I Ranked the Four Things That Make a Digital Product Sell. Product Came Last. Here Is the Bill.

A few days ago a reader, Anna Parker, asked me a question under an older piece of mine — a timeline of selling digital products from zero: what actually changed between the months when nothing sold and the moment things started to move — was it the product, or something upstream? I answered in a comment, ranked the four things in an order that surprised even me, and promised to write it out properly. One honest note before the numbers: the May piece was written as a composite timeline before I started publishing the actual ledger. Everything below is the ledger.

This is that piece. It is not a list of principles. Anyone can write principles. It is the bill — what each item cost me when I did them in the wrong order, in days, in dollars, and in one Etsy strike. I am an avionics engineer by training; when a system fails, I want the fault tree, not the moral. So here is the fault tree of my own business.

The short version of the ranking: positioning first, distribution second, audience third, product last. The long version is what the wrong order looks like from the inside, because from the inside it looks like progress.

The ledger, so you can check my arithmetic

Everything below comes from a system I built to run a digital-products business without me: about 670 Python modules, a crontab that has grown past a hundred lines, four storefronts. It started on April 28, 2026. As of this morning, September 13:

• 196 articles published to my own site, one a day, by a scheduler.
• 52 listings across Etsy, Gumroad and my own store — roughly 30 distinct products, several listed twice.
• Six paid sales. Five on Etsy, one on Gumroad. Total revenue: about $158 US. The Gumroad sale was 99 cents.
• 788 views across all fourteen Etsy listings in the last ninety days. Twenty people saved something to favourites.
• One product bought more than once. On its own it is over half of everything I have ever earned; counting the bundle it sits inside, nearly all of it.

The machine did its job every day. That is exactly the problem, and it is why the order matters.

Line item 1 — Product first: 138 days, ~30 products, $158

I did what people who like building do. I built. The generator wrote a product brief, a PDF, a set of listing images, a description; I reviewed, polished, listed. When one product did not sell, the obvious next step was another product. Notion templates for freelancers, for developers, for students. A planner. A clip-art bundle. A workflow guide. Then a “premium” version of the workflow guide. Then a bundle of the premium versions.

Here is what that looked like on the dashboard: a catalog that grew every week. Here is what it looked like in the bank: nothing. Twenty-odd products earned zero dollars each. Not “low sales” — zero. Fourteen listings on Etsy currently share 788 views over three months, which is fewer than sixty views per listing, and a handful of favourites. My most-polished Notion templates sit at 35 to 58 views each.

The cost: most of five months. Not because any single product took long — the machine made them fast — but because each one bought me another two weeks of believing the next one would be different. The catalog was a very convincing way to postpone the question the catalog could not answer: who is looking for this, and where are they standing?

There is a second, sharper cost. A marketplace with thirteen listings is thirteen surfaces for a policy mistake. One of mine was removed by Etsy for a single word in a category they had quietly reclassified. Cleaning that word out of twelve descriptions and five images, and sitting through a self-imposed freeze on edits so the shop would not look like it was thrashing, cost me a further two weeks in which I could touch nothing.

Line item 2 — Distribution assumed: 196 articles, ~12 visitors

I did know that “build it and they will come” is a lie. So I built distribution too — as software. An article a day, cross-posted; scheduled posts to four regions on X; automatic re-promotion of listings; Pinterest, LinkedIn, Facebook. On paper, a small newsroom.

The audit I published two weeks ago already showed how badly the instruments lied. What I want to add is what distribution-by-cron actually delivered: the page that matters most for conversion received about ten to twelve real human sessions per fifteen days. Not per day. Per fifteen days. The articles were indexed, they were even competent, and they brought almost nobody who wanted to buy.

Meanwhile the channels failed the way automation fails — silently. My X poster refused, for weeks, to send any post containing a link, because a budget rule I had written myself scored link posts as too expensive and simply dropped them; the daily log said “posted”, and the content went nowhere. LinkedIn has been returning a permissions error since September 4; nothing alerted me. Three days ago the API credits behind the whole writing pipeline ran out, the blog stopped for two days, and I found out from a log, not from a notification.

The cost: distribution is not a cron job. A cron job can deliver distribution once you know where the people are. I had it running before I knew, so it delivered content to nobody, on time, every day, and reported success. That is worse than not having it, because “success” is exactly what I was checking for.

Line item 3 — Audience deferred: one subscriber in July, three in August

Everything in the previous two items would have been survivable if there had been a list — a few hundred people who had raised a hand once. There was not. In July the business had one email subscriber. In August, three. The free product that was supposed to build the list sat in a catalog nobody visited, so it built nothing.

What did build an audience, slowly, was the one thing I did by hand: writing here, under my own name, about what was actually happening, including the parts that made me look foolish. The piece about the $54 got more real conversations than 149 automated articles combined. This week Medium reports ten new followers and nine new subscribers, and paid me $9.88 — which, I am aware, is a comic number, and also the only channel in this entire operation that is growing rather than merely running.

The cost: I treated audience as a by-product of product and distribution. It is the other way round. The audience is the thing that tells you what to make and where to put it; without one, every product decision is a guess dressed as a plan. Five months of guesses.

Line item 4 — Positioning last: the one thing nobody else had

This is the line that changed the ranking for me, because I found it by accident and could have found it on day one.

Of about thirty products, exactly one is bought repeatedly: a multi-agent starter in Python and FastAPI, with the code and a long guide, aimed at developers. It has more views than any two of my Notion templates together, three of my six sales outright and a fourth as part of a bundle. I did not position it cleverly. It sells because when a developer types those words into a marketplace search, it is the only thing there. Every Notion template I made competes with a free gallery of thousands; every planner competes with ten years of planners. The code kit competes with nothing, so a hundred and fifty people found it without me doing anything.

Positioning is not copywriting. It is choosing the shelf where you are the only item. I had that shelf from June and spent the summer stocking the crowded ones instead, because the crowded shelves were where “digital products” were supposed to be sold.

The cost: every product built after June that was not on that shelf. Which is most of them.

Why the order is what it is

Put the bill together and the ranking is not a philosophy; it is just the direction of dependence.

You cannot know what product to build until an audience tells you. You cannot gather an audience until something you made reaches strangers, which is distribution. And distribution to strangers only works if what arrives is instantly legible as the thing they were looking for and could not find elsewhere — which is positioning. Run it backwards, as I did, and each stage produces impressive activity that the next stage cannot use: products nobody asked for, delivered by channels pointed at nobody, to a list of one.

Run it forwards and the same machine becomes useful. I already have the code, the schedulers, the publishing. What I did not have was the shelf, the people standing in front of it, and the patience to let their downloads tell me what to build next. That patience is the least satisfying part to write about, which is probably why it is missing from most “how I did it” posts: the early numbers are too small to mean anything, and the only honest strategy is to keep the sample growing until they do.

What I am doing now, with the real order

I have stopped adding products. The catalog is frozen; the automated product factory is switched off at the source. Positioning first: the developer kit goes to the front of every storefront, and I am writing for the people who bought it — about deploying agents, about what they cost to run, about the quiet failures — because that is the shelf where I am alone. Distribution second: one channel, this one, by hand, until it reaches a hundred of the right people a week; the cron jobs get to amplify only what has already worked once manually. Audience third: the free thing goes where those readers already are, not in a catalog. Product last, and only when a buyer asks.

I will report the numbers either way. That is the deal I made with myself in July, and it is the only reason the rest of this is worth reading.

If you build agents and want to see the one thing people actually bought, it is here. If you are earlier than that — still building the catalog — I would rather you read this twice than buy anything.

Written by Andrii Klymenko — Researcher & Writer. Exploring the human side of science. One micro-shift at a time.

☕ Want to support my work or explore more?
→ andriiklymenko.carrd.co
Or tip $1 instantly via PayPal: paypal.me/AndriiKlymenko400/1

📚 Related Articles

• Solopreneur AI Business: Build Revenue With No-Code in 2026
• Claude Passive Income: 3-Step System for Digital Products
• 3 Best AI Tools for Automation & Digital Products 2026
• I Automated My Business, Then Audited It. The $12.99 Sale Never Happened, and the 71% Growth Was a Bot Farm.