Ten activations, ten demos, zero sales: what publishing an EA on the Market actually taught me
I published VitalEdge Gold EA on the MQL5 Market on 29 September 2026. As of the middle of November it has 10 activations, 10 demo downloads, no reviews and no sales. This is not a story about a launch that went wrong. It is a list of the things I got wrong in the order I got them wrong, because the Market's own documentation tells you how to publish and almost nothing about what happens afterwards.
If you are about to upload your first product, this will save you a month.
The listing is not the marketing
I spent three weeks on the robot and four days on the listing. That ratio was backwards.
The Market gives you a title, a short description, screenshots, a version history and a price. None of those things bring anyone to the page. MQL5 search does, and it is driven by the words in your description and the activity around your product. Ten demo downloads in seven weeks is not a demand problem, it is a visibility problem: almost nobody is arriving at the page to be persuaded either way.
What actually generates traffic on this site, in order of how much work it costs:
1. Articles and blog posts that get read. They surface in the site's own feeds and in Google.
2. Forum answers where the question is the same problem your product solves.
3. A CodeBase entry that is genuinely useful on its own and points back to the product.
4. Reviews and comments, which the Market weights because they signal a live product.
I did none of these for the first six weeks. Do them from day one, not after the sales fail to arrive.
The demo count is the only number that means anything early
Activations include your own test installs and anyone who clicked buy with a trial. Demo downloads are closer to real interest: somebody went to the trouble of attaching the product to a chart and running it.
Watch the ratio. Ten demos and zero sales over seven weeks tells you the page is not converting the small number of people who do arrive. That is a different problem from the traffic problem above, and it has a different fix: the screenshots, the description, and the honesty of the backtest.
Write the description for the person deciding in ninety seconds
Mine opened with the strategy logic. Wrong. The person reading it wants four things before they will read anything else:
- What instrument and timeframe, in the first line.
- What the drawdown looks like in the worst month, not the best one.
- Which prop firm rules it respects, because half the Market's buyers are evaluating for prop accounts now.
- What happens on a news event, and whether it holds over the weekend.
Prop-firm style rules deserve their own paragraph. If your EA has a daily loss cap, an equity stop and a weekend flat option, say so in the first screen of the description. Those three features are the difference between a product a funded trader can use and one they cannot, and almost no listing mentions them.
Versioning is a feature, not a chore
I am on version 3.50 and every one of those increments was a real bug fix found by someone else running it on a different broker. Two lessons:
- Test on at least three brokers with different spread and swap models before you publish, not after the first complaint.
- Keep the version history in the description and describe what changed in plain words. Buyers read it to judge whether you are still working on the product. A listing stuck on 1.00 for a year reads as abandoned, whatever the truth is.
Pricing and rental, honestly
I listed at 39 USD to buy, with rentals at 30, 34, 36 and 38 for one, three, six and twelve months. MQL5 takes 20 percent and the minimum price is 30 USD, and you can only change the price once a day, so decide deliberately rather than fiddling.
The rental ladder is the part I would do differently. Rentals convert far better than purchases for an unproven product, because the buyer is risking one month instead of the full price. If I were starting again I would set the monthly rental low enough that it is an easy yes and let the purchase price sit where the maths on a year of rental makes it look sensible.
What the robot itself had to survive
The strategy logic was the easy part. The parts that nearly broke it:
- Daily loss accounting. If your EA tracks a daily loss limit, the reset is a server time, not the user's midnight. I had to make the reset hour configurable and document it, because a 5 percent daily cap measured against the wrong snapshot is worse than no cap at all.
- Spread spikes at the rollover. A fixed stop in points is not a fixed risk in currency. Size from money, not from points.
- Restart safety. Terminals get restarted, VPS instances get migrated. If your state lives only in memory, the EA forgets its daily loss total and starts again. Persist it.
- Partial fills and requotes. On gold, in the first minute of a session, both happen. Handle them explicitly or the position size will not match what you intended.
The uncomfortable conclusion
A product on the Market is not a passive asset. It is a listing that decays without attention, in a catalogue where the median product sells nothing. That is not pessimism, it is arithmetic: there are thousands of products and a finite number of buyers, and the ones who sell have a visible reason for a stranger to trust them, usually because they wrote something useful first.
So the plan from here is not a better robot. It is writing, answering questions where my own EA had the same problem, and letting people find the product through the work rather than through the listing.
If you are building one, the order that works is: write about the problem, publish something useful in the CodeBase, answer forum threads, and only then put a price on it. I did it in the opposite order and I have the numbers to prove it.
VitalEdge Gold EA trades XAUUSD and XSIUSD on MetaTrader 5 with configurable daily loss caps and a weekend flat option. Product page: https://www.mql5.com/en/market/product/198351


