I stopped guessing what sells on Gumroad — and built two tools from my own data

When I listed my first digital product, Gumroad showed me a price field and nothing else. No market, no benchmark, no idea whether my niche was viable or dead on arrival. So I did what you do when a dashboard is empty: I went and got the data myself.

Two problems turned up, and two tools came out of fixing them. Here's the short version — in case you're about to make the same blind guesses I did.

Problem 1: you pick a niche and a price blind

44% of Gumroad products earn exactly $0. Most of them never stood a chance, because the seller skipped the research step entirely. After I pulled live bestseller rankings for 14 keywords and analyzed a platform dataset of 146,271 products and $206M in tracked revenue, a few numbers changed how I build:

• The $30–49 price band converts ~28% better than under-$10 pricing. Cheap is not a strategy.
• Downloads outsell courses 2.5x by volume. A PDF people own beats a course people start.
• 77.3% of revenue goes to the top 1% of products. This is a winner-take-most market — which means picking the right category matters more than your marketing.
• Category economics vary wildly: Software Development averages ~$60k per product, Writing & Publishing ~$15k with only 226 competitors. Your category choice is worth more than your copy.

The trap is obvious in hindsight: most people open the editor before they open the spreadsheet. I built the Gumroad Seller Toolkit to make that research step take 15 minutes instead of a weekend — live bestseller data, the 146k-product analysis, a pricing calculator, and a 5-gate niche selector that kills bad ideas in ten minutes.

Problem 2: your product looks like a printout

So you did the research, you wrote the guide, and then you export it with print-to-PDF. It looks like a document, not a product. Buyers judge by the cover — literally. Marketplace data shows well-presented products earn multiples of identical content that looks like a bare printout.

I got tired of fighting a design tool's export dialog for three hours, so I wrote a one-command tool:

python forge.py content.md --theme indigo --kicker "YOUR BRAND" -o product.pdf

In comes your markdown; out comes a product-grade PDF — a dark branded cover with a title and promise line, then styled body pages with colored headings, callout boxes, clean lists, and proper code blocks. Four themes tuned to niches (tech, creator/AI, business, marketing), CJK font support, and the only dependency is reportlab — no paid API, no telemetry, runs fully offline. That's the PDF Product Forge.

The order that actually works

1. Research first (is the category viable? what's the price anchor?).
2. Build the product, not a printout.
3. Publish and let the data tell you what to iterate.

I turned the full research — the 146k-product analysis, the live bestseller pull, the pricing model, and the rendering workflow — into those two tools so I'd stop re-doing the work every time I ship something. If you're at step 1 right now, start with the dataset; if you're at step 2, start with the PDF tool.

Happy to answer setup or pricing questions in the comments — especially if your first product is sitting at $0 and you don't know whether it's the price or the product.