A week after my "failed niche site" post: it actually brought a real inbound

Yep. And that gives you a much cleaner metric than “did they like the test?”

I’d track the exact transition from test → production conversation.

If nobody asks about the full site, the free test gave you information but didn’t validate willingness to pay. If they do, you’ve got a real commercial signal.

I’m becoming slightly obsessed with separating those two because solo founders can spend weeks collecting positive feedback that never turns into revenue.

Yep. And that gives you a much cleaner metric than “did they like the test?”

I’d track the exact transition from test → production conversation.

If nobody asks about the full site, the free test gave you information but didn’t validate willingness to pay. If they do, you’ve got a real commercial signal.

I’m becoming slightly obsessed with separating those two because solo founders can spend weeks collecting positive feedback that never turns into revenue.

Congrats — the "product wasn't the problem, distribution was" reframe is the hardest one to internalize on your own, so credit to whoever gave you that comment.

Two things from running French-market SEO for a few brands, in case they're useful for the production version:

hreflang is where most multilingual sites quietly break, and clients never notice until a Sitebulb crawl. The three failures I see over and over: (1) x-default pointing at the wrong locale, (2) return-tag reciprocity missing on one side, (3) hreflang set correctly but the canonical still pointing at the EN version. Any one of those and Google merges the localized page back into the EN one and you lose the whole point. If your engine already outputs the hreflang block, add a validator that catches those three and you're ahead of most agencies.

"Translation quality" is a proxy — the real metric is search intent match by locale. The French phrase for "how to X" is often not the direct translation of the English one; the locale-idiomatic keyword is different. Sentence-level translation accuracy can be perfect while the page still ranks nowhere, because the H1 doesn't match how French speakers actually search. A quick test: pick a target page, look at what French SERPs return for the intent, compare to your translated title/H1. If they don't line up, no amount of clean translation will rank.

Your inbound is exactly the shape that matters — an existing SaaS with EN/DE/FR already live and a concrete reason to add ES/PT. That's a paying customer, not a curious one. Push hard.

Completely agree with this perspective. Keeping things simple early on really helps avoid over-engineering. Thanks for sharing!

Congrats on the inbound—building in public and sharing the post-mortem on the failed site is usually where the actual market validation hides.

The package you put together for them (especially the hreflang snippet and meta comparison alongside the translated text) is spot on. A lot of developers/founders miss that translation isn't just about localized strings; if the hreflang tags and metadata parity aren't cleanly mapped out from day one, you end up splitting crawl budget and running into duplicate content issues across regions.

Curious how you're planning to handle dynamic content or future updates for them once it moves from a static test page to production? Keeping those localized trees synced without breaking the core theme layout can get tricky fast at scale

This is a really interesting outcome. As a web developer, I especially like the part where you kept the original branding, theme, and layout intact and changed only the content.

I think localization is one of those things that looks simple from the outside, but keeping the translated pages technically and visually consistent with the original site can get tricky — especially with hreflang, metadata, URLs, and QA.

Also, “measurement quality > distribution” is probably the biggest takeaway here. One genuine SaaS lead seems far more valuable than thousands of bot impressions.

Good luck with the review curious to see whether this turns into a production use case.

Congratulations, and the thing that actually made this work is worth naming: you sent artefacts, not a pitch. Localized pages plus a hreflang snippet plus a meta comparison plus QA notes is a deliverable somebody can evaluate in ten minutes. "Want to try my engine" is a request for their time.

Since you are now shipping hreflang into other people's sites, one detail bites almost everyone the first time and it fails silently: hreflang annotations have to be reciprocal. If your Spanish page declares an alternate pointing at the English page, the English page must declare one pointing back at the Spanish page. When the return tag is missing, Google can ignore the whole cluster. No error, no warning, no ranking change you can trace. It just quietly does nothing.

Three more with the same silent failure mode:

1. The value is a language, optionally a dash and a region: es, es-ES, pt-BR. Underscores (es_ES) are invalid and ignored.
2. Region alone is not valid. es-419 for Latin America works, 419 on its own does not.
3. Include x-default in every set, pointing at the page a visitor with no matching language should land on. Without it, that visitor is routed by a guess.

The 60-second check on any page you hand over: view source, list every hreflang alternate on it, then open one of the pages it names and confirm the same set appears there too, self-reference included. If any page in the set does not point back at all the others, the set is not doing anything.

For sanity-checking a handover before it goes out, NexusBro runs a free audit at nexusbro.com/audit that covers the SEO and technical layers in one pass, no signup.

The contrast between impressions and one real inbound is telling. I’d use the next few conversations to learn the exact trigger that made this company reach out, then repeat that path deliberately.

This is the point where I’d write down the conversion path before their review lands: what they’re evaluating, who approves production, what result counts as success, and what paid scope follows. Which event turns this from a useful free test into a purchase?

This is a great reminder that a “failed” product can still reveal a valuable wedge. The translation engine becoming the real signal is a useful lesson in watching for repeated user pull instead of defending the original thesis. I’d keep documenting which use cases convert that inbound curiosity into a concrete next step.

The inbound is a much stronger signal than the original site metrics, but the important evidence may come from what they do after reviewing the output. What would you need to see from this test to conclude there’s real production demand rather than just curiosity?