477 signups later, here’s what surprised me about user acquisition

This breakdown highlights one of the most fundamental rules of SaaS distribution: Traffic volume and Buyer intent are rarely correlated.

Here is an analytical breakdown of why your numbers look the way they do, plus how to leverage this for your next growth push:

1. Why YouTube Outconverted Directory Links by 10x:
   Directory traffic (like Toolify or generic AI aggregator lists) consists of tire-kickers and bookmarkers. They sign up because it is free and novel, but their purchase intent is near zero. Video content (YouTube) does something a landing page screenshot cannot do: it proves real-time performance. For a cloud-based AI tool where latency and GPU rendering quality are the main objections, seeing a video demo instantly destroys customer skepticism before they even click the link.

2. The Dark Social Factor Behind Your 5 Direct Paying Users:
   Notice how Direct gave you 227 signups and 5 paying users. In AI SaaS, "Direct" is almost never pure organic brand search. It is usually dark social: someone sees a YouTube video or TikTok clip on their phone, copies the link to their desktop, or shares it in a private Discord server or Telegram group. There is a very high probability that those 5 Direct buyers were actually primed by your YouTube or demo content.

3. Actionable Next Steps to Scale YouTube and Video Channels:
   • Create Split-Screen Demo Videos: Show side-by-side video (Webcam Input vs Cloud AI Output) with a live FPS / Latency overlay.
   • Target OBS & Streaming Use Cases: Live face-swap has massive utility for VTubers, streamers, and privacy-conscious video creators. Video tutorials like "How to use cloud AI face swap in OBS Studio" carry extreme buyer intent.
   • Conduct a 3-Question Post-Purchase Survey: Email your 10 paying users and ask: "What exact project or workflow were you working on when you decided to pay?" The common denominator in their answers will give you your exact landing page headline.

Congratulations on 10 paying customers! Proving willingness-to-pay early is the hardest milestone.

10 Paying customers in a couple of weeks is a big W!

The YouTube number matches what I see on my own product — most of my traffic is YouTube-referred, and it's the only channel where people arrive already understanding what the thing does. Video pre-sells in a way a directory listing can't.

Two things I'd look at before acting on the table though.

Direct at 227 is doing a lot of work there, and it's probably not really direct. Anything with a stripped referrer lands in that bucket — apps, some browsers, anyone pasting a link into a message. If YouTube is genuinely your best-converting channel, a chunk of that 227 is likely YouTube too, which would make your case stronger than the numbers show.

The other one: same channel, different device. My YouTube traffic is overwhelmingly mobile, and I found out the hard way that my mobile page was the problem, not the traffic. The hero had no visible CTA above the fold — people landed, scrolled, left. In the numbers it looked identical to a low-intent channel. Session replay is what actually showed me; analytics never would have.

At 477 signups and 10 payers, the gap between signup and paid is probably where the money is rather than the channel mix. Do you know where people stall after they sign up?

the next useful table may be smaller than a full funnel: source, first successful face-swap, free credits used, day-seven return, and paid conversion. that would show whether YouTube is bringing buyers who understand the product, or only a few lucky conversions in a sample of nine. i would keep the source-level numbers visible, but wait to move budget until the same pattern holds across a few more cohorts. the difference between traffic quality and traffic size is a very useful lesson here.

Quick contrast from our side: 3.5 months in, protein-payment
has ~100 weekly visits, 96% direct, $0 MRR. So your "what
surprised me" question is the one I think about every day.

Two things from my (small) dataset:

1. Direct traffic dominates when you post on PH or HN -
   people who already know you come. 0% from cold search.
2. The 477 signups came from where? I'd guess a single
   distribution moment (Product Hunt, HN, indiehackers
   front page). Was the signup rate before that moment
   very different from after?

Curious if your organic vs paid split shifted after
those 477. For me the 96% direct means distribution
beats conversion every time.

Worth being careful with the biggest bucket. Direct at 227 is not really a channel, it is the drawer everything falls into when the referrer gets stripped: links pasted into WhatsApp or Slack, private windows, some mobile in-app browsers, PDFs, desktop apps, and people who saw a URL somewhere and typed it. So your largest source is mostly labelled "unknown", and if even a third of it is actually YouTube word of mouth or someone acting on an AI recommendation, the ranking you drew from this changes shape.

The YouTube number has the same problem from the other side. Two conversions is two conversions. At that count the confidence interval is so wide that 22% and 5% are both perfectly consistent with what you observed.

At 10 paying users there is a better instrument available, and it is not analytics. Ask all ten how they found you, in one sentence, by email. Self-report is noisy but at this size it beats attribution, and the mismatch between what they say and what your dashboard says is itself the finding.

The one I would watch deliberately is ChatGPT at 21 signups. That channel showing up unprompted this early is not nothing.

Have you asked the ten directly yet?

The 227 direct signups are actually what caught my attention. I’d be dying to know where those people really came from.

“Direct” always feels like a giant mystery bucket when you’re trying to figure out what’s actually working 😅

The YouTube figures exemplify how the quality of traffic can differ greatly. ~

Nine signups that lead to two paying users show you something that 200 low-intent visits do not.

Biggest traffic source might mislead one on huge opportunity but it can be preposterous of assuming so. However, when you separate sign-ups from paid customers, a totally different picture might emerge.

In my opinion, one of the more difficult aspects of early adoption is determining which statistics are actually worthy of your focus.

The split that always gets me is signup volume vs the subset that actually paid. Last-click on the signup is the easy number. Did the 10 paying users arrive on the same channel as the 477, or did a quieter channel punch above its traffic?

One timing check before reallocating anything: 477 signups is a cumulative count while 100 visits/day is a snapshot, so channel pairs like YouTube 9/2 and Google 32/1 are only comparable if both accumulated over the same launch window. The next useful number is the signup date of each of the 10 paying users. If most paid users came in the first two weeks, the real signal may be that early intent users convert, with YouTube as the visible source rather than the cause. Still worth testing the demo angle first, just not proven yet by this table.

This is a good example of why signup volume can be a pretty misleading metric. YouTube brought only 9 signups, but 2 became paying users, while some much larger channels produced none.

I’d be curious to see how this looks once you break it down by activation and revenue per visitor. The channel that brings fewer people but better-fit users can easily be the better acquisition channel.

This is the part a lot of acquisition reporting misses. A channel can look great on traffic and signup volume while contributing very little revenue.

I’d be especially interested in what happens to those numbers once you have enough data to compare the channels by activation and paid conversion rate. The YouTube result is small, but 2 paying users from 9 signups is a very different signal from 1 paying user out of 32 Google signups.

It would be interesting to see if that gap holds as the sample grows.

Someone already flagged the direct/dark-social problem, so I'll take the other end: 10 payers is too few to rank channels, but it's plenty to interview. Five conversations with those people will tell you more than the whole attribution table, because what you actually want to know is what they were trying to do the minute before they signed up. My guess is YouTube converts 2/9 because a video answers "does the latency actually hold up" before they ever hit your site, and no search result can do that. If that's true the lesson isn't "do YouTube", it's "put the proof earlier" - which you can also do with a raw unedited demo on the landing page. Worth checking whether the 467 who didn't pay ever got as far as running a swap.

The Direct: 227 line is the one I'd dig into before planning anything — "direct" is where dark social lives. Somebody's Discord message, a newsletter mention, a friend sending it to a friend. Five of your ten paying users sit behind a label that tells you nothing about where they actually came from. A free-text "how did you hear about us" at signup (or in the first email) would reattribute a chunk of it, and your best channel might already be hiding in there.

The YouTube pattern is the cleaner signal even at 2 paid: someone watched a demo before signing up, which is a different intent level than a search click. For what it's worth — someone ran an AI visibility check on my product (interview prep, very different niche) and it doesn't appear in ChatGPT's tool recommendations at all. Your 21 signups from ChatGPT suggest being quotable in those answers is already a real micro-channel for visual tools.

Small n, sure. But "channel that converts 2/9" versus "channel that looks good on a traffic dashboard" is a distinction worth keeping.

Definitely. I’ve seen the same thing high traffic doesn’t always mean high-quality users. I’d rather have 10 users from a channel that converts well than 100 who never come back.

The YouTube result is especially useful because it suggests that seeing the product in action pre-qualifies visitors before they arrive.

One thing I’d also measure is first-touch versus last-touch attribution. Some of those “Direct” customers may originally have discovered the product through YouTube, ChatGPT, or another untagged mention and returned later.

Have you asked the 10 paying users where they first discovered the product and what convinced them to pay?

This breakdown highlights one of the most fundamental rules of SaaS distribution: Traffic volume and Buyer intent are rarely correlated.

Here is an analytical breakdown of why your numbers look the way they do, plus how to leverage this for your next growth push:

1. Why YouTube Outconverted Directory Links:
   Directory traffic consists of tire-kickers and bookmarkers. They sign up because it is free and novel, but their purchase intent is near zero. Video content does something a landing page screenshot cannot do: it proves real-time performance. For a cloud-based AI tool where latency and GPU rendering quality are the main objections, seeing a video demo instantly destroys customer skepticism before they even click the link.

2. The Dark Social Factor Behind Your Direct Paying Users:
   Notice how Direct gave you 227 signups and 5 paying users. In AI SaaS, Direct is almost never pure organic brand search. It is usually dark social: someone sees a YouTube video or social clip on their phone, copies the link to their desktop, or shares it in a private Discord server or Telegram group. There is a very high probability that those Direct buyers were actually primed by your YouTube or demo content.

3. Actionable Next Steps to Scale Video Channels:

• Create Split-Screen Demo Videos: Show side-by-side video (Webcam Input vs Cloud AI Output) with a live FPS and Latency overlay.
• Target OBS & Streaming Use Cases: Live face-swap has massive utility for VTubers, streamers, and privacy-conscious video creators. Video tutorials like How to use cloud AI face swap in OBS Studio carry extreme buyer intent.
• Conduct a 3-Question Post-Purchase Survey: Email your 10 paying users and ask: What exact project or workflow were you working on when you decided to pay? The common denominator in their answers will give you your exact landing page headline.

Congratulations on 10 paying customers! Proving willingness to pay early is the hardest milestone.

This is the cleanest case study of how your measurement metric determines your distribution strategy. You're measuring two totally different things with "signups" vs "paying users," and they pointed to completely different channels.

Most founders optimize for the wrong metric early on - they see 227 direct signups and 32 GitHub signups, so they build more acquisition mechanics around the big number. But what you found is that 227 signups across conversion became 5 payers, while 9 YouTube visitors became 2 payers. YouTube is 22x more efficient by the only metric that actually mattered.

The insight that matters: YouTube didn't look valuable until you started measuring what conversion actually meant. If you kept measuring "signup volume" you'd never notice YouTube was your highest-leverage channel. It looked like noise in the volume metric.

This is why measurement clarity forces distribution clarity. When you're measuring the wrong proxy metric, you can't even see the channels that actually work. You optimized for paying customers and immediately discovered you should double down on two channels that looked worthless under the signup metric.

How much time would you have spent on paid ads or organic growth hacks if you'd kept optimizing for raw signups?

Ten paying users out of 477 is the most useful number in this post, because those ten already answered the hard question. Worth asking each of them what they were doing right before they paid and what they would have used instead. That usually surfaces one specific use case worth building the landing page around, instead of describing the tool generally. Broad traffic tends to convert badly when the page speaks to everyone. Narrow the promise to the ten, and the same 100 visits a day will behave differently.

Yes, and the reason is more useful than the pattern. Traffic and conversion diverge because the channel isn't the real variable, the visitor's intent state is. YouTube converts because someone who watched a demo already saw it work and self-qualified before clicking, they arrive believing. A search or directory click arrives curious, not convinced.

So the sharper read isn't "YouTube is a good channel," it's "seeing it work before signup converts." The channel is a proxy: did they witness the product before deciding?

Caution on "SEO + YouTube next": YouTube earned it, SEO is on the list for traffic, not paying users. Don't promote it on volume. The real question isn't "which channel," it's "which delivers people who already saw it work?"

Interesting data. This is exactly why optimizing for traffic alone can be misleading—a smaller channel with stronger intent can be far more valuable than a high-volume source that produces mostly curious visitors.

YouTube converting particularly well makes sense because users arriving after watching a product demonstration may already understand the use case and have more trust in the product before signing up.

With an AI face-swap product, I'd also be interested in seeing how trust affects conversion as you scale. Clear communication around how uploaded images are processed, retained, protected, and whether they are used for AI training could become increasingly important for turning interested users into paying customers.

I work with AI/SaaS businesses on privacy, AI governance, and compliance readiness, so that would be an interesting area to evaluate alongside your acquisition experiments. The highest-converting channel may ultimately be the one bringing users who understand both the product and trust what happens to their data. 🔐

That YouTube stat is the most interesting thing in this post — 9 visits, 2 paying customers. That's not noise, that's intent signal.

Makes sense when you think about it: someone watching a face swap demo on YouTube already understands what they're buying before they click.

Curious — was that YouTube traffic from your own video or someone else covering your product?

he next thing I would add is a time window, not another channel.

Right now 2/9 from YouTube looks much better than 5/227 from Direct, but those nine may have had more time to convert or seen you somewhere else first. I have started freezing the user count and the date of the later payment read before looking. Small cohorts can produce a convincing story whichever way you slice them afterwards.

I would keep YouTube running until either 30-50 attributed signups or a fixed date, then compare signup-to-paid and days-to-paid. With ten payments total, I would treat it as a lead, not a budget decision.

That YouTube split matches what I've seen — the lowest-volume channel often carries the most intent. 2 paying out of 9 signups is ~22% conversion vs ~2% overall. Tiny sample, but I'd chase that before scaling SEO. Curious what the YouTube videos actually were — demos or tutorials?

That's interesting, especially the YouTube numbers. 2 paying users from just 9 attributed signups is hard to ignore, even with such a small sample.

Do you know what kind of YouTube traffic that was? Your own content, reviews from other people, or something else?

don't have my own numbers to compare yet, still pre-revenue, but reading this as a preview of a trap I'll probably fall into later. the instinct to chase whatever channel brings the most signups feels almost automatic, this is a good reminder to instrument for "which channel produced someone who actually did the thing" from day one instead of retrofitting attribution after the fact once the direct-traffic bucket is already muddying everything

the YouTube pre-qualifying-buyers idea from a comment above is the one I'd want to test for myself eventually, video probably filters harder for real intent than any text-based channel simply because it takes more attention to watch than to click

Your 227 Direct → 5 paid vs 9 YouTube → 2 paid split mirrors what we see: intent channel matters more than volume after first week. We started tracking exactly this for our own devtool funnels — not just source but what they were asking when they arrived. ChatGPT 21 → 1 and GitHub 32 → 1 is still early to call, but they tend to convert slower then stick longer if they do (comparison / evaluation mode). Two practical things that helped us avoid overfitting: we now log post-signup activation events per-source before looking at paid, and we split Direct by first-touch via UTM capture in localStorage — otherwise Direct inflates. For your mix, testing a short loom / teardown that same YouTube audience sees differently than landing page visitors probably compounds that YouTube lever — same traffic quality, clearer problem → solution bridge. Also agree with Gregory's point on forward vs backward gap.

477 signups is already a lot, but looking at where the paying users actually came from is the part I liked most.

9 YouTube signups and 2 of them paid — that number surprised me too.

The YouTube data is the most interesting thing in this breakdown. 9 signups, 2 paying = 22% conversion rate. Industry average for free-to-paid is 2-5%, so something about how people find you through video creates a fundamentally different kind of user - they watched you explain the product, understood the use case, showed up pre-sold.

This is also why YouTube compounds differently than a Product Hunt spike. A video that explains a problem well keeps converting for months.

The ChatGPT referral data is worth watching as your sample grows too. People who were actively asking an AI for a solution and got pointed to you - that intent signal is high.

One question: do you know if the 2 paying YouTube users came from the same video, or different ones? That would tell you whether it's the topic or the format doing the conversion work.

so great

No real channel data yet on my end, still pre-launch. But this makes me want to set up per-channel tracking from day one instead of bolting it on later. Did you have attribution in place before launch, or did you add it after you started seeing signups come in?

Really interesting seeing YouTube bring so few signups but such a high share of paying users. Definitely shows why conversion quality matters more than raw traffic. Would be great to see an update once you hit 1,000+ signups.

This is a really good example of why raw traffic can be a misleading metric. 9 YouTube signups producing 2 paying users is way more interesting than 32 signups from a channel with zero or one conversion. I’d definitely keep tracking this before drawing conclusions though — with only 10 paying users, one or two conversions can completely change the picture. Curious whether you’re able to track what happens after the first visit too, since that could reveal even more about which channels bring users who actually stick around.

The fastest lie detector I've found for paid traffic is the device mix. My first campaign had 8% CTR at nine cents a click and I thought I'd cracked it. Then I noticed 39% of clicks came from tablets and zero from desktop, for a product people research at a desk. Those were accidental taps on game banners, not customers. Rebuilt as plain search and the mix went 85% mobile, 14% desktop, 2% tablet, which is what actual humans look like. Curious whether your 477 skew anywhere weird, it's usually the first place a channel confesses.

Worth digging one layer into that "Direct" bucket, it's often people who saw you somewhere else and typed the URL later, not organic brand recognition. Tag every off-platform mention with a short link for a month and I'd bet Direct splits into YouTube-influenced, Twitter-influenced, and word of mouth. The real signal is that YouTube converted at a far higher rate than everything else, that means the content is pre-qualifying buyers before they click, worth doubling down on before you scale spend anywhere else.

Two reads on your numbers: (1) 2/9 from YouTube looks great but n=9 — track cost per paying user per channel for another month or two before shifting budget; at small samples that ratio re-ranks everything. (2) Your 'Direct: 5 payers' bucket is probably hiding some AI-search traffic — ChatGPT visits often arrive direct and only a few surfaces get attributed, so the gap between direct and attributed is where the interesting channel actually lives. We built analytics that attributes chat referrals separately for exactly this reason (https://amami.dev).

Answering your closing question with data from my side: yes, and the gap gets wider the deeper you measure. I run free browser tools. Search brings by far the most sessions, but the metric that actually changed my decisions was the percentage of sessions that complete a full run of the tool (and channels re-rank hard on that axis). Traffic sources send visitors, but only some send people who came to do the job. One caution on your table though: 10 payers is a small enough n that one YouTube video with three buyers flips the ranking. I'd let it triple before trusting the order.

Interesting numbers. Curious what the biggest surprise was on the acquisition side — was it which channel actually converted, or how different the quality of users felt across channels?

The gap between signup volume and paid conversion is much more revealing here than the traffic numbers alone. YouTube standing out with only 9 attributed signups is a particularly interesting signal.

High-traffic ≠ high-converting is one of the most common traps in early acquisition, and your numbers show it cleanly. The channel that fills the top of the funnel is rarely the one that produces paying users — traffic builds reach, but conversion needs trust. YouTube is likely under-attributed here: people watch, forget, then come back through direct traffic weeks later, which probably inflates your Direct line more than you think. The question isn't which channel brought signups — it's which one earned the attention that made the buying decision easier. Thanks for sharing the real breakdown, most people only post the impressive numbers.