CCG PMax — Account Audit

Bidding & budgets

Campaign	Bidding	Target	Budget/day	Status
Standard	Max Conv Value	none (no tROAS)	$1,000	Live
NY	Max Conversions	none (no tCPA)	$48	Paused
NY-DBT	Max Conversions	none (no tCPA)	$48	Paused
Harrison	Max Conversions	none (no tCPA)	$48	Paused
Bethesda	Max Conversions	none (no tCPA)	$48	Paused
Arlington	Max Conversions	none (no tCPA)	$48	Paused
DC	Max Conversions	none (no tCPA)	$48	Paused
Parsippany	Max Conversions	none (no tCPA)	$48	Paused
Red Bank	Max Conversions	none (no tCPA)	$48	Paused

No targets anywhere. The live campaign maximizes value with no tROAS (no efficiency floor); the locals maximized conversions with no tCPA against the wrong goal. The 8 locals shared identical $48/day budgets — too little signal each.

The decaying trend (live campaign)

Month	Spend	Conv	Value	ROAS
2026-01-01	$30,373	91	$50,134	1.65×
2026-02-01	$30,023	97	$59,034	1.97×
2026-03-01	$30,363	138	$86,778	2.86×
2026-04-01	$30,232	106	$76,320	2.52×
2026-05-01	$30,354	72	$52,020	1.71×
2026-06-01	$29,487	52	$37,080	1.26×

June through the 30th; offline-conversion lag will revise it up slightly, but the Mar→May decline is on complete months.

Prior period — H2 2025 (scaling context)

Month	Spend	Conv	Value	ROAS
2025-07	$9,688	152	$66,891	6.90×
2025-08	$9,521	100	$42,983	4.51×
2025-09	$13,569	7	$5,040	0.37×
2025-10	$19,437	13	$10,800	0.56×
2025-11	$30,552	25	$43,920	1.44×
2025-12	$30,209	64	$31,559	1.04×

The campaign was far more efficient at lower spend. In Jul 2025 it ran ~$9.7K and returned 6.9× (152 conv / $66.9K). As budget scaled toward $30K/mo (Nov+), efficiency fell to the ~1.3–2.0× seen in 2026. Sep–Oct 2025 also shows a near-zero-conversion stretch (7 & 13 conv) — a likely tracking break or learning reset worth confirming. Read together with the 2026 decline, this is evidence the campaign has been scaled past its efficient frontier, not just fatiguing.

Dayparting (live campaign, by weekday)

Day	Spend	Conv	Value	ROAS
Monday	$27,249	83	$54,750	2.01×
Tuesday	$23,330	74	$47,859	2.05×
Wednesday	$24,047	68	$45,093	1.88×
Thursday	$25,896	94	$61,524	2.38×
Friday	$22,783	89	$57,223	2.51×
Saturday	$28,946	74	$49,063	1.69×
Sunday	$28,580	72	$45,851	1.60×

Fairly even across the week (Thu/Fri marginally strongest); no material daypart signal to act on. PMax has no ad-schedule bidding control regardless — this informs budget/messaging only.

Settings hygiene

Check	Finding	Flag
Conversion goal config	Live = PURCHASE biddable only (Won) · locals = PURCHASE + CONTACT×3 origins ($0 taps) biddable	core misconfig
Brand exclusion	No brand list applied — brand traffic runs through PMax	MISSING
Final URL expansion	Not exposed via API v24 — confirm in UI	VERIFY
Geo target type	PRESENCE / PRESENCE (not interest-based)	PASS
Negative keywords	359 applied to the live campaign	PASS
Channel mix	7,263 placements; impressions skew to WEBSITE/Display (110.8K) vs Google-owned (41.7K) & YouTube (5.2K)	REVIEW
Budget	Capped at $1,000/day, spending to cap monthly	CAPPED

Prioritized actions

Now — measurement & read-accuracy

Lock the live campaign’s conversion goal to Won + Consult only; demote ad-extension taps to secondary. (resets learning)
Add a brand exclusion list; remove the “counseling center group” and “free therapy” search-theme signals.
De-duplicate conversion actions — one primary lead + one primary call action per step.

Next — fix the decay

Refresh creative / activate asset-group diversity — five months on one group while ROAS halved. (resets learning)
Hold budget flat until ROAS recovers; don’t scale into a decline.
Introduce a tROAS near the trailing achieved ratio after creative re-stabilizes. (resets learning)

Next — geo & device

Diagnose DC (1.02× on ~$21K); lean into Florida (2.69×).
Prioritize mobile post-click (4× better than desktop).

Later — structure & locals

Don’t relaunch the 8 locals as-is; if revived, set them PURCHASE-only (turn off the 3 biddable CONTACT goals) and consolidate the eight $48/day budgets.
Review account-level placement exclusions for the Display long tail.
Clean asset-group naming (geo-mismatched groups in the wrong campaigns).

Audit coverage (vs. the PMax framework)

What this audit covers, partially covers, or cannot close due to Google Ads API limits.

#	Step	Status	Notes
1	Frame + conversion setup	✓ covered	Window, data-driven attribution, offline/qualified-lead feedback covered. Enhanced conversions & consent mode = UI-only.
2	Top-line vs target	✓ covered	CPA/ROAS vs (no) target, budget cap, volume, lag — covered.
3	Channel / network distribution	◑ partial	Placement-TYPE impressions shown. PMax spend & conversions BY channel are not exposed by the API — needs the in-UI script. The one true black-box gap.
4	Search terms & brand split	✓ covered	Brand cannibalization quantified; brand-exclusion gap flagged.
5	Asset groups & assets	◑ partial	Ad strength + video presence covered. Per-asset Low/Good/Best ratings are not returned in API v24.
6	Listing groups / feed	— n/a	N/A — lead-gen, no Merchant Center feed (confirmed empty).
7	Audience signals & insights	✓ covered	Signals, search themes, off-target signals — covered.
8	Geo / device / time	✓ covered	Geo + device + dayparting included.
9	Bidding & budget	✓ covered	Strategy, targets, cap, learning. Search campaigns paused → no live overlap.
10	Incrementality / cannibalization	◑ partial	Prior-period trend + brand segmentation as proxy. True read needs a geo holdout / pause test — not yet run.
11	Prioritized actions	✓ covered	Above.

Account settings & hygiene