free newsletter posts with others.

All photos/drawings used are taken/created by Randy unless otherwise noted.