Google Weekly Briefing — October 12, 2026

Period: 30-day rolling (September 12 – October 11, 2026) · Data: Windsor Google Ads 🟢 (pulled Oct 12, data through Oct 11) | Windsor GA4 🟡 (pulled Oct 12; Oct 11 incomplete, Oct 7–10 read low and need a re-read next week) | Windsor Search Console 🟢 (pulled Oct 12, window Sep 10 – Oct 9, 3-day lag by design) | media_audit_metrics.json 🟢 (generated Oct 12) | newsletter intel via google_intel.py (14 days) · Inflation ratios: July 2026 vintage, measured Aug 21 — 52 days old 🔴; past 45 days, so every est. real figure carries an unquantified drift error until P1 refreshes them. Registers: docs/Google_PPC_Action_Plan_2026-10-08.md (P1–P11, D1–D5) · docs/Google_SEO_AEO_Action_Plan_2026-10-08.md (S1–S21, D1–D3). First run of the google-weekly-review skill (LIVE).

Snapshot

The read. Branded-Search is carrying the account (est. real ~6.6x on a third of spend). Seasonal-Scents-PMAX's late-September bounce did not hold: the settled week (Sep 28 – Oct 4) came in at 1.13x est. real, the provisional week at 1.35x, both under the 1.5x line in the Sep 28 decision framework, so this briefing recommends the $150 → $100/day cut (Proposed Action 1). October MER is 8.5 points over the 26% ceiling, which closes the scale gate: nothing on Google scales this month except by substitution from cuts. On the organic side, Google organic sessions fell for the first time in thirteen readings, and Search Console shows clicks down ~28% in the Oct 6–10 stretch on flat impressions — a visibility-without-clicks pattern, cause not yet known (SEO Metrics).

Campaign Table

30 days, Sep 12 – Oct 11. Conv value is Windsor's rule-adjusted, Google-attributed figure (L107); est. real = reported ÷ July ratio. "Budget-limited" = search_budget_lost_impression_share ≥ 0.10 on recent days (mean of the last 7 days shown).

Seasonal is both below the ~2.0x margin floor and limited by budget: Google would spend more if allowed, at a return that already loses money on the margin. That is the combination the cut is for. Branded-PMAX is flat at the floor's edge, unchanged (P8 pause test still waits on P3).

NCA signal (Pull A2, last 7 days, Oct 5–11): new_customer_lifetime_value is non-zero on all three campaigns — Branded-Search $1,587 of $6,612 conv value (24%), Seasonal $517 of $1,984 (26%), Branded-PMAX $42 of $813 (5%). The field populates, so the "reads zero where unset" test in P3 no longer applies; whether the goal is in "bid higher for new customers" mode is still unverified — confirm in Chrome (P3).

Weekly Trend

ISO weeks (Mon–Sun), reported / est. real. Settled = the week §2.1 verdicts use (Sep 28 – Oct 4). Provisional = still attributing (Oct 5–11); shown, never decided on. The current week (Oct 12–) has no data yet.

• Revision: the Sep 28 briefing showed Seasonal at 3.68x / 2.66x on a 6-day partial (Sep 22–27). The full settled week (Sep 21–27) reads 4.07x / 2.95x — the bounce was real and larger than reported.
• Verdict (P6, playbook §2.1): settled week est. real 1.13x < 1.5x → recommend cut to $100/day, on est. real values, L107 interim (PPC D5 still open). The Sep 28 framework said exactly this: "if it reverts below 1.5x, execute the $100/day cut." The 1.5x / 2.0x thresholds were set Sep 28 and Oct 8 and are unvalidated; the verdict turns on them.
• Daily spend Sep 28 – Oct 11 held at ~$100 / ~$150 / ~$45 per campaign: no budget change has been made since Aug 21.
• Branded-Search had its two best weeks since August (7.5x and 7.4x est. real) on 24–26% lost impression share; the brand cap is binding, as intended (D1 open).

MER line

Google is 2.7 points of a 34.5% October MTD MER (through Oct 11) against a 26% ceiling. Google's share is small; the ceiling is breached by the whole, so the scale gate (NC-ROAS ≥ 2.5x and MER trending under ceiling) is closed for Google this month regardless of campaign ROAS.

Execution Tracker — PPC register

Evidence sources this run: Pull A daily spend (budget changes), Pull A2 (NCA), live checks. Weeks since the register last moved: 0 (created Oct 8); weeks since any PPC action completed: 7 (Aug 21). This is the eighth consecutive briefing with no completed PPC action.

Execution Tracker — SEO/AEO register

Evidence sources this run: Shopify GraphQL (title_tag / description_tag metafields, collection seo{}), live-page fetches (curl) for schema and FAQ headings, GSC page rows. Weeks since the register last moved: 0 (created Oct 8); SEO actions executed by this program since Aug 21: none — one item (S4) is found done on the live site by an unknown hand.

Tier A page tracker

Query-level rows vs the Sep 28 baseline (playbook §4.2). GSC window Sep 10 – Oct 9. No title or meta has been pushed, so these are drift, not effect.

The two "/3mo" rows in the playbook are re-baselined this run at page level, 30 days (the query-level rows for them were not measurable):

The "house smell good" head term halved in impressions (14,955 → 7,212) at a stable position — the window moved from late August into October and the query is seasonal, but the page-level 105K impressions at 0.6% CTR is the number that matters: each 0.1 point of CTR on that page is ~105 clicks a month. That is what S1/S3 are for.

SEO Metrics

Organic sessions (GA4, 30d). The thirteen-reading growth streak ends.

What is inside the −14.5%: Sep 12 – Oct 6 ran ~314 sessions/day (a 30-day pace of ~9,400, already ~8% under the prior window); Oct 7–10 ran 156–211/day (~40% under pace); Oct 11 reads 115 and is almost certainly incomplete (GA4 via Windsor lags a day). Search Console corroborates the drop and rules out pure lag: site-wide clicks averaged 228/day Sep 26 – Oct 5 and 163/day Oct 6–10 (−28%) while impressions stayed at 9.1–10.4K/day. Clicks fell on flat visibility. Cause unknown this run — possible causes are an AI Overview expansion on the informational cluster (see Outside Signal), a ranking change inside the top 10 that GSC's average hides, or a Google update. The next run re-reads Oct 7–11 with settled GA4 data; if it holds, this is an SEO event, and the answer is the Tier A work, not more rank.

GSC totals, Sep 10 – Oct 9: 3,950 clicks on 137,549 impressions. Branded (hand-matched "grow fragrance / candle / spray / car / air" variants) 2,739 clicks on 6,943 impressions; non-branded 1,211 clicks on 130,606 impressions (0.9% CTR).

Top non-branded queries (clicks):

Striking distance (non-branded, position 4–20, by impressions):

Cluster totals (non-branded; the review's own matching, so not comparable to the Sep 28 "~34K" figure, which used a narrower set):

Car freshener remains the #1 non-branded click source (298 clicks across the cluster). The "smell good" cluster is 44K impressions at 0.5% — the AI-answered query family; citation and CTAs, not rank, is the play there. Candles: 7.5K impressions, 27 clicks (S11 and the candle pages are not earning clicks at position 9–13).

AEO Metrics

GA4 source/medium, 30 days (Sep 12 – Oct 11), GA4-attributed. Same 🟡 lag caveat as organic.

Milestone trend: May 49 → Jun 392 → Jul 457 → Aug 31 652 → Sep 21 817 (ATH) → Sep 28 807 → Oct 12 719. Still inside the 800–1,200/month December target band's neighborhood but the first real pullback; ChatGPT is 97% of the total. Gemini doubled from a tiny base (9 → 15). Most AI influence is dark; referral counts alone do not judge the program.

AEO-segment conversion (Pull B2, S10 measurement): 719 sessions → 31 purchases → $1,747 GA4-attributed revenue, CVR 4.3%. Site-wide CVR in the same pull ≈ 4.3% (≈2,824 purchases on ≈65,200 sessions, hand-summed from the source/medium rows). Google organic converts at 6.5% and Google cpc at 6.2% in the same window. AI referrals convert like the site average, not better — the ChatGPT Shopping feed sessions do better (12 purchases on 160 sessions, 7.5%).

Prompt Audit status

Due the first Monday of each month in a live session with Dan. Sep 7 and Oct 5 were both missed; 5 weeks overdue from Sep 7. Next first Monday is Nov 2 — run one before then. Latest scorecard, Aug 5 baseline: category 62.5% (5/8), brand 100% (5/5). Named gaps — "non-toxic room spray," "pet-safe air freshener," "clean home fragrance 2026" — are open, unverified since Aug 5. Prompt set: playbook §5.2 (8 category, 5 brand, 3 diffuser; Gemini, Google AI Mode, ChatGPT, Perplexity).

What this briefing said last week

The Sep 28 briefing proposed five actions (the Oct 5 run produced no file).

1. Hold Seasonal at $150/day; decide on Sep 29 – Oct 5. Adopted — the hold happened (daily spend unchanged). The framework now fires: the settled week reverted to 1.13x est. real, below the 1.5x line, so the cut is this week's Proposed Action 1. The "dead-cat bounce" caution was right.
2. Build the diffuser campaign. Declined / not done — no campaign in Pull A. Carried as P2 (ninth briefing).
3. Refresh Customer Match + set the NCA goal. Not done; rationale superseded in part — Pull A2 shows new-customer value populating on all three campaigns, so "reads zero where unset" is no longer the test; the goal-mode check still needs Chrome (P3).
4. Refresh inflation ratios. Not done — 52 days (P1).
5. Run the September prompt audit. Not done; October's also missed (S18).

The Sep 28 "strategy assessment" said a fourth consecutive below-floor Seasonal week would trigger a lever-C downgrade. Two settled weeks have now read below 1.5x (Sep 28 – Oct 4; Oct 5–11 provisional), but the playbook's three-week count starts after a cut, and no cut has been made. Not triggered; the clock starts when the budget changes.

Outside Signal

AI for Ecommerce, Sep 30 — "The Click Is Being Designed Out of Commerce" (score 3, verified). An Ahrefs study of ~1,000 domains after Google's July AI Overviews expansion: sites with heavy AI Overview exposure lost 23–25% of median CTR in nine days with rankings barely moving; impressions can rise while clicks fall; informational queries lost most, branded queries held. That is the shape of this week's Search Console read — clicks −28% in Oct 6–10 on flat impressions and stable positions, concentrated away from brand (branded clicks still 2,739). It does not prove the cause, but it says what to measure next week: CTR by page on the informational cluster, not rank. The same item notes Google is auto-enrolling eligible US Shopify merchants in native AI checkout inside AI Mode and Gemini, opt-out under Shopify Sales channels → Agentic — worth a look in the Shopify admin to confirm whether Grow is enrolled and whether that is wanted (Dan decision; no action taken).

Proposed Actions

Priority order. Every Google Ads change runs only in a live session with Dan's per-change confirmation (R84).

1. Cut Seasonal-Scents-PMAX $150 → $100/day. Settled week 1.13x est. real, provisional 1.35x, 30-day 1.77x — all under the ~2.0x margin floor; the Sep 28 framework's "below 1.5x → cut" condition is met. Also budget-limited (26% lost IS), so Google is asking for more of the same. Impact: ~$1,500/month less spend at a return where marginal dollars lose money; on the settled week, $1,013 of spend returned ~$1,146 est. real Google-attributed value. Label: on est. real values, L107 interim — needs D5 (approve the exception in DECISIONS.md, or pull P11 forward first). After the cut, three settled weeks < 1.5x → "strategy update proposed" (lever C to $0 for the quarter). Never pause Seasonal. Path: Windsor set_campaign_budget (inspect the schema first), live session, Dan confirms. 5 minutes.
2. One Chrome session for P1–P5 (~2 hours). Ratios at 52 days are the single biggest accuracy risk in this document; the diffuser campaign (P2) is the biggest value gap (strategy lever C, $10–30K H2, unreviewed, ±20%) — product live ~9 weeks, GSC shows non-brand diffuser queries at 992 impressions and 1 click; P3 goal mode; P4 draft delete; P5 listing-group report and the two summer asset groups. Creative assets for P2 are the dependency — ask for them this week. Path: Claude in Chrome, Dan present, authuser=1, account 858-556-1537. Agenda below.
3. Treat the organic click drop as a watch item, not a conclusion. Next run re-reads Oct 7–11 with settled GA4; if the −28% click rate holds, it is an SEO event. Impact if it persists: ~65 fewer GSC clicks/day ≈ 1,950/month; at Google organic's 6.5% CVR and ~$70 GA4-attributed revenue per purchase that is on the order of $9K/month, GA4-attributed. Path: this review (data) + 5 minutes in Chrome on Search Console → Manual actions / Security issues to rule those out. The structural response is S1/S3 (CTAs and FAQ on the informational pages), which is why item 4 matters now.
4. Decide whether S1 may run ahead of the Amazon-first claims sweep. Three blog titles and metas (not Baby-Safe) after the §0 word pass: "non-toxic" allowed on informational pages, "100% plant-based" and "no synthetic chemicals" out. The "7 Ways" page carries 105K impressions a month at 0.6% CTR; the page-level baseline is now set, so a push is measurable. Claims gate: SEO D1 and D3 are open; §0 item 4 says Tier A rides the sweep. This is Dan's call, not the review's. Path: Claude via Shopify articleUpdate (global.title_tag / global.description_tag), 20 minutes, after Dan's yes and a read-back of each string against §6e.
5. Build P11, the Google Hub export, this week. etl/export_google_briefing_data.py → dashboards/google_briefing_data.json (Datahub-verified Google-attributed revenue from fact_orders, settled vs provisional rows, weekly spend, same-week-last-year), wired into daily_refresh.run_exports and worker.js (L84). It removes the L107 interim label from every future cut recommendation and resolves D5 the right way. Path: Claude, no Chrome, mirrors etl/export_media_briefing_data.py; ~half a day; Dan reviews the first output.

Strategy check: all five are consistent with playbook §3. No strategy update is proposed this week; the lever-C downgrade clock starts only after the Seasonal cut.

Chrome Agenda

Cumulative; one session clears it. Google Ads UI, authuser=1 (dan@growfragrance.com), account 858-556-1537 — never 439-754-1327.

Total for the core session (P1–P5 + the two 5-minute checks): about 2.5 hours.

System Notes

• Data quality: Windsor Google Ads 🟢 (through Oct 11, 1 day old; totals differ from last week, so not a stale pull). GA4 🟡 — Oct 11 incomplete; Oct 7–10 read ~40% under pace and are corroborated by GSC, but re-read next week before calling a trend. GSC 🟢 (Sep 10 – Oct 9). media_audit_metrics.json 🟢 (generated Oct 12; MER "MTD through Oct 11"). Intel script: 2 Google-relevant items in 14 days; one used.
• Inflation ratios: July vintage, 52 days 🔴, past the 45-day line — est. real figures carry an unquantified drift error (the ±3–4% band quoted in September briefings had no derivation and is not repeated). P1.
• Settled vs provisional: verdict on Sep 28 – Oct 4 (settled). Oct 5–11 shown as provisional; next week restates it.
• Register maintenance (LIVE run): PPC register — P6 and P7 status cells updated with this week's evidence, P3 note added. SEO/AEO register — S4 marked ✅ Oct 12 with the live-page evidence; S21 marked running. Playbook §4.2 — the two "/3mo" rows re-baselined at page level from this run's 30-day pull; §5.3 availability fact corrected. No ✅ on inference.
• Legacy briefing found and archived (D4): docs/Google_Briefing_2026-10-12.md already existed at 11:43 today in the old Master-Plan format (23-action tracker, media_audit_metrics dated Oct 9). This scheduled task's only recorded run is 12:17. Provenance matches the Cowork google-weekly-briefing task that was supposed to be retired; a Weekly_Media_Briefing_2026-10-12.html appeared in dashboards/briefings/ at 11:47 from the same window. Moved to archive/Google_Briefing_2026-10-12_legacy_1143.md (uncommitted; not deleted). Dan: confirm the Cowork tasks are off (D4), or two Google emails arrive every Monday.
• Parallel-session hygiene (L80a) — one miss to know about: at deploy time two files another session had staged (dashboards/Grow_Dashboard.html, dashboards/Meta_Weekly_Briefing.html) were unstaged first (working tree untouched). Between that step and the commit, the daily-refresh session staged nine more files (Quadruple_Whale_Dashboard.html, briefings/Weekly_Media_Briefing_2026-10-12.html, and seven dashboards/.json exports), and deploy.py --only filters only what *it stages — git commit takes the whole index. So commit 1d35e07 carries those nine refresh outputs under the Google briefing message. They are today's generated data files, not hand edits, and were left in place rather than reverted; the gap in deploy.py is flagged as a follow-up. Local was level with origin at run start.
• Claims surfaces seen, not touched: the live home page title "100% Plant-Based Room Sprays & Candles" and meta description ("100% plant-based, toxin free and safe") — brand-line decision (claims review D-A / SEO D3); the Tier A copy word pass (§0). Nothing pushed.
• Open Dan items: D5 (L107 interim exception — blocks Action 1), D1 (brand budgets), SEO D1/D3 (claims gate), D2 (passkeys / josh@), D3 (email to Katelyn and Mike, pending since Aug 21), D4 (Cowork tasks), prompt audit before Nov 2, Shopify "Agentic" sales channel check.
• Standing rejections unchanged: AI Max, Demand Gen, broad-match adds, "target all eligible products" auto-apply, one-jump brand raises.