In the last episode, I talked about reframing MAKE around a simpler thesis: creators need fulfillment infrastructure and discovery, not a coordination layer. To prove that thesis, I need to launch internal products that actually win.

This episode covers how I parsed signal at scale and am using it to make decisions about what to sell.

Why? Every product I launch costs real time and real dollars. I need to ideate the concept, iterate a design for aesthetics and function, tune for printability, create marketing assets, list on the site, and run ads. Each launch is a discrete bet. Get it wrong and those hours and dollars are gone. Get it right and MAKE earns the credibility to attract creators.

I needed a way to make informed bets before spending anything.

20 Million Data Points

3D model platforms are abundant, and each one boasts millions of downloads. Across them, there's a wealth of engagement metrics: downloads, prints, remixes, comments from hobbyists who own printers and create. That's demand signal at scale. Millions of people voting with their time on what's worth printing.

But in raw form, it's useless. What engagement metrics actually matter? How do I spot the difference between a viral fluke and a category pattern? And how do I evaluate this data while accounting for the gap between a hobbyist who prints for fun and a consumer who shops online?

I built pipelines that classify the data so I can extract patterns from the noise.

Scraping

I chose four platforms and scraped data from their most downloaded models. In short, the scraping tool visits the pages for the top models from each site and parses their HTML into structured data. Across them, I captured 880 models representing around 20 million total downloads.

Classification

I built a classification system with 14 top-level facets, several with sub-dimensions, all designed specifically for commercialization decisions.

Product facets capture what the thing actually is:

• Functional purpose. What job does it do? Utilitarian, decorative, entertainment, personal, tech, or specialty.
• Aesthetic style. Primary visual style across six categories: form (minimalist, geometric, organic), era (art deco, mid-century modern), cultural (Scandinavian, Japanese kawaii), genre (steampunk, cyberpunk, fantasy), mood (cute, whimsical, elegant, industrial), and contemporary (cottagecore, brutalist). A secondary style from a different category is captured when relevant.
• Mechanism type. Static, articulated (print-in-place joints), or mechanical (gears, hinges).
• Placement. Where does it live? Mounting type (tabletop, wall-mounted, wearable, handheld) or room zone (office, kitchen, garage, gaming setup).
• Printing characteristics. Size class from tiny to extra-large, plus whether the model requires assembly or prints in one piece.
• Product category. Google Product Taxonomy path, mapping each model directly to the category structure used by Shopping ads and merchant feeds.

Consumer psychology facets capture who buys it and why:

• Age group. Children, teens, adults, or all ages.
• Emotional trigger. Primary purchase motivation: practical need, self-reward, gift giving, nostalgia, fandom, collecting, novelty delight, cozy comfort, self-expression, or impulse.
• Price tier. Under $15, $15 to $30, $30 to $60, or over $60.
• Seasonality. Evergreen or tied to a specific holiday or season.

IP and metadata determine commercial viability:

• Intellectual property. Three sub-facets: whether it's original or fan art, the parent franchise if applicable, and the franchise media type (film, video game, anime, etc.).
• License type. Whether the model can legally be sold commercially.

Research facets support ongoing analysis:

• Description. A generated catalog-style product description capturing what the item is and its buyer appeal.
• Tags. 5 to 15 research tags per model covering subjects, themes, emotions, gift contexts, and niche aesthetics.

Insights

People love cute things that move

The highest-performing models across every platform are articulated toys. Dragons whose joints ripple when you pick them up. Turtles that curl into a ball. Octopuses with tentacles that wiggle. These print as a single piece with no assembly, and they move in ways that are captivating and unique. The collecting and novelty triggers were the two highest-performing emotional facets in the entire dataset, and articulated toys dominated both.

The top performers skew toward organic, animal forms. Creatures with personality. Mechanical articulation like robots and vehicles underperforms by comparison. People want cute things they can fidget with, display, and gift.

There's a sub-category worth separating here. Many of the top articulated toys are really fidget toys. The buyer psychology is different: sensory satisfaction, stimming, stress relief. That's a distinct audience (ADHD, desk toy buyers, anxiety relief seekers) with distinct marketing angles. Lumping them in with collectibles misses the opportunity to speak directly to what actually motivates the purchase.

Anyone can design exactly what they need

There's something deeply satisfying about seeing a problem in your house and solving it on your printer by dinner. A bag clip that actually grips. A cable organizer shaped to your exact desk. A mount for that one weird thing no manufacturer has ever thought about. This is why functional categories like tech accessories, containers, tools, and mounts dominate download charts.

The designs get good fast because iteration is basically free. Someone publishes a bag clip. Someone else remixes it with a stronger hinge. A third person makes it bigger. By the time a design hits 50K downloads, it's been pressure-tested by thousands of people who actually use it. You can't riff on an injection mold. The virality comes from the satisfaction: someone prints a clever solution, posts it, and the reaction is always "I need that."

Modular systems create ecosystems

Gridfinity is a modular storage system built on a standardized grid. You print baskets, dividers, and holders that all snap into the same base plate, then configure them however you want. Multiboard is a similar concept for wall organization: a pegboard system where every hook, shelf, and holder is designed to fit the same mounting pattern. Both are open source, which means thousands of creators have contributed their own compatible pieces.

They've built massive followings in the hobbyist community, and it's easy to see why. 3D printing lets you create any shape, which means a modular system isn't limited to the six configurations that a manufacturer decided to injection mold. You design your exact drawer layout, your exact tool wall, your exact storage configuration. The system adapts to you instead of the other way around. Mass production can't touch that.

The question I want to test is whether consumers who don't own printers respond to this same appeal when MAKE handles the printing for them. The value proposition is the same: perfect customization. The delivery mechanism is just different.

Home goods show consistent demand

Planters, lighting, vases, and decorative objects all show consistent demand across platforms. The styles that perform best (minimalist, geometric, Scandinavian) read as design objects, which makes sense given that home goods live in visible spaces. That's a useful signal for which aesthetics to prioritize, not a limitation on the category.

A self-watering planter with 36K downloads and Japandi vases at 35K suggest real appetite for home items that are both functional and beautiful. This is a category I'm excited to test.

What Comes Next

My goal is to launch 1 product a day for an entire week.

Each product gets two kinds of fuel. Paid ads for fast signal. Organic content for long-term traction. They do different jobs.

Paid tells me what resonates. Five hundred dollars total across seven products, roughly seventy each. At realistic CPCs for a new brand ($2 to $2.50), that's 200 to 250 total visitors. Not enough to draw conclusions from conversions, but enough to see which products earn clicks, which creatives stop the scroll, and where intent shows up. It's a radar, not a revenue strategy.

Organic builds the foundation. For each product, I'm using AI pipelines to generate content at a pace I couldn't sustain manually: product videos, lifestyle images, short-form clips. SEO builds slowly, but it builds free. Social proof accumulates. A product that gets one organic sale this month might get ten next month without another dollar spent. Every product that works becomes a permanent traffic source.

That's the flywheel I'm trying to start. As the catalog grows, organic traffic grows with it. Each new product brings visitors who discover other products. And when creators eventually join the platform, they bring their own audiences too.

These first two weeks are about proving the demand patterns hold. If they do, I've earned the right to keep building.

I'll follow up with results.