Month Three: I Measured the Engine I Was Feeding
135 subscribers, a notes engine I caught converting at 0.13%, and a reel kit that finally has a launch date.
Hi, I’m Daniel — a software engineer, solo builder, and the person behind Digital Craft Workshop.
I’ve spent ten years writing TypeScript, .NET, and React for a living. The last few of them I’ve been building on the side of the day job: B2B SaaS, AI tools like Reel Pipeline Kit and Drippery, internal infrastructure, and a Unity 2D narrative game. Some shipped. Some I killed. Both are useful.
Digital Craft Workshop is for developers, AI builders, and indie founders who’d rather understand the craft than ride the hype. The kind of reader who distrusts gurus and wants AI that’s cheap, practical, and accountable to one engineer in a workshop.
Each essay is something you can read in one sitting and use the same week: an architecture pattern, an AI workflow a single developer can actually run, a build log, or a postmortem on a project that didn’t make it.
Everything here, I actually ran — including the parts that broke.
TLDR
Month two I stopped pushing my tools. Month three I stopped trusting them and measured instead.
Subscribers went 74 → 135. Same +82% as May, but nearly double the absolute jump, still on $0 ads.
I finally pulled the stats on my Notes engine and didn't like what I found. So I rebuilt it, launched a second account to test my reels in public, and gave the Reel Pipeline Kit a launch date: July 7.
For two months I fed a Notes engine and assumed it was working. In June I finally checked the gauge. It converts at 0.13%.
Subscribers went 74 → 135 anyway, and almost none of it came from the engine I was pushing.
A month ago I wrote Month Two: I Stopped Pushing the Tools. The lever that surprised me there was recommendation swaps, quietly out-growing the Notes engine I kept feeding. I ended that post promising to cut Notes from ten a day to four.
I didn't. Before cutting anything, I looked at the data.
The numbers
The percentage held from last month. The absolute nearly doubled.
I'd assumed the +34 in May was roughly the ceiling for a month on free channels. June cleared it without me changing much, and I still can't fully explain why.
The 0.13%
I built a system called grownote to draft and schedule my Notes in my own voice. By June it had posted 401 of them.
So I wired up Substack's note-stats API and asked it two things. How many of those 401 Notes had ever produced a single subscriber? And across every impression they earned, what was the actual conversion rate?
The first answer was seven. Seven Notes out of 401 had ever converted anyone. The second answer was 0.13%, subscribers per impression across everything the engine had posted.
I'd poured two months of evenings into that engine. It shipped Notes on time, in my voice, with the angles I asked for, and still brought no one home.
I wrote the full breakdown up on its own, numbers and all, because if I'm building in public the bad rate belongs out here too.
I rebuilt the notes engine around conversion
The old pipeline planned by volume and variety. The new one plans three Notes a day, each mapped to an archetype the data actually rewarded:
Every Note now has to open on something concrete. A real number, or a specific result. No mood-setting, no question standing in for a hook.
It's a smaller engine, and it respects what the 0.13% told me. I don't have a clean before-and-after number yet, since June was its first month running.
But the +15 week landed while it was live, and three of those subscribers came in tagged to a Note. That's the first time Notes have shown up in my attribution at all.
Where the growth actually came from
Same honest answer as last month, with better data behind it.
I scrubbed the source field across roughly 50 days of new subscribers. Notes are the single biggest named source, at about a third of conversions.
The bigger bucket, around 45%, is Substack's own discovery machinery: direct-to-app, signup flow, reader-profile discovery. People find me on the platform itself, not through anything I posted that day.
So Notes drive about a third of my conversions and roughly two-thirds of everything the Substack ecosystem sends me. That's the finding that made the rebuild worth doing instead of the cut I'd planned.
The channel was never the problem. It was underperforming what it could do, and now I know by how much.
I've watched this platform effect before. A year ago a single embed quietly turned a Medium reader into a Substack subscriber while I wasn't looking. The plumbing does a lot of the work if you give it something to point at.
I launched a second account to test my reels in public
The reel pipeline needed somewhere to prove itself that wasn't my own newsletter. So I started Reads in Motion, a curation account that turns other writers' best posts into short reels.
Every Monday a routine builds a Top 10 of the week's most-engaging reels, ranked by real interactions, and drafts a leaderboard post with a rendered video. I approve and publish each one by hand.
It does two jobs at once. It's a live stress test of the exact pipeline I'm about to sell, and a distribution surface that grows on its own engagement, not on mine.
Watching which reels actually pull interactions there has taught me more than a month of theory would have.
The Reel Pipeline Kit finally has a date
Last month I said I'd finish the kit and ship the sales page. That one I kept.
The Reel Pipeline Kit is a local app. You paste a Substack URL and get a 9:16 MP4 back. Your keys, your machine, nothing routing through a server of mine.
The six-post tutorial series behind it is done and published, from Haiku writing the scenario to the GPU clip stage.
The catalog page is live. Two pre-launch sales came in before I'd properly announced. The official launch is July 7, with a Pro tier in progress behind it.
Two sales isn't a business. But it's the first money this workshop has earned from a product I built to sell, and it arrived before launch. That's the signal I wanted.
What I said I'd do, and what happened
Keeping myself honest against last month's list:
What I got wrong
I spent April and May treating the Notes engine as a black box that was obviously working, because it was obviously running. It took me two full months to point the stats API at it.
The 0.13% was knowable on day one. I didn't look because looking risked finding out. I'd shipped the engine and never once checked whether it converted, and those turn out to be two very different jobs.
What month four looks like
74 → 135 in a month, still on nothing but free channels. The chart is steeper than last month's, and this time I know which lever I'm pulling.
That's month three. Workshop's still open.
— Daniel
I ignored my note conversion rate for two months because I assumed it was fine. It was 0.13%. Restack this if there's a number in your own work you've been avoiding, you don't have to say which one.
P.S. The Reel Pipeline Kit launches July 7, the whole engine behind these reels running on your own laptop with no subscription. If you want to see what it produces first, the build series is I Taught My Articles to Make Their Own Reels.
