Add 12-month growth plan, baseline, weekly log and measurement loop

Week-by-week plan grounded in current GSC data (4,745 impressions, 9 clicks,
CTR 0.19%, hosting cluster at position 27.5, 1,090 impressions on page-1
zero-click queries). One bounded change per week, measured the following week.
Snapshot script: ~/.hermes/scripts/growth_week.py.

growth/PLAN.md

@@ -0,0 +1,161 @@
+# odoo-expertos.com — 12-month improvement plan
+
+One change per week. Measure the previous week before starting the next.
+North-star metric: **affiliate clicks** (Plausible goal `Affiliate Click`).
+Supporting metrics: GSC clicks, money-keyword position, outbound CTR.
+
+Guiding rules:
+
+- A week ships **one** bounded change, deployed and verified live, logged in `WEEKLY-LOG.md`.
+- Every change states a hypothesis and the metric it should move, **before** it ships.
+- Next week starts by reading the result of last week. Keep, revert, or adjust — then move on.
+- Impressions are the guardrail: if GSC impressions drop more than 10% week over week, stop
+  new work and diagnose before continuing.
+- No invented proof. No fake testimonials, certifications or client names, ever.
+
+Baseline (2026-10-24, see `BASELINE.md`): 4,745 impressions, 9 clicks, CTR 0.19%, avg
+position 24.7, money cluster at position 27.5, affiliate clicks ≈ 0.
+
+---
+
+## Phase 0 — Instrumentation (week 0)
+
+Done before the loop starts: Plausible tag on 1,206 pages, `Affiliate Click` /
+`Outbound Click` events, UTM on all 479 outbound links, weekly snapshot script, this plan.
+
+**Week 1 — Product marketing context.**
+Write `.agents/product-marketing.md`: ICP, pains, objections, proof, positioning. Every
+later copy decision reads from this file.
+*Metric:* none (enabler). *Done when:* file exists and is committed.
+
+---
+
+## Phase 1 — Remove what disqualifies us (weeks 2–7)
+
+**Week 2 — Strip the fabricated proof.**
+Remove the three invented testimonials, the ISO 27001 / SOC 2 / "200+ proyectos" badges, and
+the placeholder phone number (8 pages). Replace with the real customer list.
+*Hypothesis:* E-E-A-T and trust signals improve; removes legal exposure.
+*Metric:* no ranking loss (guardrail week), Plausible outbound CTR on those pages.
+
+**Weeks 3–6 — CTR rescue on page-1 zero-click queries.**
+17 queries sit on page 1 with 0 clicks and hold 1,090 impressions (23% of all impressions).
+Rewrite title + meta description for 4 landing pages per week, ordered by impressions:
+`odo ia` (279), `odoo de donde es` (154), `de donde es odoo` (115), `odoo honduras` (81),
+`.odo ia` (66), `odoo origen` (60), `odoobot que es` (56), `docker odoo 20` (32 ×2),
+`odoo 20 docker`, `odoo de que pais es` (32), `hosting economicos` (30), `odoo inmobiliaria`
+(27), `odoo bot` (25), `odoo de donde es originario` (25).
+*Hypothesis:* these positions already earn the impression; the snippet is what fails.
+*Metric:* CTR on those queries (target 0.19% → 1.5% site-wide by week 8).
+
+**Week 7 — Kill the empty page.**
+`pt/odoo-ia/odoo-sistema-reservas-booking-ia-2025/` is a 0-byte file in the sitemap that
+returns an empty 200. Fill it with the translated ES content or remove it and regenerate the
+sitemap.
+*Metric:* index coverage.
+
+---
+
+## Phase 2 — Make the money pages worth ranking (weeks 8–19)
+
+Context: `odoo hosting` = 298 impressions at position 31.5. `hosting odoo` = 85 at 28.5.
+The affiliate funnel is correct but nothing reaches it.
+
+**Weeks 8–10 — Consolidate the thin hosting pages (ES first).**
+40 near-duplicate `/odoo-hosting/*` pages per language cannibalize each other. Keep one hub,
+keep the 6 genuine country pages, 301 the rest into the closest survivor, regenerate the
+sitemap. Do not delete pages that have impressions without checking GSC per URL first.
+*Hypothesis:* consolidating authority lifts the hub for `hosting odoo`.
+*Metric:* position of `hosting odoo` / `odoo hosting`, hub impressions.
+
+**Weeks 11–13 — Rebuild the ES hosting hub as a real comparison.**
+Actual prices ($214/$395/$595/$950+), what each plan includes, backups, regions, migration
+path, who it is wrong for. Comparison table above the fold. `Service` + `FAQPage` schema.
+One primary CTA with our UTM.
+*Hypothesis:* a page that answers "which plan and why" earns links and converts.
+*Metric:* affiliate clicks from `/odoo-hosting/`, position for `odoo hosting`.
+
+**Weeks 14–16 — Same rewrite for EN, then DE.**
+DE matters: `odoo beratung` has 155 impressions at position 51.5 — the German section is
+underperforming relative to its size.
+*Metric:* per-language affiliate clicks.
+
+**Weeks 17–19 — Internal links into the money pages.**
+Add contextual links from the 50 highest-impression articles to the hosting hub, in the
+article's own language. Anchor text describes the destination, not "click here".
+*Metric:* hub entry pages, affiliate clicks.
+
+---
+
+## Phase 3 — Content that deserves to rank (weeks 20–30)
+
+**Weeks 20–23 — Upgrade the top 20 impression pages.**
+Add what competitors don't have: screenshots from a real instance, benchmark numbers,
+version-specific detail (19 vs 20), "what this costs" sections.
+*Metric:* average position and CTR on those 20 URLs.
+
+**Weeks 24–27 — Article-level CRO.**
+Contextual CTA blocks inside articles at natural stopping points, matching the article's
+topic (hosting article → hosting CTA; module article → module CTA). No popups.
+*Metric:* affiliate clicks per 1,000 sessions; scroll-to-CTA rate.
+
+**Weeks 28–30 — Contact path.**
+The only conversion today is `mailto:`. Replace with a short form (name, email, what they
+need) posting to the existing webhook, with a real success state.
+*Metric:* form submissions (new goal), goal conversion rate.
+
+---
+
+## Phase 4 — Locale quality (weeks 31–36)
+
+**Weeks 31–33 — Audit ar/pt/fr.**
+Find pages whose main content is untranslated boilerplate. Either translate the body or
+`noindex` them. Never cross-locale canonical, never hreflang a page you don't want indexed.
+*Metric:* index coverage ratio per locale, locale impressions.
+
+**Weeks 34–36 — Localise the money pages properly.**
+Currency, payment methods, local hosting expectations, local phone/hours where real.
+*Metric:* affiliate clicks from non-ES locales.
+
+---
+
+## Phase 5 — Authority (weeks 37–46)
+
+**Weeks 37–40 — Original data asset.**
+Publish "Odoo hosting prices compared" with a method, a date, and a table others can cite.
+This is the link magnet; everything else is maintenance.
+*Metric:* referring domains, positions for comparison queries.
+
+**Weeks 41–43 — Directory + OCA presence.**
+Submit to relevant directories; get the OCA module tracker listed where Odoo users look.
+*Metric:* referring domains, referral sessions.
+
+**Weeks 44–46 — Outreach round.**
+20 personalised pitches to Odoo blogs/communities pointing at the data asset. No bought links.
+*Metric:* links earned, referral traffic.
+
+---
+
+## Phase 6 — Scale what works (weeks 47–52)
+
+**Weeks 47–49 — Template the winner.**
+Take the best-converting page structure and apply it to the 6 country pages.
+*Metric:* affiliate clicks per country page.
+
+**Weeks 50–51 — Decay sweep.**
+Refresh the 20 posts with falling impressions (update prices, versions, dates).
+*Metric:* recovered impressions.
+
+**Week 52 — Year review.**
+Compare every phase against its metric. Kill what didn't work, codify what did into the
+templates and into the `odoo-expertos-site` skill.
+
+---
+
+## Weekly loop (what the cron job does)
+
+1. Snapshot metrics, diff against last week, print the delta.
+2. Read last week's log entry: what shipped, what it was supposed to move.
+3. Record the verdict: keep / revert / adjust.
+4. Take the next week's action from this file, implement it, deploy, verify live.
+5. Write the log entry with hypothesis + metric before moving on.