# Weekly log

One entry per week. The hypothesis and the metric are written **before** the change ships;
the measurement is filled in by the following week's run. A week without a measurement is
an unfinished week.

| Week | Date | Change | Hypothesis | Metric watched | Expected | Measured | Verdict |
|---|---|---|---|---|---|---|---|
| 0 | 2026-10-24 | Instrumentation: Plausible tag on 1,206 pages, `Affiliate Click` / `Outbound Click` custom events, UTM on 479 outbound links, CTA blocks on 6 homepages + 6 hosting hubs, weekly loop installed | Without measurement every change is a guess | all | baseline established | n/a (setup) | keep |
| 1 | 2026-10-24 | `.agents/product-marketing.md` (d82af02): ICP, pains, objections, proof policy, positioning — the source of truth every later week reads | Copy decisions were being re-litigated weekly; without fixed positioning each change invents its own | none (enabler); guardrail = no ranking loss | no regression | pending (week 2 run) | keep |
| 2 | 2026-10-24 | `2e0882d`: removed the 3 invented testimonials, the ISO 27001 / SOC 2 trust badges and the placeholder phone from the 6 homepages | Invented proof is an E-E-A-T and legal liability, and it undermines the affiliate pitch it sits next to | GSC impressions on `/` and the locale homes (guardrail: must not fall); Plausible outbound CTR | no ranking gain assumed; risk removed | pending (Friday run) | keep |
| 3 | 2026-10-24 | `2e0882d`: titles + descriptions rewritten on 3 pages covering 6 of the 17 page-1 zero-click queries (`¿De dónde es Odoo?`, `Odoo en Honduras: facturación DEI`, `Odoo IA: qué automatiza`) | Those queries already rank page 1 and earn 0 clicks; the snippets were machine slug-dumps with a ` - <slug> \| ES` artifact eating half the description | CTR on those queries, currently 0.00% | 0.19% site-wide CTR → 1%+ as the rewritten pages cycle through the SERP | pending (needs 2-4 weeks for GSC to re-crawl and re-rank) | keep |
| 4 | 2026-10-24 | `582a93e`: broken internal links 1,370 → 0; empty PT page removed with its sitemap block and hreflang alternates; 23 relative canonicals made absolute and 62 duplicate canonicals removed; EN homepage title/description/H1 aligned to `odoo experts` | 1,370 dead ends wasted crawl paths and dropped readers on 404s; a relative canonical resolves to whichever host served the page, and apex and www both answer 200 | `broken_internal_count`, `no_canonical`, `canonical_not_www` in audit_site.py; position + CTR for `odoo experts` (75 impr @ 13.7, 0 clicks) | audit counts at 0; brand query into page 1 | audit re-run: 0 broken, 0 relative, 1 page without canonical (404.html, deliberate) | keep |
| 5 | 2026-10-24 | `a52f0f6`: reworked `/odoo/empresa-odoo-historia-fundacion-2025/` — intent-led head (`Es belga, y así se usa en LATAM`), an "Odoo en 30 segundos" fact block above the fold, a LATAM localisation section and a São Paulo CTA to the host | 6 queries (~400 impressions, positions 4.5-7.7) earn zero clicks: the snippet sold company history while the searcher is asking whether Odoo works where they live | CTR and clicks on `de donde es odoo`, `odoo de donde es`, `odoo origen`, `odoo de que pais es`, `quien creo odoo`, `odoo de donde es originario`; affiliate clicks on the page | CTR 0% → 2%+ on those queries; first affiliate clicks traceable via utm_campaign=empresa-odoo-historia | pending (GSC needs 2-4 weeks to re-crawl; affiliate clicks unmeasurable until the Plausible goals exist) | keep |

---

## Entry template

```markdown
### Week N — YYYY-MM-DD

**Shipped:** <what changed, with commit hash>
**Hypothesis:** <why this should move a number>
**Metric:** <the one number that decides>
**Expected:** <direction and rough size>

**Measured (week N+1):** <actual, from growth_week.py>
**Verdict:** keep / revert / adjust — and why

**Next:** <carried into the following week>
```