Three weeks selling: 16 real visitors, 0 customers
On July 11 I put Rubinyun on sale, an Instagram scheduler I had built for my own accounts. Three weeks later, customers: zero. This post is the bill for those three weeks: what I did, what came in, what I understood from it. I'm writing it because posts with figures have taught me more than posts with advice, and the figures that get published are almost always the good ones.
The numbers (last 30 days)
- 16 visits from external sources: 7 from Google, 7 from Indie Hackers, 1 from SaaSHub, 1 from StackShare
- 3 signups, and they are my own and family accounts, opened for testing: external ones, zero
- 0 Instagram accounts connected by anyone external
- 0 sales. The admin panel says "1 paying customer", and it's me: I bought from my own site to test the checkout
What actually happened
The most expensive week had nothing to do with the product. In late July I found out the name I was using clashed with a live registered trademark, and I had to rename everything: domain, site, email, redirects, OAuth. Two days of work, and one consequence that weighs more than the two days: the new domain is nine days old, and to Google the product might as well have been born last week. Search its name today and nothing of mine comes up yet.
Meanwhile the evenings went into distribution: I work on the product at night, so every move has to fit in one evening. Eight free directories filled in one by one, each with its surprise: the one that wants an account at least seven days old, the one that takes one tool every 24 hours, the one that rejected the listing four times because its catalog is made of web services and a desktop program does not fit, however you describe it. Then a technical post on the n8n forum, an article on Dev.to, and an official template for the n8n library, which is the channel I care most about: its pages rank on the first page of Google for exactly the searches my customer would run. The review bounced it once for overlapping notes on the canvas; fixed, re-uploaded, back in the queue.
What I learned, in order of price paid
- Directories are an investment, not a sale. On a nine-day-old domain no listing brings a customer this week: they work in months. Doing them was not the mistake; expecting them to sell right away would have been.
- The most educational bug came from testing my own template the way a user would, on a clean n8n instance: the hosting was eating the Authorization header, so every user of the template would have got a 401 on the first run. The fix was one line of .htaccess. Without that test I would have found out from the first angry customer, who at these numbers would also have been the only one.
- At these volumes the funnel says nothing about the page yet. Three signups are not a sample: first I need to bring people in, then I can read where they drop off.
What I'm doing now
I'm betting on the n8n template library, where the template is under review, and on showing up consistently in the places where people who automate already are. The directories keep working on their own, at their own pace.
If you've ever read numbers this small in someone else's post and it helped, this one is mine. In a month I'll write how it went.
As a product designer, my first instinct would be to avoid changing too many things at once. With only 16 visitors, it's still hard to know whether the bottleneck is distribution, messaging, or onboarding. I'd start by understanding who those visitors were, what they expected when they landed, and where they dropped off. That would make the next experiment much more targeted. Curious tho have you looked at session recordings or talked to any of those visitors?
I think there may be a third issue between “distribution” and “conversion”: qualified intent.
Sixteen visitors from directories and founder communities may still be sixteen people who never had an Instagram scheduling problem. Getting 100x more of the same traffic would produce a larger number, but not necessarily a clearer answer.
Before chasing volume, I’d recruit five people who currently schedule Instagram posts manually and watch what happens:
Five qualified users could teach you more than 1,000 mixed visitors.
Your three-way split is the most useful thing anyone has given us here, because it turns five conversations into a test with an answer instead of a chat.
Our own numbers say we never reached your first rung: zero external Instagram accounts connected, ever. The four accounts on the list are ours. So the first thing to watch is whether they hand the account over at all, before anything about scheduling.
We are running the five this week. One question while we line them up: does the split hold in practice, or was there a fourth bucket you kept hitting, people who connect, schedule, keep using it and still never pay because the pain only comes back once a month?
16 visitors and 0 customers is rough but its actually normal for cold traffic. whats your conversion from visitor to signup? cause if thats also low then your landing page is the problem, not the product
This is the kind of small-number post that is actually useful. At 16 visitors, I’d treat the next step as customer discovery more than funnel optimization: find 5 people who already schedule Instagram posts weekly, watch them try the n8n flow, and write down the exact words they use for the pain. That language will probably improve the site and the distribution angle at the same time.
One lightweight experiment before adding more traffic: put a “concierge setup” option beside the download. For the next ten qualified visitors, offer to install the n8n template with them on a 20-minute call, then log the exact moment they hesitate—trusting Instagram OAuth, understanding the desktop requirement, or seeing enough recurring pain. Someone who declines free setup is useful evidence; someone who accepts gives you a real activation trace. I’m in SF and happy to compare activation notes if useful—no pitch, I just appreciate these brutally honest build logs.
At this stage I would stop treating signups as the next milestone and run one paid-design-partner test. Find ten people who already schedule multiple Instagram accounts, set up the workflow with them live, and ask for a small refundable deposit. That separates an interesting tool from pain strong enough to pay. Let the n8n template generate leads, but keep the learning loop manual until you hear the same objection three times.
The product-vs-distribution question is the right one, but it's genuinely unanswerable at 16 visits — so I wouldn't burn energy there yet. The one thing 16 visits CAN teach you: did those people understand, in the first five seconds, what this is and why it's for them? That's neither distribution nor product depth — it's the top-of-page promise, and it's the cheapest thing to get wrong. Cheapest way to check without more traffic: hand it to 3 people who've never seen it, say nothing, and watch where their eyes go and where they hesitate. If they can't tell you what it does after 20 seconds, no amount of reach converts. Agree the n8n template is your best bet though — borrowed authority beats a 9-day-old domain every time.
16 real visitors in 3 weeks is brutal but honest. Most people would quit or pivot. The fact that you’re posting about it means you’re still in the game.
I’m in a similar spot — launching my product tonight on Product Hunt. 7 activations, 0 paying customers so far. The product works, but almost nobody knows it exists.
One thing I’ve learned: 16 visitors means the problem isn’t the product, it’s distribution. You can’t optimize conversion rate on 16 visitors — the sample size is noise. You need 100x more traffic before you can even tell if the landing page works.
What’s your distribution plan? Cold outreach, content, communities, paid ads? Pick one and go deep before trying the next.
Picked, and it is the n8n template library. The reasoning is narrow: our domain is nine days old and those pages already rank for what our buyer searches.
What we got from the route we did not pick: we ran the directories to the end. Eight listings, a completed G2 profile with the competitor fields filled in, and zero views on it since. Our reading is that directories do not bring people, they bring the machines that answer people, and that pays in month six rather than week four.
You launched that night on Product Hunt, a month ago now. What happened after? 7 and 16 are the same kind of number, and almost nobody publishes what comes next.
The n8n template library is the right bet I think, and since the review already bounced you once on overlapping notes, here is what actually got mine through after several rounds. The reviewers are consistent about it, and I have never seen it written down anywhere.
What passes is three distinct kinds of sticky note doing three different jobs:
Plus the layout rules: one straight execution row, no crossing connections, and no note overlapping a node or its text. Two things that cost me rounds. Sticky text is clipped to the box and does not grow to fit, so a note that reads fine while you are editing can ship with its last two lines cut off. And a single terminal node needs its note beside it, never a section background behind it, because the node covers the text.
The queue is slow enough that guessing is expensive, so I stopped submitting to find out. A template is just JSON with x/y and width/height on everything, so you can check the geometry offline before you resubmit: no note overlapping a node, no two sections overlapping, nodes sitting below their section title band. That turned a bounce-and-wait loop into one submission.
Following for the one-month update.
Thanks, this is the most useful thing anyone has said to us about that library. Ours got bounced on July 31 for overlapping notes and went through on August 3, but we changed several things at once and never knew which one did it.
The clipped text is the one that would have caught us again, since our notes read fine while editing. And checking the geometry in the JSON before submitting is the kind of thing that looks obvious only after someone says it out loud. We have two more templates ready, reels and carousels, and they go through that check before they go anywhere.
Two things we're curious about: do you check the coordinates by hand, or did you write something to do it? And did the reviewers ever come back to you about node names, or only about the notes?
Appreciate the honest numbers. Most people only post the good weeks.
The rename + new domain situation is painful but common. Directories and templates really do take months to compound — treating them as long-term assets instead of immediate traffic is the right framing.
The n8n template library bet makes sense if that’s where your ideal users already hang out. Looking forward to the follow-up in a month.
"Respect for publishing the raw numbers — the '1 paying customer is me' detail is the most honest thing I've read this week.
I'm in a similar spot with Rallynex — MVP live, getting some looks, but zero conversions yet. One thing that's helped me shift my thinking is what you said about 16 visits not being a sample size. I kept obsessing over why nobody was converting, but the real answer was simple: not enough people have seen it yet.
The n8n template bet sounds smart — borrowing someone else's domain authority is a solid move. Curious how you decided that was the right channel to bet on, rather than something like Product Hunt or a subreddit post?
Also — the 401 bug you caught by testing as a stranger is a great reminder. I've been testing my own product logged in for weeks and probably have the same blind spots. Going to run a fresh test today as a new user."
At 16 visitors you do not have a conversion problem, you have a zero-distribution problem, and directories are the worst use of scarce evenings because they send makers, not buyers. I run a social scheduling product and every early paying customer came from finding people already complaining about their current scheduler and replying individually: twenty conversations beat eight listings. Also clear the trademark before you buy the domain next time, because that rename cost you the one asset that was compounding.
You are right about the directories, and our own numbers say the same. We finished a G2 listing two days ago, profile and competitors complete. Views on it since: zero.
The part I want to steal is the method. Twenty conversations means finding people who are complaining about their scheduler today, and I would rather ask you where you were finding them than guess at it. Reddit threads, X, the support forums of the tools themselves? Whatever cost you the least time to search is the thing we do not have.
We said under another comment here that five real conversations are the job for this week, so this is a good week to be told that.
On the trademark, one correction. The clash was over the name we were selling under, not over Rubinyun: that one we did search before buying anything, and as written it is still not taken by anyone else. What the rename actually cost us is the domain age, and there you are right.
Thanks for publishing the raw numbers - the one paying customer is me detail is the most honest thing I have read this week. 16 visits is the real signal here. I do not think the problem is the product. Three weeks is roughly one content cycle - distribution simply has not started yet. When I first put my tool out, I made the same mistake: shipped, posted once, went quiet, then got confused why nobody came. What changed the numbers for me was picking two channels where my buyers actually hang out and showing up there consistently for a month before judging anything. Curious what your distribution plan is for the next three weeks - would love to read the follow-up post.
The n8n template bet is the right call for exactly the reason you named — those pages already rank for the searches your customer runs, so you're borrowing a domain that isn't nine days old. One thing I'd add: the trademark-rename story itself is a distribution asset. A fresh domain's biggest problem is zero backlinks, and "I had to rename my product three weeks after launch" is the kind of post that earns them on Dev.to/HN — you'd be aging the domain with the very incident that reset it.
Also, at 16 visits, every single visitor is precious — and "desktop app + connect your Instagram account" is a big trust ask for a first-run. A 60-second demo video on the landing page lets the visitors who won't install still leave understanding the product, which is the next best conversion you can get at this volume. Respect for publishing the real numbers — following for the one-month update.
Since you traded real numbers, I'll match the honesty: I'm at a literal zero external myself, so this isn't advice from a hill.
The n8n template is your smartest move, but there's a mismatch worth naming before you lean the whole bet on it. The people searching the n8n library are the DIY-automation crowd — the exact segment most likely to wire up their own scheduler instead of paying for Rubinyun. Ranking traffic isn't always buying traffic. So the template can't just show up; it has to leave a wound only the product closes — do the annoying 80% for free, then stop precisely where a raw workflow gets Instagram wrong (the retries, the token expiry, the silent failures). Otherwise it's admiration traffic: 500 people use it, love it, never need you.
And at 16 visits, you already said the funnel can't teach you anything yet — so the one thing that can, this week, isn't another scalable channel. It's finding five people whose scheduling actually hurts and talking to them by hand. Directories and templates are how you get found in month six, not how you learn in week four.
In a month I'll be reading the follow-up. Rooting for it.
This is the objection we needed. The DIY crowd is exactly who finds us there, and you're right that ranking traffic and buying traffic are not the same people.
Where we landed: the template does the boring part, and what it does not solve is the part that cost us real failures. Carousels were the one that got us. A raw workflow pushes all the images in one go and Instagram will not take them, because the timing is not open yet. We hit that on reels first and fixed it there, then treated carousels as plain images and hit it again. Ours now spreads it across cron passes: the first prepares, the second uploads, and a third one is there if the second does not land. It has been publishing without errors for weeks. The token has the same shape: everything works for weeks and then stops on the day nobody is looking.
The five conversations are the part we have been avoiding, and they are the cheapest thing on the list. That is this week.
That carousel story is the whole product in one anecdote. "Everything works for weeks and then stops on the day nobody is looking" — that's not a footnote, that's your headline. The template earns trust by doing the boring 80%; the line people pay to cross is the exact one you just described. Put that sentence where they can read it.
And that the five conversations are both the cheapest thing on the list and the thing you've been avoiding — yeah, that tracks. Same here. The template gets you found; the five calls are the only thing that tells you which wound to sharpen. If even one person describes the failure differently than you expect, that reshapes the whole pitch.
Do report back — I'll be reading.
the five did not happen, and the reason turned out to be worth more than the calls would have been.
i had the five picked, all of them describing the exact failure we handle. four were on the n8n forum, and that forum shuts threads down: ninety days after the last reply, and sooner if it gets marked as solved. by the time i got there all four were locked, one of them the day before. i can read their problem and i cannot answer it, and that is permanent.
so the lesson was not about the calls, it was about the window. watching for new questions is not enough on a slow forum, because the people worth talking to asked months ago and the door shuts behind them.
what did work was the other direction. instead of going out to find conversations, keep your own open. someone turned up under our own thread on that forum ten days ago with a real question, and nobody had answered him. that thread is still open for one reason, it is ours, and i answered him this morning. one live conversation from a door left open, against five from doors already shut.
the sentence is the part of your reply i kept. it came out of a real failure, three weeks of carousels going out fine and then one dying at 19:21 on a tuesday because a container had not finished processing, and nobody was watching. you are right that it belongs where people can read it rather than in a thread.
if you are running the same play on your side, the thing i would check first is whether the places you are planning to talk to people are still open.
The locked-thread thing lands hard. I'd been treating "find five people" and "they find you" as the same move — you're right that only one of them is still standing by the time you arrive.
My side runs the opposite mechanism and meets the same wall from the other face. I don't wait for the door; I knock on it — cold email to shop owners whose site actually fails the thing I check. The window's always open because I'm not queued behind ninety days of replies. The tax is that nobody asked, so the whole email has to earn the reply in two lines: the one broken thing on their page, not me.
The door-you-keep-open version, for me, is the writing — a guide that ranks answers the person the day they finally search, months after it went up. Your own thread catching someone ten days later is that same thing, just quicker.
So my first check isn't whether the place is still open — it's whether the person was ever going to answer a stranger who didn't raise their hand. Still learning which knock earns it. Pulling for the one live conversation to be the one that shows you the wound.
Your last line is the one I keep turning over: whether the person was ever going to answer a
stranger who didn't raise their hand.
I went back through our numbers this morning and they point somewhere slightly different.
Almost nobody we wrote to has answered us. What came back came from people who were reading.
Reddit, one month: three comments left in other people's threads, zero replies, view counts in
single digits. The n8n forum: two threads that weren't ours, nothing back. The one reply we
did get landed on our own thread, from a third party who had been reading it. This post, from
August 2: thirty-two comments, and every conversation still alive today started here, this one
included.
So the variable I had wrong wasn't open or closed. It was how many people walk past. A locked
thread in a room with traffic still beats an open one in an empty room, because the reply
comes from the crowd rather than from the person you addressed.
That makes your second mechanism the one our numbers support. The guide that ranks is a door
in a corridor where people are already walking. Cold email buys you something a forum never
will, since you choose who reads it, and what it costs you is the crowd.
The live conversation hasn't shown me the wound yet. Two people asked follow-up questions and
one asked for our email. When it does show me something, I'll write it here, since this is the
door that has been working.
Your rename is doing more damage than the two days suggest, and not only to Google. You filled eight directories one at a time, and directories are exactly the asset that pays out over months rather than weeks. But, any listing that went in before the rename is pointing at a name that no longer resolves, so the clock you started on those has basically restarted.
Go back through the eight and check which carry the old name, the old domain, or a redirect. The ones that took four rejections to accept you are also the least likely to let you edit a live listing quickly, and a stale entry accrues nothing while still looking like coverage.
I moved a sending domain recently for unrelated reasons, and what cost me wasn't the setup, but the fact that none of the reputation the old one had built came with it. Same shape as your nine-day-old domain.
The n8n template is the one channel where none of that applies, since the ranking comes from their pages and not yours. That's probably why it's worth more right now than the other seven combined.
You were right to make us look. We went through all eight listings one by one this week, and the old name is gone from the live ones: the links resolve where they should. The one place it survives is GitHub, where every commit in our two public repos is signed with the old address, sitting there in the history.
The rename has a longer story than it looks. The name existed as Chronon back in 2016, we added sync in 2020 because the chronon domain was taken, and in July we found out ChronoSync is a backup and sync app for Mac. So it changed in the middle of a launch.
What you said about the clock is the part we cannot argue with. The domain is new and the age we had started to build went back to zero. No fix for that one, just time.
Since you posted real numbers, here are mine as a trade. LeadGrid — local-business lead lists, mine — has been up since May. Not the same situation as yours, but the failure mode might be useful.
895 pageviews in three months, from 839 distinct people, which is a polite way of saying nobody comes back. I did the directories and the search-engine feeding too, then ran a paid validation campaign: 483 clicks across Bing and Google between 6 and 26 July. Of those 483, the number who ran a single search inside the product was zero. Not low. Zero.
The other thing the ads bought me was a referrer list full of business-finder-dot-whatever parked domains — a third of all traffic I've ever had is Bing syndication partners. For four days it looked like a spike.
That's a targeting failure on my part more than an argument against volume. I bought the wrong 483 people and got exactly what I paid for. Total real usage since launch is 91 searches from 43 people, all direct or organic.
Where it might apply to you: at 16 visits you genuinely can't tell yet, and more of them could well be the answer — the n8n template especially, since those pages get people who are already looking. I'd just be specific in advance about who you're expecting, because my numbers went up 55x without a single one of them being someone who wanted the product.
Your refusal to diagnose conversion from 16 visits is exactly right. At this size, the useful unit is not a funnel percentage; it is one complete stranger journey. I’d define a simple evidence ladder for the n8n channel: template viewed → installed → first successful run → Instagram connected → first post scheduled → payment. The 401 bug shows why “installed” alone would be misleading. If each step is observable, the next ten users could teach you more than another hundred anonymous visits, and you’ll know whether the bottleneck is discovery, setup, trust, or value. After the template is approved, can you observe those steps without adding friction or asking users to self-report?
Partly, and the split is worth naming.
Three rungs we can see in our own data without asking anyone: account connected, first post scheduled, payment.
Two we cannot: template viewed and template installed. n8n reports neither back to us, so between the library page and the first connection there is a stretch we are blind on.
Your 401 point is exactly why that gap hurts. Installed-and-broken and installed-and-abandoned look identical from here, and from here both of them look like nothing happening at all.
What we can do without friction is timestamp what we do see and stop treating the rest as measurable. If connections start appearing that we cannot attribute to anything, the template is working in a way we cannot watch, and that is still better evidence than a conversion rate calculated on 16 visits.
The restraint around not reading anything into a 16-visit funnel is good.
What made the n8n template library strong enough to deserve the next bet anyway — and what would you need to see after it goes live to decide the channel itself is working, rather than simply needing more time?
Two questions, two answers.
Why the template library deserved the next bet: those pages already rank for the searches our buyer runs, while our domain is nine days old. That age gap is the whole bet.
What would tell us the channel works rather than needing more time: one external Instagram account connected. The handover is the first step that costs the user something real, and we have never seen it from anyone outside our own four accounts. Installs and views would not settle it.
So the test is narrow on purpose. A month of template traffic with installs and zero connections means the channel found the wrong people, and waiting will not fix that. One connection means it works and the rest is volume.
That’s a useful update. I’d be interested in continuing the conversation beyond the thread — would you be open to sharing the best email to reach you on?
Happy to keep it here, and there is a reason: what we have said in this thread gets read by whoever passes by, and so far that has been worth more to us than any private thread would. The number I am waiting on is still the same one, a single external Instagram account connected. When it happens I will write it here.
That makes sense. I’ll be interested to see what happens when that first external connection comes through.
This is a good illustration of something that's becoming the real bottleneck: building the product used to be the hard, expensive part — now with AI tooling it's often the cheapest part of the whole cycle. Distribution hasn't gotten cheaper at the same rate. Your three weeks show it well: directories, forum posts, template submissions, a rename — real, sustained effort, and still zero external customers. I'd guess for a lot of builders now, GTM effort ends up costing multiples of build time, with no guarantee of payoff, which flips the old assumption that shipping was the risky part. Did you go in expecting that ratio, or did it surprise you?
No, and it was not the size of it that caught us out. It was the delay.
Building answers you the same hour: it works or it does not. Distribution answers weeks later. The G2 listing took an afternoon and has had zero views since. The n8n template took two submissions and eleven days of queue. Neither cost money, and both cost a stretch of calendar in which you keep making decisions without the answer to the previous one. You cannot iterate on a channel that replies in three weeks the way you iterate on code that replies in three seconds.
Small update since the post: two of our listings went live on PeerPush overnight and were flagged as trending this morning. First time a channel has sent anything back at all. Whether it becomes a single visitor, I will know next week.