TejeLab: I built a site from scratch to become my own SEO and UX client

View contents

In fifth grade my mom taught me to knit and an aunt taught me the chain stitch and the double crochet. I understood none of it. My first project came out terrible.

Twenty-odd years later I still make things, mostly crochet, which is the part I like best. I have made more scarves than anything else. I love starting projects and putting them away, watching knitting videos, buying yarn. The only part I dislike is frogging, which is of course what I end up doing most.

On 19 August 2026 I launched tejelab.com, a site of calculation tools for crocheters in Chile, Spain and Latin America. I built it with Claude, from scratch. This post is about that, but mostly it is about something more uncomfortable: what happens when the consultant and the client are the same person.

Why a knitting site and not something else

In consulting I have spent years recommending things that someone else then has to approve, prioritise, fit into a sprint and defend to a stakeholder. That is part of the job and I am not complaining: that negotiation usually improves the recommendation.

But it also means I almost never see a recommendation implemented exactly as I designed it, or the full result from day zero. When you arrive at an existing site, you inherit decisions. A site that starts with you is a different thing: every choice about architecture, URLs, content and structured data is yours, and so is the blame when something breaks.

I chose knitting because it is the subject where I have my own judgement as a user, not only as a professional. I can look at a yarn substitution calculator and know whether it works, because I would need it myself.

What TejeLab is today

As of 1 September it is 66 prerendered static pages and 144 commits in thirteen days:

• Eleven tools: pattern grading by size, yarn substitution across brands, how much yarn a project needs, how much to charge for handmade work, a tapestry crochet chart generator, a voice-controlled row counter. Not all of them are calculators: there is a tutorial library, a colour analysis test and a printable measurements sheet.
• A reference for 32 yarns from brands actually sold in Latin America and Spain, not the ones that appear in US patterns.
• A stitch glossary with each country's terminology.

That last part is the one that most resembles my day job. A "punto alto" in Chile is a "punto vareta" in Mexico and a "double crochet" in English. The decision was to keep a single source of terminology in the code, and to ship every regional variant in the static HTML even though the interface shows only one. I did not decide that for SEO. I decided it because a Colombian crocheter arriving with her own word has to find the page, and then read it in her word.

Being my own client

What I did on TejeLab is, almost point for point, what I recommend to a client: prerendering to static HTML so the content exists without JavaScript, structured data pointing at the right canonical, design tokens instead of hard-coded colours, an accessibility floor of AA contrast and 44px touch targets, and a rule that no edit counts as done until you have seen it in the built HTML.

That last rule exists because I broke it several times in the first week. I moved the glossary to a new route and left two dead links in the header and the footer. The build log said "14 entries" while the file had 20. None of it was visible in the code: it was visible in dist/.

It is exactly the same error I audit on other people's sites, committed by me, with full access to the repository and nobody else to blame. I am telling it because I think saying it is worth more than the version where everything came out clean.

The first twelve days, with data

These are the Search Console numbers from 19 to 30 August, unfiltered and unretouched:

Impressions	156
Clicks	2
Average position	around 10
Countries with impressions	17
Mobile / desktop	99 / 54 impressions

The first impression arrived on 20 August, one day after launch. The first click arrived on the 25th, six days after launch and five after that first impression. The second arrived on the 30th. Chile contributed 48 impressions and Spain 35, one click each; then come Mexico, Argentina, Colombia, Uruguay, Ecuador, Peru and a dozen more with single-digit numbers.

That first figure is the one that matters to me, and I want to separate it carefully, because it is easy to read backwards.

Two clicks are not traffic. With 156 impressions there is nothing statistically solid here, and anything I say about user behaviour is a hypothesis, not a finding. I am saying it before anyone has to say it to me.

But two clicks are evidence of something else. tejelab.com is a twelve-day-old domain. Zero backlinks, zero history, zero authority, zero brand: nobody was searching for it by name. Even so, there were impressions the day after launch, meaning Google indexed it in under twenty-four hours, and within twelve days it sat at an average position of 10 — page one — across several long-tail queries, with visibility in 17 countries. Someone saw it there and clicked.

In my experience, that is not the normal course of things. I have seen new sites with considerably larger budgets spend weeks stuck in "Discovered, currently not indexed" before showing up anywhere at all. For me that is the milestone of these twelve days, more than the clicks.

And I do not think it was luck. I think it comes down to four decisions taken before the first deploy: prerendered static HTML, so there is nothing waiting to render; a flat architecture, with every tool one click from the home page; a correct sitemap from the initial commit; and content that answers concrete questions almost nobody is answering in Spanish. None of the four is a trick. They are exactly the ones I recommend in consulting, and the ones that are hardest to get prioritised, because you cannot see them.

That said, twelve days of data cannot prove causation. I can say what I did and what happened afterwards, which is not the same thing.

What surprised me anyway

With that distinction above, three things I did not expect.

The page with the most impressions is the one ranking worst. The row counter has 36 impressions and zero clicks, with an average position of 22. But that average mixes queries: for "contador de vueltas crochet online" it shows at position 9, and for plain "contador de vueltas crochet", at 54. A good reminder that a page's average position is not a position, it is an average of very different situations.

The first click did not come from the most-searched tool. It came from the pricing calculator, which has 28 impressions at position 7. The tool with the most demand is not the one that converted first.

The query I find most interesting is one I did not design for. Someone searched "cuánto cobra una tejedora por hora" — how much a knitter charges per hour. We show at position 8, with three impressions. That query is not about crochet: it is about what a knitter's handmade work is worth, in a market that pays it badly. It is the question I would most like to research properly, and I did not see it coming.

What I still do not know

I do not know whether the site will grow or settle at this plateau. I do not know whether the tools solve the problem crocheters actually have or the one I think they have, because so far I have zero conversations with users and only search data, which tells you what people type but not why. It is the same limitation I point out to a client who arrives with a dashboard and no interviews.

So the next step is not more SEO. It is talking to knitters.

What I do know is that I am somewhat obsessed with this, and that there is a good deal more coming.

If you crochet, tejelab.com is there and it is free. If something helps, or if something is calculated wrong (entirely possible), write to me at [email protected]. Corrections about the maths of knitting are as useful to me as corrections about SEO.

---

Share