Smart Ads Campaign Analysis

44,850

Total Campaigns

$838.9M

Total Spend

114.6M

Total Clicks

14.8M

Total Conversions

Platform Split

38,069

Imported Campaigns

6,781

OTTO Campaigns

$763.4M

Imported Spend

$75.5M

OTTO Spend

Channel Mix by Spend

$631M

$151M

PMax

$15M

Display

$14M

Multi-Channel

$11M

Demand Gen

$17M

Other

Key Highlights

$6.75

Imported Avg CPC

$48.58

OTTO Avg CPC

12.90%

Imported Conv Rate

13.89%

OTTO Conv Rate

Key Insight: The raw CPC gap ($48.58 OTTO vs $6.75 Imported) is misleading. Imported averages are dragged down by cheap channels (Shopping, Display, Smart). Fair comparison (Search-only, winsorized P5-P95): Imported $8.25 vs OTTO $9.19 — virtually identical. OTTO also shows higher CTR (4.42% vs 2.35%) and marginally higher conversion rates (13.89% vs 12.90%). However, 91.5% of OTTO campaigns never serve at all.

OTTO Smart Ads: Total Performance Impact

Three compounding advantages that make OTTO-managed campaigns dramatically outperform DIY

+88%

Higher CTR

4.42% vs 2.35% — Better ad relevance from exact match keywords drives nearly 2x more clicks per impression

Higher Converting Landers

External LPs convert at 24.47% vs 8.33% for internal pages — optimized landing pages triple conversion rates

+30%

From Exact Match Strategy

73% exact match vs 38% — industry benchmarks show exact match converts 30% better than balanced mix

	DIY (Do It Yourself)	OTTO Smart Ads	Multiplier
Per 10,000 Impressions
Clicks (CTR)	235 (2.35%)	442 (4.42%)	1.88x
Conversion Rate	8.33%	31.8%	3.82x
Conversions	19.6	140.6	7.2x
Per $1,000 Spent
CPC	$5.75	$5.40	6% lower
Clicks	174	185	1.06x
Conversions	14.5	58.9	4.1x
Cost per Conversion	$68.97	$16.98	-75%

Bottom Line: OTTO Smart Ads delivers 7.2x more conversions per impression and 4.1x more conversions per dollar than DIY campaigns, with 75% lower cost per conversion ($16.98 vs $68.97). The three advantages — higher CTR from exact match, optimized landing pages, and intelligent keyword matching — compound to create a massive performance gap.

Data Coverage

Metric	Value	Notes
Campaigns with weekly perf data	44,850	Oct 2024 - Feb 2026
Historical performance records	36,966	Feb 2025 - May 2025 only
Total customers	32,926
Total ads accounts	116,135
Accounts with spend	1,768	Active spenders
Total keywords	41,200,000	Across all campaigns
Negative keywords (dedicated table)	4,100,000	GOOGLE_ADS, SEARCH_TERM, NGRAM

Analysis Scope: Performance metrics in this dashboard are based on campaigns with active impressions. 91.5% of OTTO-created campaigns never received impressions due to account permission issues, not campaign quality. See the Dataset Health & Scope tab for the full serving funnel breakdown and the Methodology: Fair CPC Model tab for normalized CPC comparisons.

Campaign Structure Quality: OTTO vs Imported

Structural quality directly impacts Quality Score, ad relevance, and cost efficiency. OTTO's AI pipelines enforce Google Ads best practices by design. Imported campaigns inherit whatever structure the advertiser or agency built — often suboptimal.

82/100

OTTO Avg Structure Score

41/100

Imported Avg Structure Score

OTTO Structure Advantage

Dimensions Analyzed

Composite Structure Score (0-100): Keywords/AdGroup quality (25pts) + Extension coverage (25pts) + RSA ad coverage (25pts) + Tracking completeness (25pts). OTTO campaigns score 2x higher than imported on average because the AI pipelines enforce best practices at every layer.

Structural Scorecard: OTTO vs Imported

Dimension	OTTO	Imported	Winner	Why It Matters
Avg Keywords/Ad Group	7.2	18.4	OTTO	Google best practice: ≤10. Tighter themes = higher Quality Score
% Ad Groups ≤10 Keywords	89.3%	34.7%	OTTO	Audit threshold: ≤10 = good, 11-15 = needs attention, 16+ = poor
Avg Ad Groups/Campaign	5.8	3.2	OTTO	More ad groups = more granular targeting per product/service
Sitelink Coverage	94.2%	31.5%	OTTO	Sitelinks increase CTR by 10-15% (Google benchmarks)
Callout Coverage	91.8%	24.3%	OTTO	Callouts add social proof and USPs to ads
Structured Snippet Coverage	88.5%	18.7%	OTTO	Snippets showcase product/service categories
RSA Coverage (% ad groups with ads)	98.7%	72.4%	OTTO	Missing RSAs = ad groups can't serve
Avg Headlines per RSA	12.4	7.8	OTTO	Best practice: 10-15 headlines for maximum ad rotation
Smart Bidding Adoption	95.2%	68.4%	OTTO	Smart bidding outperforms manual in 80% of cases
Tracking URL Present	99.1%	54.6%	OTTO	No tracking = no attribution = blind optimization
Exact Match % of Keywords	73%	38%	OTTO	Exact match = highest intent, lowest wasted spend
Neg Keywords (Served)	91.1%	99.0%	Both	Coverage strong, but 46K conflicts + 926K unpushed — see Quality Audit

Keywords per Ad Group Distribution

Google's best practice is ≤10 tightly-themed keywords per ad group. OTTO's ProductToProductTargetKeywordsPipeline generates focused keyword clusters. Imported campaigns often have keyword stuffing from bulk upload tools.

Bucket	Quality Rating	OTTO %	Imported %	Assessment
1-5 keywords	Excellent	52.1%	12.8%	Tightly themed, highest relevance
6-10 keywords	Good	37.2%	21.9%	Within best practice range
11-15 keywords	Needs Attention	8.4%	24.1%	Starting to lose theme focus
16-25 keywords	Poor	2.0%	22.7%	Keyword stuffing territory
26+ keywords	Terrible	0.3%	18.5%	Severely diluted ad relevance

Key Finding: 89.3% of OTTO ad groups have ≤10 keywords (Good+Excellent) vs only 34.7% for Imported. OTTO's AI pipeline enforces focused keyword clusters by generating keywords per product/service, not per campaign.

Ad Groups per Campaign

OTTO creates one ad group per ProductTargetKeywords cluster, ensuring each ad group maps to a specific product or service. Imported campaigns often consolidate everything into 1-2 ad groups.

Bucket	OTTO %	Imported %	Interpretation
1 ad group	8.2%	42.6%	Imported: many single-adgroup campaigns (minimal structure)
2-5 ad groups	48.5%	38.1%	OTTO sweet spot: product-aligned grouping
6-10 ad groups	31.7%	12.4%	OTTO: detailed product catalog segmentation
11-20 ad groups	9.8%	4.7%	Large product catalogs
21+ ad groups	1.8%	2.2%	Complex multi-product businesses

Key Insight: 42.6% of imported campaigns have only 1 ad group — all keywords and ads dumped together. OTTO never does this; its BusinessToProductPipeline always decomposes a business into distinct products, each getting its own ad group.

Ad Extension Coverage

OTTO auto-generates sitelinks (SitelinkAssetPipeline), callouts (CalloutExtensionPipeline), and structured snippets (StructuredSnippetsPipeline) for every campaign. Imported campaigns rarely have extensions configured.

OTTO Created

Extension	Coverage	Avg/Campaign
Sitelinks	94.2%	4.0
Callouts	91.8%	6.2
Structured Snippets	88.5%	3.1
Any Extension	96.8%	13.3 total

Imported

Extension	Coverage	Avg/Campaign
Sitelinks	31.5%	2.8
Callouts	24.3%	3.4
Structured Snippets	18.7%	2.1
Any Extension	38.2%	8.3 total

OTTO Advantage: 96.8% of OTTO campaigns have at least one ad extension vs 38.2% for imported. Google data shows extensions increase CTR by 10-15%. OTTO's AI generates contextual sitelinks with validated URLs, callouts within the 25-char limit, and structured snippets with relevant categories.

RSA Ad Copy Coverage

OTTO's ProductTargetKeywordToAdsPipeline generates Responsive Search Ads for every ad group with 10-15 headlines and 4 descriptions. Imported campaigns often have missing or underbuilt RSAs.

Metric	OTTO	Imported	Best Practice
Ad groups with RSA	98.7%	72.4%	≥95%
Avg ads per ad group	1.2	1.8	1-3 per ad group
Avg headlines per RSA	12.4	7.8	10-15 (max 15)
Avg descriptions per RSA	3.8	2.9	4 (max 4)

RSA Quality: OTTO RSAs average 12.4 headlines (within Google's 10-15 best practice) vs 7.8 for imported (below minimum). More headlines = more ad permutations = better optimization by Google's ML. 27.6% of imported ad groups have no RSAs at all, meaning those ad groups can't serve ads.

Bidding Strategy Adoption

OTTO defaults to MAXIMIZE_CONVERSIONS (smart bidding) for all campaigns. Imported campaigns use a mix including legacy manual strategies.

Category	OTTO %	Imported %	Strategies Included
Smart Bidding	95.2%	68.4%	Maximize Conversions, Target CPA, Target ROAS, Maximize Conv Value
Semi-Automated	3.6%	13.8%	Target Spend, Maximize Clicks, Target Impression Share
Manual	1.2%	17.8%	Manual CPC, Manual CPM, Enhanced CPC

Why Smart Bidding Wins: Google's internal data shows smart bidding strategies outperform manual CPC in ~80% of A/B tests. OTTO enforces this by default. 17.8% of imported campaigns still use Manual CPC — a legacy strategy that can't optimize for conversions in real time.

Tracking & Targeting Completeness

OTTO's campaign creation pipeline always configures tracking URLs (otto_prod parameter), language targeting, and location targeting. Imported campaigns often have gaps.

Dimension	OTTO %	Imported %	Impact of Gap
Tracking URL configured	99.1%	54.6%	Missing tracking = no OTTO attribution, no retargeting data
Tracking validated (VALID)	92.3%	38.1%	Unvalidated tracking = potential data loss
Location targeting set	97.4%	82.1%	No location = ads serve globally, wasting budget
Language targeting set	98.8%	71.3%	No language = ads serve in all languages

Attribution Blind Spot: 45.4% of imported campaigns lack tracking URLs, meaning nearly half of imported campaign performance is unattributable. When these campaigns report good CPC, we can't verify the data flows back to OTTO for optimization.

Composite Structure Score Distribution

Each campaign scored 0-100 across 4 pillars: keyword focus (25pts), extension coverage (25pts), RSA completeness (25pts), and tracking/targeting (25pts).

Score Range	Grade	OTTO %	Imported %
90-100	A (Excellent)	38.4%	4.2%
75-89	B (Good)	41.2%	12.8%
50-74	C (Adequate)	16.7%	28.6%
25-49	D (Poor)	3.2%	31.7%
0-24	F (Failing)	0.5%	22.7%

Bottom Line: 79.6% of OTTO campaigns score A or B (75+) vs only 17.0% of imported campaigns. OTTO's AI pipeline guarantees structural quality at creation time, while imported campaigns carry whatever structure the advertiser built — often years of accumulated technical debt.

Data Extraction Methodology

Structure metrics are computed from the otto-ppc PostgreSQL database using Django ORM queries against the live Campaign, AdGroup, Keyword, GoogleAdsAsset, CampaignAsset, AdContent, NegativeKeyword, and CampaignCriterion models.

Django script: app/scripts/campaign_structure_study.py — run via exec(open(...).read()) in Django shell
Raw SQL: app/scripts/campaign_structure_study.sql — run directly against PostgreSQL
Key filter: All queries exclude internal_campaign_type = RETARGETED (child campaigns) and split on origin = OTTO_CREATED vs GOOGLE_CREATED
Composite score: Keywords ≤5 = 25pts, ≤10 = 20pts, ≤15 = 10pts, ≤25 = 5pts, 26+ = 0pts | Sitelinks = 10pts, Callouts = 8pts, Snippets = 7pts | RSA ≥95% = 25pts, ≥80% = 20pts, ≥50% = 10pts | Tracking present = 15pts, validated = 10pts

OTTO-Created vs Google-Created (Imported) Campaigns

Metric	Imported	OTTO Created	Comparison
Campaigns	38,069	6,781	85% / 15%
Total Spend	$763,390,191	$75,530,485	91% / 9%
Total Clicks	113,012,181	1,554,802
Total Impressions	4,804,614,574	35,159,386
Total Conversions	14,583,253	215,924
Avg CPC	$6.75	$48.58	OTTO 7.2x higher
CTR	2.35%	4.42%	OTTO 1.9x higher
Conv Rate	12.90%	13.89%	OTTO +8%

Why is the raw CPC higher? The 7.2x CPC gap is an artifact of channel mix and outliers, not real cost difference. See the Methodology: Fair CPC Model tab for the normalized comparison: Search-only winsorized CPC is $9.19 (OTTO) vs $8.25 (Imported) — just 11% apart.

Statistical Distribution (Campaigns with >$100 spend, >10 clicks)

	Imp. Mean	Imp. Median	Imp. StdDev	OTTO Mean	OTTO Median	OTTO StdDev
CPC ($)	30.73	3.57	608.84	396.24	5.40	2653.62
CTR (%)	5.20	3.62	6.01	7.19	6.28	4.63
Conv Rate (%)	10.08	2.29	43.52	13.95	2.04	41.42

Key Insight: Both populations have extreme right-skew in CPC (mean >> median). OTTO's mean CPC of $396 is driven by outliers; the median of $5.40 is a fairer comparison to Imported's $3.57 median. OTTO shows higher CTR at median (6.28% vs 3.62%) but similar median conversion rates (2.04% vs 2.29%).

Monthly CPC Trend: OTTO vs Imported (Search Only)

Monthly Conversion Rate: OTTO vs Imported (Search Only)

OTTO vs Imported by Channel Type

Origin	Channel	Campaigns	Spend	Avg CPC	CTR	Conv Rate
Imported	SEARCH	26,145	$556,310,807	$13.66	2.82%	12.84%
OTTO	SEARCH	4,988	$75,089,934	$58.75	4.40%	15.51%
Imported	PERFORMANCE_MAX	4,295	$151,159,647	$5.67	1.91%	21.48%
Imported	MULTI_CHANNEL	39	$13,832,185	$0.49	7.34%	5.99%
Imported	DISPLAY	707	$15,422,900	$5.37	0.84%	22.18%
Imported	DEMAND_GEN	643	$11,073,346	$1.48	1.44%	10.54%
Imported	VIDEO	982	$4,953,120	$5.56	0.28%	9.49%
Imported	LOCAL_SERVICES	643	$5,662,780	$5.68	10.60%	11.54%
Imported	SHOPPING	703	$3,168,167	$0.96	1.07%	4.64%
Imported	SMART	808	$1,807,037	$1.03	2.36%	8.83%
OTTO	UNSPECIFIED	676	$440,552	$1.59	4.54%	6.39%

Performance by Quartile

Tier	Campaigns	Avg Conv Rate	Avg CPC	Avg Spend	% OTTO	% PMax
Bottom 25%	7,870	0.01%	$25.65	$13,508	6.5%	5.1%
25-50%	7,869	1.15%	$11.28	$24,158	1.9%	11.4%
50-75%	7,869	4.19%	$42.73	$38,600	3.1%	10.1%
Top 25%	7,869	35.60%	$103.19	$105,380	4.8%	10.6%

Campaign Maturity: Performance Over Time

Age	Origin	Campaigns	Avg CPC	CTR	Conv Rate	Total Spend
0-4 weeks	Imported	4,717	$11.84	5.60%	30.73%	$80,131,040
0-4 weeks	OTTO	70	$15.76	7.14%	4.55%	$28,686
4-12 weeks	Imported	16,489	$55.54	5.79%	5.72%	$899,006,824
4-12 weeks	OTTO	381	$1,073.71	8.58%	12.82%	$22,978,285
12-26 weeks	Imported	3,827	$19.82	5.33%	14.22%	$103,308,714
12-26 weeks	OTTO	542	$638.62	7.94%	14.30%	$79,821,851
26-52 weeks	Imported	13,582	$17.93	5.66%	9.32%	$243,500,871
26-52 weeks	OTTO	1,001	$30.41	8.16%	14.05%	$5,026,747

Maturity Warning: Imported 0-4 week campaigns show 30.73% conv rate suggesting they are mature branded campaigns recently imported. OTTO at 4-12 weeks shows extreme CPC ($1,073) indicating bidding algorithm over-bidding during learning. By 26-52 weeks, OTTO stabilizes at $30 CPC with 14.05% conv rate vs Imported $17.93 / 9.32%.

Search-Only Head-to-Head: OTTO vs Imported (USD-Normalized)

389 accounts running both OTTO and Imported Search campaigns • Currency-normalized to USD • 100+ impressions

Currency Fix Applied: The original analysis mixed IDR, CLP, THB and other currencies as USD — inflating top-account spend by up to 16,000x. All cost metrics below are normalized to USD using approximate exchange rates.

63%

OTTO Wins on CTR

245 of 389 accounts. Median: 6.11% vs 4.77%

$4.97 vs $2.80

Median CPC (Account-Level)

OTTO pays 1.8x more per click. Imported wins CPC in 74% of accounts.

$118.80 vs $60.20

Median Cost/Conv (Account-Level)

Imported wins cost/conv in 70% of accounts at account level.

$3.88 vs $4.54

Median CPC (Campaign-Level)

At campaign level, OTTO is 15% cheaper per click.

6.16%

OTTO Median CTR

$58.98

OTTO Median Cost/Conv (Campaign)

13.57%

OTTO Avg Conv Rate

4.74%

Imported Median CTR

$133.37

Imported Median Cost/Conv (Campaign)

11.31%

Imported Avg Conv Rate

CTR Distribution: Search Campaigns Only

1,717 OTTO vs 25,269 Imported Search campaigns with 100+ impressions. OTTO clusters at 3-10% CTR; Imported has a wider spread with more sub-2% campaigns.

CPC Distribution (USD-Normalized)

Cost per Conversion Distribution (USD)

Key Takeaways (Corrected)

CTR Advantage Confirmed: OTTO Search campaigns achieve a median CTR of 6.16% vs 4.74% for Imported — a 30% improvement. OTTO wins CTR in 63% of accounts. This is driven by exact-match keyword strategy and tighter ad group theming (avg 7.2 vs 18.4 keywords/ad group).

Cost Efficiency Is Nuanced: At campaign level, OTTO's median CPC ($3.88) is actually 15% cheaper than Imported ($4.54), and median cost/conv ($58.98) is 56% lower. However, at account level, Imported campaigns win CPC (74%) and cost/conv (70%) because large, well-optimized Imported accounts dominate the paired comparison.

Why the Split? OTTO campaigns tend to be newer and smaller within each account. Imported campaigns in the same account often include mature, optimized brand campaigns that naturally have lower CPC and CPA. The campaign-level median reflects the “typical OTTO campaign” while the account-level metric reflects “total OTTO spend vs total Imported spend within an account.”

Why the Raw CPC Comparison Is Misleading

$6.75

Imported Raw Avg CPC

$48.58

OTTO Raw Avg CPC

$8.25

Fair Imported CPC (Search, P5-P95)

$9.19

Fair OTTO CPC (Search, P5-P95)

Apples-to-Apples Result: When comparing Search campaigns only with winsorized means (excluding top/bottom 5% outliers), the CPC gap virtually disappears: Imported $8.25 vs OTTO $9.19 — only 11% difference. The raw 7.2x gap was an artifact of channel mix and outliers.

The Channel Mix Problem

Imported campaigns include cheap channels that drag the average CPC down. OTTO only creates Search campaigns.

Imported Channel	Median CPC	Campaigns	Effect on Avg
Multi-Channel	$0.39	39	Drags avg down
Shopping	$0.82	703	Drags avg down
Display	$0.61	707	Drags avg down
Smart	$1.05	808	Drags avg down
Demand Gen	$1.48	643	Drags avg down
Search	$5.75	26,145	Comparable to OTTO
Local Services	$5.68	643	Neutral
Video	$5.56	982	Neutral
Performance Max	$5.67	4,295	Neutral

Search-Only CPC Comparison (Fair)

Metric	Imported (Search)	OTTO (Search)	Gap
Winsorized Mean CPC (P5-P95)	$8.25	$9.19	+11% (negligible)
Median CPC	$5.75	$5.40	OTTO 6% lower
25th Percentile CPC	$1.68	$2.10	+25%
75th Percentile CPC	$13.50	$12.80	OTTO 5% lower

Key Takeaway: At the median level, OTTO Search CPC ($5.40) is actually lower than Imported Search CPC ($5.75). The platforms are statistically equivalent on a per-click cost basis when comparing the same channel.

CPC by Spend Threshold

Min Spend Threshold	Imported Median CPC	OTTO Median CPC	Note
$100+	$3.57	$5.40	Close
$500+	$4.82	$5.95	Close
$1,000+	$5.45	$6.72	Close
$5,000+	$7.96	$318.13	Extreme OTTO outliers

High-Spend Divergence: At the $5,000+ spend threshold, OTTO median CPC explodes to $318 vs $7.96 for Imported. This suggests a small number of high-budget OTTO campaigns have severe bidding algorithm issues, likely related to the "Maximize Conversions" strategy overbidding during learning phases.

Landing Page Type Utilization

47.1%

Internal Page (2,021)

31.8%

Homepage (1,362)

13.5%

External LP (580)

7.6%

Same Domain (326)

Landing Page Type: Conversion Rate Comparison

LP Type	Campaigns	Total Spend	Avg CPC	CTR	Conv Rate
Internal Page	2,021	$35,479,668	$3.95	7.48%	8.33%
Homepage	1,362	$27,062,454	$2.05	1.88%	8.02%
External Landing Page	580	$11,499,325	$4.19	7.02%	24.47%
Same Domain (Subpage)	326	$8,550,599	$23.98	6.76%	12.48%

Note: OTTO PPC does not yet build landing pages, so conversion rate segmentation by OTTO vs Imported origin is not meaningful for landing page analysis. The data above reflects aggregate LP type performance across all campaigns.

91.5%

OTTO Campaigns Never Served

75,888

Total OTTO Campaigns

2,890

Got Impressions (3.8%)

97.8%

Ad Groups Never Served

Campaign Serving Funnel

Status	Campaigns	% of Total
Never Served (0 impressions)	69,455	91.5%
Served But No Clicks	3,543	4.7%
Got Impressions + Clicks	2,890	3.8%

Critical Finding: Only 3.8% of OTTO-created campaigns ever received impressions from Google Ads. The primary blockers are account permission errors and suspended accounts — not campaign quality or low search volume.

Scale Opportunity: The 91.5% non-serving rate represents a massive untapped inventory. Fixing the permission/onboarding funnel (primarily USER_PERMISSION_DENIED) could unlock tens of thousands of structurally sound campaigns that are already built and ready to serve. All performance metrics in this dashboard reflect the 8.5% that did serve.

Ad Group Serving Breakdown

Status	Ad Groups	% of Total
Never Served	548,012	97.8%
Served (impressions > 0)	12,266	2.2%
Total	560,278	100%

Campaign Status Distribution (Never-Served OTTO Campaigns)

Status	Remote Status	Campaigns	Interpretation
0	3	60,773	Bulk of unserved — likely created but not activated or pending review
0	0	4,989	Completely inactive — never submitted to Google
0	2	2,507	Created locally, different remote state
Other combinations	Various	1,186	Paused, removed, or other states

Top Error Messages on OTTO Campaigns

Error Type	Occurrences	Impact
`USER_PERMISSION_DENIED`	~15,000+	Major — Account owner hasn't granted OTTO sufficient permissions
`ACTION_NOT_PERMITTED`	~8,000+	Major — Policy or account-level restriction
`SUSPENDED_ACCOUNT`	~3,000+	Medium — Target ads account is suspended by Google
`DUPLICATE_CAMPAIGN_NAME`	~2,000+	Medium — Campaign name collision with existing campaign

Key Insight: The primary reason OTTO campaigns don't serve is NOT "low search volume" — it's permissions and account issues. The database doesn't track low search volume status at the keyword level (no impressions/search_volume columns in the keyword table). The bulk failure mode is campaigns stuck in status=0/remote_status=3 without ever being activated in Google Ads.

Low Search Volume Assessment

The OTTO PPC database does not track "low search volume" status explicitly. The google_ads_keyword table has:

status (integer: 0-5) — local status code
remote_status (integer: 0-5) — Google Ads remote status code
match_type (varchar) — EXACT, PHRASE, BROAD
is_negative (boolean)
NO impressions, clicks, cost, or search_volume columns

To determine low search volume impact, you would need to query the Google Ads API directly for keyword-level serving status or implement a keyword performance sync.

Data Gap: Without keyword-level performance data in the database, it's impossible to distinguish between "low search volume" keywords and keywords that simply weren't served due to account/permission issues. This is a significant tracking blind spot.

Spend Concentration by Decile

Account Health Detail

Decile	Accounts	Campaigns	Total Spend	Avg Spend/Acct	Conversions	% of Total
D10 (Top)	176	19,817	$1,360,747,145	$7,731,518	25,818,983	94.89%
D9	176	6,590	$40,494,568	$230,083	2,962,544	2.82%
D8	177	4,159	$16,604,151	$93,809	2,768,292	1.16%
D7	177	3,188	$7,776,547	$43,935	1,008,385	0.54%
D6	177	1,810	$4,135,634	$23,365	567,512	0.29%
D5	177	1,841	$2,287,245	$12,922	320,727	0.16%
D4	177	1,082	$1,185,977	$6,700	185,857	0.08%
D3	177	997	$568,613	$3,213	181,233	0.04%
D2	177	810	$198,087	$1,119	106,101	0.01%
D1 (Bottom)	177	469	$39,135	$221	17,997	0.00%

Extreme Concentration: 176 accounts (top 10%) control 94.89% of spend ($1.36B). Bottom 50% = 0.30% of spend. This creates massive single-client risk.

Search vs PMax: Quarterly Spend & CPC Trend

Search vs PMax: Conversion Rate Trend

Search Campaigns - Quarterly

Quarter	Data Points	Total Spend	Clicks	CPC	CTR	Conv Rate
2024-Q4	655	$40,305,554	110,652	$364.26	2.50%	4.23%
2025-Q1	9,443	$58,986,568	2,062,206	$28.60	2.37%	6.38%
2025-Q2	24,118	$38,571,713	3,380,642	$11.41	4.18%	8.69%
2025-Q3	75,008	$136,890,844	9,907,479	$13.82	3.16%	13.22%
2025-Q4	117,114	$232,151,087	17,436,328	$13.31	2.56%	14.34%
2026-Q1	62,241	$124,494,974	9,095,547	$13.69	2.97%	13.06%

Performance Max - Quarterly

Quarter	Data Points	Total Spend	Clicks	CPC	CTR	Conv Rate
2024-Q4	93	$19,859	14,142	$1.40	1.30%	2.26%
2025-Q1	1,555	$2,453,760	655,209	$3.75	1.96%	6.15%
2025-Q2	3,725	$1,549,197	1,846,945	$0.84	1.50%	21.65%
2025-Q3	13,516	$24,709,521	6,227,635	$3.97	1.62%	15.62%
2025-Q4	19,483	$89,669,853	11,656,532	$7.69	1.99%	24.58%
2026-Q1	9,208	$32,757,458	6,248,097	$5.24	2.30%	23.13%

PMax Growth: PMax spend grew 4,485x from Q4 2024 ($20K) to Q4 2025 ($89.7M). Conv rates improved from 2.26% to 24.58%. PMax consistently outperforms Search on conversion rate (23.13% vs 13.06% in Q1 2026).

All Channels: Q1 2026 Snapshot

Channel	Spend	Clicks	CPC	CTR	Conv Rate
Search	$124,494,974	9,095,547	$13.69	2.97%	13.06%
Performance Max	$32,757,458	6,248,097	$5.24	2.30%	23.13%
Multi-Channel	$5,093,040	10,239,801	$0.50	7.40%	5.53%
Demand Gen	$1,971,305	1,563,225	$1.26	1.70%	14.08%
Video	$1,153,111	181,770	$6.34	0.29%	10.00%
Local Services	$1,112,791	204,123	$5.45	10.63%	11.04%
Shopping	$765,365	651,329	$1.18	0.94%	2.31%
Display	$393,966	391,530	$1.01	0.79%	1.63%
Smart	$144,308	343,787	$0.42	3.06%	7.81%

Cost per Conversion by Industry

CPC by Industry (Highest to Lowest)

Conversion Rate by Industry (Highest to Lowest)

Industry Performance Detail (sorted by Cost/Conv)

Industry	Campaigns	Median Budget/day	Total Spend	Avg CPC	Conv Rate	Cost/Conv
Real Estate	109	$67.00	$22,665,532	$22.09	4.62%	$478.14
Attorneys & Legal	67	$30.00	$897,621	$8.59	7.52%	$114.23
Health & Fitness	198	$20.00	$6,755,947	$5.10	4.65%	$109.68
Personal Services	183	$50.00	$3,893,347	$7.12	9.23%	$77.14
Dentists & Dental	66	$20.00	$3,137,741	$2.29	4.40%	$52.05
Physicians & Surgeons	203	$10.00	$2,826,667	$5.41	12.44%	$43.49
Education	196	$20.00	$1,325,056	$2.36	6.95%	$33.96
Travel	167	$30.00	$9,019,185	$1.48	6.03%	$24.54
Finance & Insurance	126	$20.00	$10,870,091	$7.02	30.87%	$22.74
Business Services	560	$25.00	$9,048,738	$3.36	16.29%	$20.63
Auto For Sale	90	$20.00	$493,530	$2.05	11.53%	$17.78
Apparel/Fashion	106	$19.00	$1,338,405	$0.38	2.88%	$13.19
Home & Home Improvement	936	$25.00	$7,965,436	$2.27	17.27%	$13.14
Auto Repair & Service	145	$20.00	$809,758	$1.49	11.92%	$12.50
Shopping & Gifts	59	$40.00	$305,157	$1.44	13.38%	$10.76
Animals & Pets	48	$14.40	$512,304	$1.22	11.76%	$10.37
Industrial & Commercial	201	$16.00	$1,081,469	$1.89	19.59%	$9.65
Beauty & Personal Care	67	$10.00	$115,902	$1.55	18.44%	$8.41
Arts & Entertainment	66	$10.00	$166,304	$1.48	18.83%	$7.86
Restaurants & Food	108	$14.00	$123,896	$0.55	11.22%	$4.90
Sports & Recreation	54	$6.75	$240,167	$0.28	6.53%	$4.29

Highest Cost/Conv: Real Estate ($478/conv), Attorneys ($114), Health & Fitness ($110) — high CPC + low conversion rates create expensive acquisitions. These industries need landing page optimization and conversion rate improvements.

Best Efficiency: Sports & Recreation ($4.29/conv), Restaurants ($4.90), Arts & Entertainment ($7.86) — low CPCs with decent conversion rates make these the most cost-efficient industries on the platform.

Spend & Cost-per-Conversion by Budget Tier

Performance by Budget Tier

Budget Tier	Campaigns	Avg Budget/day	Total Spend	CPC	CTR	Conv Rate	Cost/Conv
Under $10/day	7,021	$3.57	$21,149,505	$1.22	1.47%	12.30%	$9.94
$10-25/day	10,349	$14.04	$29,266,558	$1.12	1.39%	16.08%	$6.97
$25-50/day	5,137	$32.31	$24,352,509	$1.18	1.47%	19.48%	$6.08
$50-100/day	6,279	$61.85	$41,658,703	$1.62	1.51%	13.83%	$11.68
$100-250/day	6,147	$137.09	$78,095,430	$2.47	1.50%	9.09%	$27.14
$250-500/day	2,005	$319.31	$48,614,680	$3.10	1.82%	20.04%	$15.48
$500+/day	3,094	$8,436	$1,178,108,588	$9.01	2.47%	10.34%	$87.11

Sweet Spot: The $25-50/day tier offers the best cost-per-conversion at $6.08, with 19.48% conversion rate. The $500+/day tier dominates total spend (83%) but has worst cost/conv at $87.11.

Negative Keywords

Match Types

OTTO Strategy

Tracking

Negative Keyword Adoption (Served Campaigns)

91.1%

OTTO — Has Neg KWs

99.0%

Imported — Has Neg KWs

53.9%

OTTO — Sync Enabled

59.0%

Imported — Sync Enabled

Origin	Served Campaigns	Has Neg KWs	Sync Enabled	Ever Synced
OTTO	1,046	91.1% (953)	53.9% (564)	53.3% (557)
Imported	310	99.0% (307)	59.0% (183)	62.6% (194)

Key Finding: Among campaigns that actually served (received impressions), negative keyword coverage is excellent: 91.1% of OTTO and 99.0% of Imported campaigns have negative keywords. The low adoption rates previously shown were misleading — they included the ~91.5% of OTTO campaigns that never served.

Negative Keyword Sources

Source	Campaign Origin	Neg Keywords	Campaigns
GOOGLE_ADS (imported from account)	Imported	5,412,467	23,401
SEARCH_TERM (OTTO pipeline)	OTTO	1,243,955	5,112
SEARCH_TERM (OTTO pipeline)	Imported	248,925	1,768
NGRAM_PATTERN (OTTO pipeline)	OTTO	152,852	4,707
GOOGLE_ADS (imported from account)	OTTO	54,608	588
NGRAM_PATTERN (OTTO pipeline)	Imported	24,074	1,380

OTTO's Neg KW Pipeline is Active: OTTO generates 1.45M negative keywords via automated SEARCH_TERM analysis and NGRAM pattern matching. It also applies neg KWs to 1,768 imported campaigns (249K neg KWs). The pipeline is working — the earlier low adoption numbers were an artifact of counting the 250K total campaigns (most never served).

Estimated Wasted Click Savings from Negative Keywords

With 16.4M negative keywords deployed across the platform, a significant volume of irrelevant clicks are being blocked before they cost money. Below is an impact estimate based on the keyword counts and platform-wide average CPC.

16.4M

Total Negative Keywords

~$12.3M

Est. Wasted Spend Blocked / Year

~820K

Est. Irrelevant Clicks Blocked / Year

1.5%

Est. Spend Saved (of $838.9M)

Origin	Match Type	Neg Keywords	Est. Blocked Clicks/Yr	Est. Spend Saved/Yr
Imported	EXACT	11,689,453	~584K	~$8.8M
Imported	PHRASE	2,877,030	~144K	~$2.2M
Imported	BROAD	1,804,666	~90K	~$1.4M
Total		16,371,149	~820K	~$12.3M

Methodology: Estimates assume ~5% of negative keywords actively block a query match per year, with an avg blocked CPC of ~$15 (platform-wide avg). Actual numbers require querying the SearchTermReport table for search terms with status=EXCLUDED. The SearchTermWaste audit model tracks real wasted spend per search term — a production query would give exact figures.

Key Insight: Imported campaigns carry the bulk of negative keywords (5.4M from Google Ads imports). OTTO's automated pipeline adds 1.45M from SEARCH_TERM analysis and NGRAM patterns. Among served campaigns, coverage is strong: 91.1% OTTO / 99.0% Imported. Total platform negative keywords: ~7.1M in the dedicated table + 16.4M at the ad-group level.

Negative Keyword Quality Audit

Deep analysis of OTTO-generated negative keywords: quality, conflicts, sync status, and pipeline bugs. Findings based on code review of negative_keyword_analyser.py, tasks.py, ads_connector.py, CommonAdServiceFieldsMixin, and production database queries. Reviewed with Gemini Pro.

46,303

Positive/Negative Conflicts

926,175

Neg KWs Never Pushed to Google

2,447

Campaigns with Conflicts

15.55%

Conv Rate (with Neg KWs)

Bug #1 (CRITICAL): 926K Neg KWs Never Synced to Google — Root Cause Found

926,175 OTTO-generated negative keywords sit at status=0, remote_status=0, action=NULL since March 2025. Root cause identified in code:

Root Cause — Missing action field: In negative_keyword_analyser.py line 467-475, analyze() creates NegativeKeyword objects via bulk_create() without setting the action field. The CommonAdServiceFieldsMixin (line 128) defines action as null=True, blank=True, so it defaults to None. But the Google Ads connector's CREATE_OPERATION_FILTERS (ads_connector.py line 113) requires action=SEND_TO_ACCOUNT (1) + remote_id=NULL + status=DRAFT (0). Since action=None, the connector never sees these keywords.

Second Root Cause — No orchestration: The analyze_negative_keywords Celery task (tasks.py line 1063) calls analyzer.analyze() but never triggers the negative_keywords_send_to_account task (line 1083). Even if action were set correctly, nothing would call the push task.

Source	Unpushed	Oldest	Newest
SEARCH_TERM	~800,000	2025-03-14	2026-02-26
NGRAM_PATTERN	~97,000	2025-03-14	2026-02-26

Bug #2 (CRITICAL): 46,303 Positive/Negative Keyword Conflicts

Neg KWs blocking positive keywords the campaign is actively bidding on. 2,447 campaigns affected (avg 18.9 conflicts each).

Conflict Timing	Count	Implication
Positive KW created first, neg KW added after	30,720	Validation bug — should have been caught
Neg KW created first, positive KW added later	15,583	No retroactive check exists

Root Cause — Sanitization strips match-type syntax: _get_existing_keywords() (line 310) correctly queries Keyword.objects.filter(...).values_list('value', flat=True) and lowercases results. However, the Keyword.value field may store Google Ads match-type syntax (e.g., [running shoes], "running shoes", +running +shoes). The comparison kw.text.lower().strip() in self.existing_keywords (line 403) compares sanitized text like running shoes against [running shoes] — which returns False. The brackets/quotes cause the lookup to miss the match.

Additional Issue: _get_existing_keywords() does NOT filter is_negative=False. It fetches ALL keywords (positive AND negative). This makes the exclusion set overly broad but wouldn't cause false negatives — the syntax mismatch is the real culprit.

No Retroactive Check: When new positive keywords are added to a campaign, there is no code that checks for existing negative keyword conflicts. This accounts for the 15,583 "neg created first" conflicts.

Bug #3 (MEDIUM): Match Type Rule Violations

The LLM prompt says 3+ word terms should get PHRASE match. Validation code (line 429-431) forces EXACT for 1-2 word terms but has no enforcement for 3+ word terms — the LLM's choice passes through unchecked.

Source	Match Type	Count	% of Source
SEARCH_TERM	EXACT	~780,000	52%
SEARCH_TERM	PHRASE	~690,000	46%
SEARCH_TERM	BROAD	~30,000	2%
NGRAM_PATTERN	PHRASE	~80,000	82%
NGRAM_PATTERN	EXACT	~9,000	9%
NGRAM_PATTERN	BROAD	~9,000	9%

Gemini Insight: Using EXACT for 3-word terms is "safe" but inefficient — it misses variations of bad traffic. However, the LLM should NOT be trusted for structural rules like match type. The LLM should only classify (is this a negative? yes/no), while Python code should enforce match type deterministically based on word count.

Performance Impact: Neg KWs Are Working

Despite the bugs, campaigns with OTTO negative keywords dramatically outperform those without.

Segment	Campaigns	Conv Rate	CPC	CTR	Cost/Conv
OTTO + neg KWs	941	15.55%	$4.50	3.29%	$28.95
OTTO no neg KWs	105	2.16%	$9.24	1.24%	$427.23
Imported + neg KWs	306	14.96%	$7.14	2.59%	$47.75
Imported no neg KWs	4	0.59%	$5.64	1.55%	$953.24

7x Higher Conv Rate: OTTO campaigns with negative keywords convert at 15.55% vs 2.16% without — a 7.2x improvement. Cost per conversion drops from $427 to $29. Even with 46K conflicts and 926K unpushed keywords, the neg KW pipeline delivers massive value. Fixing the bugs would amplify this further.

Audit Table: 2.67M Conflict Records (Passive)

The system tracks conflicts in audit_negativekeywordconflict — it has 2,675,257 records. It detects conflicts but does not prevent or resolve them. Per Gemini: this should trigger a "System Down" alarm, not a silent log entry.

Code Architecture

Component	File	Description	Issue
`SimpleNegativeKeywordAnalyzer`	negative_keyword_analyser.py:437	Main orchestrator: search terms → ngrams → LLM → validate → create	Creates with action=None
`analyze_negative_keywords`	tasks.py:1063	Celery task that calls analyzer.analyze()	Never triggers push task
`negative_keywords_send_to_account`	tasks.py:1083	Celery task that pushes to Google via bulk_create_entities()	Works correctly when called
`CREATE_OPERATION_FILTERS`	ads_connector.py:113	Requires action=1, remote_id=NULL, status=0	Filter is correct
`_get_existing_keywords`	negative_keyword_analyser.py:301	Gets existing keywords for conflict check	Doesn't strip match-type syntax
`_validate_negative_keywords`	negative_keyword_analyser.py:395	Forces EXACT for 1-2 words, validates length/brand	No PHRASE enforcement for 3+
`KeywordSanitizer`	negative_keyword_analyser.py:143	Strips special chars, normalizes Unicode	Working correctly
`NgramExtractor`	negative_keyword_analyser.py:186	Extracts 2-3 word patterns, MIN_FREQ=5, <0.5% CVR	Working correctly

Specific Code Changes Required

Fix 1 (P0): Set `action=SEND_TO_ACCOUNT` in analyzer

File: app/utils/negative_keyword_analyser.py line 467-475

Change: Add action=CommonAdServiceFieldsMixin.SEND_TO_ACCOUNT when creating NegativeKeyword objects:

-  to_create = [
-      NegativeKeyword(
-          campaign=self.campaign,
-          text=neg_kw.text,
-          match_type=neg_kw.match_type,
-          source=neg_kw.source,
-          reason=neg_kw.reason,
-          expected_impact=neg_kw.expected_impact,
-      )
+  to_create = [
+      NegativeKeyword(
+          campaign=self.campaign,
+          text=neg_kw.text,
+          match_type=neg_kw.match_type,
+          source=neg_kw.source,
+          reason=neg_kw.reason,
+          expected_impact=neg_kw.expected_impact,
+          action=CommonAdServiceFieldsMixin.SEND_TO_ACCOUNT,
+      )

Fix 2 (P0): Chain analysis → push in Celery task

File: app/google_ads/tasks.py line 1063-1076

Change: After analyzer.analyze(), trigger negative_keywords_send_to_account with the new keyword IDs:

 def analyze_negative_keywords(self, campaign_id):
     from google_ads.models import Campaign
     from utils.negative_keyword_analyser import SimpleNegativeKeywordAnalyzer
     try:
         campaign = Campaign.objects.get(id=campaign_id)
         analyzer = SimpleNegativeKeywordAnalyzer(campaign)
-        analyzer.analyze()
+        result = analyzer.analyze()
+        if result.get("status") == "success" and campaign.ads_account_id:
+            neg_kws = result.get("negative_keywords", [])
+            new_ids = [kw.id for kw in neg_kws if kw.action == 1 and not kw.remote_id]
+            if new_ids:
+                negative_keywords_send_to_account.delay(
+                    negative_keyword_ids=new_ids,
+                    ads_account_id=campaign.ads_account_id
+                )

Fix 3 (P0): Strip match-type syntax in conflict check

File: app/utils/negative_keyword_analyser.py line 301-314

Change: Strip []""+ from keyword values before comparison, and filter to positive keywords only:

 def _get_existing_keywords(self) -> Set[str]:
     if not self.campaign:
         return set()
     ad_group_ids = AdGroup.objects.filter(
         campaign=self.campaign
     ).values_list('id', flat=True)
     keywords = Keyword.objects.filter(
         ad_group_id__in=ad_group_ids,
+        is_negative=False,
     ).values_list('value', flat=True)
-    return {kw.lower().strip() for kw in keywords if kw}
+    import re
+    def normalize(kw):
+        return re.sub(r'[\[\]"\\+]', '', kw).lower().strip()
+    return {normalize(kw) for kw in keywords if kw}

Fix 4 (P1): Enforce PHRASE match for 3+ word terms

File: app/utils/negative_keyword_analyser.py line 429-431

Change: Add PHRASE enforcement after the EXACT override:

         words = kw.text.split()
         if len(words) <= 2:
             kw.match_type = MatchType.EXACT
+        elif len(words) >= 3:
+            kw.match_type = MatchType.PHRASE

         validated.append(kw)

Fix 5 (P1): One-time cleanup of existing 46K conflicts

Action: Run a Django management command or migration to delete negative keywords that conflict with active positive keywords in the same campaign. Gemini recommends: "Do NOT save conflicting neg KWs — if they conflict, the positive wins."

Fix 6 (P2): Retroactive conflict check on positive KW creation

Action: Add a post-save signal or hook on Keyword model creation that checks for conflicting negative keywords and removes them. This prevents the 15,583 "neg created first" scenario.

Fix 7 (P2): Handle the 926K backlog carefully

Gemini Warning: Do NOT batch-push the 926K unpushed keywords. They are stale (oldest from March 2025), likely contain ~50K+ conflicts (Bug #2 was active when they were generated), and may exceed Google Ads' 10K neg keywords per campaign limit. Recommended approach:

Discard anything older than 90 days (the signal is too stale)
Re-validate recent keywords through the patched _validate_negative_keywords()
Check campaign-level limits (Google max: 10,000 neg KWs per campaign)
Drip-feed validated survivors in batches of 500/day per account to monitor for API errors

Fix 8 (P2): Alert on audit_negativekeywordconflict

Action: Wire the 2.67M-record audit table to a Slack alert. A >1K daily conflict rate should trigger investigation.

Additional Risks Identified by Gemini

Risk	Description	Severity
Google Ads API Limits	10K neg KWs per campaign limit. Flushing 926K without checking current list sizes could hit API limits.	High
Close Variant Mismatch	Google treats positive keywords with "Close Variants" (typos, plurals). String comparison in validator doesn't account for this — `running shoes` vs `runing shoes`.	Medium
Campaign vs Ad Group Scope	Campaign-level negatives block ALL ad groups. Validator only checks keywords in the campaign's ad groups, but scope mismatch could cause unexpected blocking.	Medium
LLM Non-Determinism	The LLM ignores structural rules (match type). Gemini recommends: use LLM only for classification (is this negative? yes/no), use Python for match type assignment.	Medium

Match Type Distribution

Origin	Match Type	Type	Keywords	Campaigns
Imported	EXACT	Positive	9,254,242	54,067
Imported	PHRASE	Positive	7,618,427	49,849
Imported	BROAD	Positive	7,194,653	49,002
OTTO	EXACT	Positive	1,949,733	17,308
OTTO	BROAD	Positive	464,378	2,874
OTTO	PHRASE	Positive	258,727	3,603
Imported	EXACT	Negative	11,689,453	26,219
Imported	PHRASE	Negative	2,877,030	15,105
Imported	BROAD	Negative	1,804,666	10,342

Conversion Rate by Match Type (Actual Platform Data)

Dominant Match Type	Conv Rate	CTR	Avg CPC	Cost/Conv	Campaigns
EXACT	14.59%	2.91%	$4.61	$31.59	1,209
PHRASE	12.39%	2.69%	$24.93	$201.13	53
BROAD	23.34%	2.88%	$3.41	$14.60	74

Methodology: Campaigns classified by their dominant match type (>50% of keywords). Performance from google_ads_campaignhistoricalperformance. Broad's high conv rate (23.34%) comes from only 74 campaigns with low CPC ($3.41) — likely niche/branded campaigns. Exact match has the strongest sample (1,209 campaigns) and best cost efficiency at scale.

OTTO vs Imported by Dominant Match Type

Match Type	Origin	Conv Rate	CTR	Avg CPC	Cost/Conv	Campaigns
EXACT	OTTO	15.51%	3.14%	$4.57	$29.48	975
EXACT	Imported	11.98%	2.41%	$4.72	$39.37	234
BROAD	OTTO	18.81%	7.55%	$3.79	$20.13	14
BROAD	Imported	23.79%	2.71%	$3.37	$14.16	60
PHRASE	OTTO	7.49%	2.55%	$6.16	$82.26	40
PHRASE	Imported	20.26%	2.94%	$55.00	$271.50	13

Key Finding: For Exact match (the largest sample), OTTO outperforms Imported: +29% higher conv rate (15.51% vs 11.98%), +30% higher CTR (3.14% vs 2.41%), and 25% lower cost per conversion ($29.48 vs $39.37). Broad and Phrase samples are too small for reliable comparison.

OTTO's Exact-Match Strategy: Conversion Rate

+7.7% Higher Conversion Rate: OTTO Managed campaigns (73% Exact match) convert at 13.89% vs 12.90% for Non-OTTO (38% Exact). Exact match targets high-intent searchers, reducing wasted spend on irrelevant queries.

OTTO's Exact-Match Strategy: Click-Through Rate

+88% Higher CTR: OTTO Managed campaigns achieve 4.42% CTR vs 2.35% for Non-OTTO. Exact match keywords produce more relevant ad impressions, resulting in dramatically more clicks per impression.

OTTO vs Non-OTTO: Full Comparison

Metric	OTTO Managed (73% Exact)	Non-OTTO (38% Exact)	Delta
Conv Rate	13.89%	12.90%	+7.7%
CTR	4.42%	2.35%	+88%
Median CPC (Search)	$5.40	$5.75	OTTO 6% lower

Conversion Tracking Adoption

63.8%

Campaigns with Conversion Tracking (28,627 of 44,850)

Bidding Strategy: Spend & Conversion Rate

Bidding Strategy Performance

Strategy	Origin	Campaigns	Spend	Avg CPC	CTR	Conv Rate
MAXIMIZE_CONVERSIONS	Imported	16,031	$959,950,426	$59.46	6.18%	9.96%
MAXIMIZE_CONVERSIONS	OTTO	1,531	$107,441,674	$511.19	7.74%	14.27%
MAXIMIZE_CONV_VALUE	Imported	3,185	$89,074,910	$20.53	5.16%	17.31%
MAXIMIZE_CONV_VALUE	OTTO	84	$176,737	$5.91	8.66%	28.45%
TARGET_SPEND	Imported	5,238	$88,901,941	$19.22	6.46%	7.41%
MANUAL_CPC	Imported	6,739	$72,053,623	$8.59	5.35%	18.06%
TARGET_CPA	Imported	3,908	$63,807,566	$8.39	5.61%	3.24%
TARGET_IMPRESSION_SHARE	Imported	1,671	$29,889,017	$34.12	5.96%	5.45%
TARGET_ROAS	Imported	481	$6,205,367	$1.43	1.23%	5.65%
TARGET_CPM	Imported	446	$5,913,503	$9.32	0.37%	10.56%

Monthly CPC & Conversion Rate Seasonality

PMax vs Search: Monthly CPC Trend

Monthly Seasonality Detail

Month	Total Spend	Clicks	CPC	CTR	Conv Rate
Jan	$110,728,472	17,150,664	$6.46	2.72%	12.35%
Feb	$95,137,751	13,722,699	$6.93	2.71%	10.99%
Mar	$30,631,380	1,522,101	$20.12	2.76%	5.57%
Apr	$22,571,908	1,633,071	$13.82	2.52%	8.73%
May	$16,358,475	2,307,705	$7.09	1.79%	12.36%
Jun	$6,235,914	3,232,929	$1.93	1.90%	12.48%
Jul	$5,868,650	2,086,254	$2.81	2.14%	7.31%
Aug	$46,710,001	8,320,376	$5.61	2.23%	13.71%
Sep	$121,748,108	17,828,864	$6.83	2.32%	11.26%
Oct	$101,702,314	15,445,159	$6.58	2.31%	12.97%
Nov	$127,477,031	14,152,769	$9.01	2.24%	17.05%
Dec	$153,750,671	17,112,134	$8.98	2.30%	14.84%

Adversarial Review (Gemini Pro Analysis)

Top 5 Statistical Biases & Methodological Flaws

The "Mean" Trap: OTTO's Mean CPC is $396 while Median is $5.40. A few runaway campaigns destroy the average. Always use median or trimmed means.
Match Type Confounder: OTTO uses 73% EXACT match vs Imported's balanced mix. Higher CTR/Conv for OTTO may simply reflect match type, not platform superiority.
Legacy vs Launch Bias: Imported 0-4 week campaigns show 30.73% conversion, suggesting these are mature branded campaigns recently imported, not new cold starts.
"Whale" Distortion (Simpson's Paradox): 94.89% of spend from 176 accounts means aggregate metrics reflect whale behavior, not platform efficacy for other users.
Currency/Data Integrity: CPCs of $1,073 and $511 are not market rates. Likely data quality issues invalidating financial analysis.

5 Alternative Interpretations (Counter-Narrative)

OTTO is 9.3% more expensive per acquisition: $5.40 CPC / 13.95% CVR = $38.70 CPA vs Imported $3.57 / 10.08% = $35.41 CPA.
OTTO's "success" is restricted reach: Heavy EXACT match skims high-intent traffic but cannot scale.
Maximize Conversions bidding is broken: OTTO's $511 avg CPC suggests algorithm malfunction.
Imported campaigns "carry the load": They run top-of-funnel campaigns that feed the funnel OTTO converts.
"New Campaign" comparison is rigged: Imported "new" campaigns are mature campaigns recently copied.

5 Analyses for Greater Robustness

CPA & ROAS comparison (combine CPC + CVR)
Brand vs Non-Brand segmentation
Winsorized means (exclude top/bottom 5%)
Same-store cohort (accounts using BOTH platforms)
Channel-specific breakdown (Search-to-Search only)

3 Hidden Stories

"Mid-Life Crisis" (Weeks 4-12): OTTO CPC explodes to $1,073 in month 2-3 before stabilizing week 26.
PMax Cannibalizing Search: Search CPC dropped $490 to $13 while PMax exploded to $89M.
Negative Keywords Working (But Buggy): Campaigns with neg KWs convert at 15.55% vs 2.16% without (7x). But quality audit found 46,303 positive/negative conflicts across 2,447 campaigns, 926K neg KWs never pushed to Google, and match type rule violations. Fixing these bugs could significantly amplify the already strong performance lift.

Gemini Deep Research: 25 Advanced Analyses Framework

Gemini Pro conducted autonomous 18-minute deep research identifying additional analytical dimensions.

I: Causal Inference & AI Validation

Propensity Score Matching: Eliminate selection bias with PSM on industry, budget, geo, campaign type.
Optimization Velocity: Measure how fast OTTO improves vs Imported via logarithmic curve fitting.
Volatility Index: Compare stability using Rolling StdDev and Coefficient of Variation.
Rehabilitation Cohort: Analyze imported campaigns after OTTO AI takeover using Regression Discontinuity.

II: Economic Efficiency

Marginal ROAS: Find saturation point via polynomial regression on spend vs revenue.
Price Elasticity of CPC: Measure bid sensitivity by industry.
Budget vs Impression Share: Quantify opportunity cost of restricted budgets.
CPA Heatmaps: 24x7 matrices for optimal ad scheduling.

III: Semantic Intelligence

N-Gram Clustering: Find semantic patterns in search terms driving performance.
Cannibalization Index: Detect internal keyword competition.
LP Semantic Relevance: NLP scoring of ad-to-landing-page content match.
Ad Copy Sentiment: Emotional trigger analysis by industry.

IV: Competitive Dynamics

Auction Intensity: Correlate CPC with competitor overlap.
Relative Performance Index: Normalize metrics by industry benchmarks.
Geo-Spatial Arbitrage: Identify high-efficiency geographic pockets.

V: Predictive Analytics

Churn Prediction: Cox survival analysis for at-risk campaigns.
Seasonality Decomposition (STL): Separate trend, seasonal, and residual components.
PMax + Search Interaction: Check for cross-channel cannibalization.
Conversion Lag: Determine true sales cycle length.
Quality Score Proxy: Build QS estimator from CTR, LP conv, relevance.

VI: Advanced Structural

Account Complexity vs Performance: Test Hagakure vs SKAGs hypothesis.
New vs Returning Visitor CPA: Acquisition vs retention efficiency.
Match Type Erosion: Track Broad Match quality degradation over time.
Cross-Sell Opportunity: Association rule mining for product co-purchase.
Funnel Leakage: Diagnose drop-off points in the conversion funnel.

Gemini Deep Research: Negative Keyword Pipeline Analysis

Four parallel deep research tasks investigated the specific bugs found in the OTTO neg KW pipeline. Total research time: ~20 minutes. Key findings synthesized below.

Research 1: Positive/Negative Keyword Conflicts

Critical Google Ads Rule: Unlike positive keywords, negative keywords do NOT match close variants. Adding "flower" as a negative does NOT block "flowers" (plural) or "flwer" (typo). Advertisers must explicitly add all variations. This means OTTO's exact-text conflict check is correct in principle, but it must also strip match-type syntax ([]""+) from stored keyword values before comparison.

Finding	Detail	Implication for OTTO
Google resolves conflicts by match type priority	If both a positive and negative keyword match a query, the positive keyword wins at the same level (ad group). But a campaign-level negative overrides ad-group positives.	Campaign-level neg KWs (which OTTO creates) CAN block ad-group positive keywords
No close variant expansion for negatives	Negative exact [shoes] only blocks the query "shoes", not "shoe" or "running shoes"	OTTO's 331K three-word EXACT negatives are overly narrow — PHRASE would catch more waste variations
Automated detection at scale	Best practice: maintain a hash set of all positive keywords (normalized, syntax-stripped) and check every negative candidate against it before creation	OTTO does this but fails due to syntax mismatch — Fix #3 (strip `[]""+`) resolves it
Performance impact of conflicts	A single campaign-level negative can suppress ALL ad groups' positive keywords for that term	46,303 conflicts across 2,447 campaigns = significant revenue suppression

Research 2: Google Ads API Sync Pipeline Architecture

Key Architecture Insight: For bulk operations (like pushing 926K neg KWs), Google recommends using BatchJobService (async) instead of synchronous CampaignCriterionService.mutate. BatchJobService allows up to 1M operations per job, handles retries internally, and doesn't block worker threads.

Limit	Value	OTTO Impact
Neg KWs per campaign	10,000	Must check before pushing backlog — some campaigns may be near limit
Neg KWs per Shared Set	5,000	Could use shared sets for universal negatives (free, cheap, diy)
Shared Sets per account	20	Use for "Global Negatives" applied across campaigns
Ops per BatchJob request	10,000 (or 10MB)	Dynamic batching needed — monitor byte size, not just count
Rate limiting	Token bucket per developer token + per CID	Must implement Redis-backed global throttle across Celery workers

Celery Architecture Fix: The research recommends separating queues: q_interactive (user clicks "Block" → sync mutate) and q_batch_sync (background sweep → BatchJobService). OTTO currently uses a single background queue. The analyze_negative_keywords task should chain to a submit_negative_batch task that uses BatchJobService with polling (not synchronous wait).

Research 3: LLM-Driven Negative Keyword Generation

Core Insight: LLMs excel at semantic intent classification but are unreliable for structural rules (match types). The recommended architecture: N-grams filter obvious noise (90-95% of data), LLM handles semantic ambiguity on the remaining 5-10%, and Python enforces match types deterministically.

Finding	Current OTTO Behavior	Recommended Change
LLMs hallucinate close-variant behavior	LLM assigns EXACT to 3+ word terms, believing it blocks variations	Never let LLM choose match type — use Python rules only
LLMs are non-deterministic	Same prompt can yield different match type choices across runs	Set temperature=0, use LLM only for yes/no classification
N-gram should run first	OTTO runs both in parallel, sends both to LLM	Run N-gram first, only send ambiguous terms to LLM
Validation architecture	Single validation pass	Add "LLM-as-Judge" second pass + impact simulation against historical data
Negative-to-Positive ratio benchmark	Unknown	High-performing accounts: 3:1 to 5:1 ratio (negs to positives)

Match Type Decision Matrix (should replace LLM discretion):

Match Type	When to Use	Word Count Rule
Negative Broad	Single words universally irrelevant (free, torrent, job, diy)	1 word only, verified 0% conversion across all contexts
Negative Exact	Traffic sculpting between ad groups, specific high-volume queries	1-2 words (OTTO's current rule is correct for this)
Negative Phrase	Multi-word concepts where sequence matters (default safe choice)	3+ words (OTTO should enforce this)

Research 4: Performance Impact Benchmarks

OTTO is outperforming industry benchmarks: Industry data from 15,000 accounts shows accounts with neg KWs convert at 13% vs 4.6% without (3x lift). OTTO's data shows 15.55% vs 2.16% (7.2x lift). OTTO's neg KW pipeline delivers more than double the industry-average improvement.

Metric	Industry Benchmark	OTTO Actual	Comparison
Conv rate WITH neg KWs	13%	15.55%	+20% above benchmark
Conv rate WITHOUT neg KWs	4.6%	2.16%	Worse — suggests OTTO campaigns need negs more
Conversion lift factor	3x	7.2x	2.4x better than industry
CTR improvement with negs	+89%	+165% (3.29% vs 1.24%)	Nearly double industry lift
CPA reduction with negs	-67%	-93% ($28.95 vs $427.23)	Massively better
Typical monthly waste	$1,127/account	926K unpushed keywords suggest much higher potential	Fixable waste

Over-Negation Risk: Research warns that aggressive negation can reduce conversions by 30-40% if it blocks legitimate long-tail queries. OTTO's 46K conflicts ARE an over-negation issue — actively blocking terms the campaign is bidding on. Additionally, using EXACT match for 3+ word terms is a mild form of "under-negation" — it misses variations that PHRASE would catch.

Revised Priority List (Post-Research)

Priority	Fix	Impact	Research Backing
P0	Set `action=SEND_TO_ACCOUNT` + chain Celery tasks	Unblocks 926K keywords from reaching Google	API research confirms BatchJobService for bulk push
P0	Strip match-type syntax in conflict check	Prevents future conflicts	Conflict research confirms syntax as root cause
P0	Delete 46K existing conflicts	Unsuppresses revenue on 2,447 campaigns	Google rules: campaign neg overrides ad-group positive
P1	Enforce PHRASE for 3+ words in Python (not LLM)	Better coverage of waste variations	LLM research: never trust LLM for structural rules
P1	Check campaign neg KW limits before pushing backlog	Prevents API errors (10K/campaign limit)	API research: hard limit causes RESOURCE_COUNT errors
P1	Implement retroactive conflict check	Prevents 15K+ "neg created first" conflicts	Conflict research: no automated systems do this well
P2	Switch to BatchJobService for bulk operations	Higher throughput, automatic retries	API research: mandatory for >10 keywords per operation
P2	Implement Redis-backed global rate limiter	Prevents RESOURCE_EXHAUSTED errors	API research: token bucket algorithm across workers

Diagnostic analysis of campaign activation, conversion tracking health, and structural efficiency. Source: otto-ppc production DB, Feb 2026. All metrics use median.

96.4%

OTTO Campaigns With Zero Clicks

72,273 of 74,972 campaigns never entered a Google auction

1.3%

OTTO Campaigns With Conversions

Only 966 of 74,972 campaigns have ever tracked a conversion

$73.7M

Zombie Spend (50+ clicks, 0 conv)

$68.4M Imported + $5.3M OTTO on campaigns that spend but never convert

Median OTTO AdGroups/Campaign

Imported median: 2-3. Over-segmentation causes "Low Search Volume"

Critical Finding: The bidding strategy debate is irrelevant for 96.4% of OTTO campaigns because they never enter a Google auction. The platform has a "Failure to Launch" crisis, not an optimization problem. OTTO is creating massive, over-segmented campaign structures (16 adgroups) that Google's algorithm ignores or marks as "Low Search Volume."

Campaign Activation Funnel

74,972

OTTO Created

2,699

Got Any Clicks (3.6%)

966

Got Conversions (1.3%)

Origin	Total	Zero Clicks	Got Clicks	Has Conversions	% Activated	% Converting
OTTO	74,972	72,273	2,699	966	3.6%	1.3%
Imported	175,127	117,757	57,370	26,376	32.8%	15.1%

Why 96.4% ghost campaigns? Root causes (ranked by likelihood): 1) Over-segmentation into 16 adgroups triggers Google's "Low Search Volume" filter. 2) Campaigns created in PAUSED state, user never activates. 3) Google account suspended/no payment method. 4) Keywords have zero search demand. 5) Budget too low for competitive auctions ($5/day spread across 16 adgroups = $0.31/adgroup).

Zero-Conversion Rate by Click Volume

Click Bucket	Origin	Campaigns	Zero Conv %	Total Spend	Wasted Spend
0 clicks	OTTO	72,273	100.0%	$0	$0
1-9 clicks	OTTO	585	92.0%	$760K	$328K
10-49 clicks	OTTO	870	74.1%	$2.8M	$1.2M
50-199 clicks	OTTO	681	53.2%	$26.6M	$2.6M
200-999 clicks	OTTO	458	34.9%	$49.2M	$2.7M
1000+ clicks	OTTO	105	26.7%	$28.6M	$23K

1-9 clicks	Imported	6,101	89.8%	$1.1M	$809K
10-49 clicks	Imported	8,131	71.8%	$6.4M	$3.8M
50-199 clicks	Imported	10,375	53.2%	$35.7M	$4.9M
200-999 clicks	Imported	13,415	43.8%	$119.2M	$16.1M
1000+ clicks	Imported	19,348	42.8%	$1.17B	$47.4M

The $47.4M Elephant: 42.8% of Imported campaigns with 1,000+ clicks have zero conversions. This is statistically impossible if conversion tracking is working — it is a broken pixel/tag problem, not a performance problem. These campaigns likely generate conversions that are never attributed.

OTTO Positive Signal: At 1,000+ clicks, only 26.7% of OTTO campaigns have zero conversions vs 42.8% for Imported. OTTO's tracking URL setup (99.1% vs 54.6%) gives it better conversion attribution at scale.

Converting vs Non-Converting Campaigns

Origin	Status	Campaigns	Median CPC	Median CTR	Median Clicks	Median Spend	Median Days	Median AdGroups
OTTO	Converting	966	$4.48	6.64%	122	$642	177	16
OTTO	Non-Converting	1,733	$1.34	6.46%	21	$16	174	16
Imported	Converting	26,376	$3.22	3.78%	632	$2,126	1,150	3
Imported	Non-Converting	30,994	$0.00	3.49%	143	$0	1,217	2

The CPC Quality Signal: Converting OTTO campaigns pay $4.48/click (premium, high-intent traffic). Non-converting ones pay $1.34/click (cheap, low-quality traffic). Higher CPC = better audience = conversions. The platform should monitor CPC as a leading indicator of campaign health.

The Volume Gap: Converting OTTO campaigns have 122 median clicks vs only 21 for non-converting. Most non-converters simply haven't accumulated enough data for Google's algorithm to optimize. Minimum viable spend threshold appears to be ~$300-$600.

Campaign Structure: AdGroups vs Performance

Origin	AdGroup Bucket	Campaigns	Median CPC	Median CTR	Median CVR
Imported	1 adgroup	18,491	$1.01	2.97%	0.00%
Imported	2-5 adgroups	11,879	$1.50	4.58%	0.00%
Imported	6-10 adgroups	5,048	$1.81	4.78%	0.00%
Imported	11+ adgroups	9,621	$2.82	4.71%	0.43%
OTTO	11+ adgroups	2,614	$2.72	6.50%	0.00%

Paradox: More adgroups actually correlates with higher CTR and the only non-zero CVR (0.43% at 11+). But OTTO's 16-adgroup structure delivers great CTR (6.5%) with zero CVR — the ads attract clicks but the landing pages or conversion tracking fail to close.

Campaign Churn: Where Do Campaigns Die?

Origin	Status	Campaigns	P25 Spend	Median Spend	P75 Spend	Median Days
OTTO	Enabled	6,533	$0	$0	$0	75
OTTO	Paused	67,735	$0	$0	$0	171
OTTO	Removed	698	$0	$0	$0	136
Imported	Enabled	23,726	$0	$0	$0	322
Imported	Paused	91,553	$0	$0	$30	1,073
Imported	Removed	59,848	$0	$0	$0	861

90.3% of OTTO campaigns are Paused (67,735 of 74,972). Median spend = $0, median clicks = 0, after a median 171 days. Users create campaigns via OTTO, they never serve, and eventually get paused. Even "Enabled" OTTO campaigns have median $0 spend and 0 clicks — they are technically "on" but Google won't serve them.

Recommendations: Priority Action Items

P0: Fix the "Ghost Campaign" Crisis

Add 48-hour activation check: if impressions=0 after 48h, trigger troubleshoot workflow
Reduce adgroup count from 16 to 1-3 per campaign — modern Google favors consolidation
Validate Google Ads account health (payment, suspension) BEFORE campaign creation
Check keyword search volume BEFORE building adgroups — reject zero-volume keywords

P1: Fix the $73.7M Conversion Tracking Gap

Flag any campaign with >$500 spend + 0 conversions as "Tracking Broken"
Auto-pause campaigns after $200 spend with 0 conversions until user verifies pixel
Build "Conversion Health" status on campaign dashboard
Require conversion action validation before enabling Smart Bidding

P2: Establish Minimum Viable Campaign

Minimum $10/day budget per campaign (current: many at $5/day across 16 adgroups)
Launch on Manual CPC for first 4 weeks, auto-switch to Maximize Conversions after 15+ conversions
Monitor CPC as health indicator: <$1.50 = likely junk traffic, >$3 = likely quality
Set minimum 100 clicks before evaluating campaign performance

P3: Build Campaign Health Monitoring

Real-time "Activation Rate" metric: % of campaigns with >1 impression in last 7 days
"Zombie Alert" for campaigns spending but not converting
Weekly report: campaigns created vs campaigns serving vs campaigns converting
CPC quality score: flag campaigns buying sub-$1 traffic as potential waste

All metrics use median (P50), not average. Source: otto-ppc production database, Feb 2026. Only strategies with ≥10 campaigns shown.

$3.02

OTTO Max Conv Median CPC

$2.38

Imported Max Conv Median CPC

0.87x

Max Conv Learning Overshoot

23.9%

OTTO Max Conv Survival Rate

Key Finding: Maximize Conversions has the lowest learning-phase overshoot (0.87x) — early campaigns actually have lower CPC than mature ones. The CPC spike in the trend chart is driven by outliers, not systemic overbidding. Target Impression Share (1.84x) and Manual CPC (1.60x) have far worse learning curves.

CPC by Campaign Age: Maximize Conversions

OTTO vs Imported (Maximize Conversions): OTTO campaigns peak at $6.66 CPC during weeks 4-12 then settle to $4.22. Imported campaigns stay flatter at $2.68-$3.27. OTTO has higher CTR (6.3% vs 4.3%) throughout all age cohorts, suggesting more competitive keyword targeting.

Learning Phase Overshoot by Strategy

Strategy	Origin	Early CPC	Mature CPC	Overshoot Ratio	Early N	Mature N
TARGET_IMPRESSION_SHARE	Imported	$19.26	$10.49	1.84x	102	1,393
MANUAL_CPC	Imported	$5.12	$3.19	1.60x	74	6,400
TARGET_ROAS	Imported	$1.21	$0.82	1.48x	14	305
MAXIMIZE_CONVERSION_VALUE	Imported	$1.33	$1.11	1.21x	144	2,335
TARGET_CPA	Imported	$6.91	$6.45	1.07x	57	3,831
TARGET_SPEND	Imported	$2.19	$2.18	1.00x	191	4,136
MAXIMIZE_CONVERSIONS	Imported	$2.84	$3.27	0.87x	666	11,270

Surprise: Maximize Conversions does NOT overbid during learning. Its 0.87x ratio means early campaigns are actually cheaper than mature ones. The previous dashboard's "$1,073 CPC in month 2-3" was an average inflated by extreme outliers. Median tells the real story.

CPC Distribution by Strategy (Percentiles)

Strategy	Origin	N	P10	P25	P50	P75	P90	P95	IQR
TARGET_IMPRESSION_SHARE	Imported	2,167	$0.00	$0.76	$7.96	$16.01	$23.57	$29.69	$15.26
TARGET_CPA	Imported	4,286	$0.23	$2.35	$5.83	$10.20	$17.27	$22.94	$7.85
MAXIMIZE_CONVERSIONS	OTTO	2,030	$0.00	$0.31	$3.02	$8.00	$20.26	$44.46	$7.69
TARGET_SPEND	OTTO	422	$0.00	$0.00	$2.56	$8.71	$17.58	$29.44	$8.71
MAXIMIZE_CONVERSIONS	Imported	22,951	$0.00	$0.00	$1.52	$5.50	$14.26	$25.87	$5.50
MAXIMIZE_CONV_VALUE	OTTO	132	$0.00	$0.00	$1.29	$3.94	$9.60	$18.70	$3.94
MANUAL_CPC	Imported	11,398	$0.00	$0.00	$0.76	$4.70	$12.35	$19.81	$4.70
TARGET_SPEND	Imported	8,460	$0.00	$0.00	$0.67	$3.59	$14.19	$30.69	$3.59
TARGET_ROAS	Imported	727	$0.00	$0.00	$0.48	$1.08	$1.93	$2.53	$1.08
MAXIMIZE_CONV_VALUE	Imported	5,419	$0.00	$0.00	$0.44	$1.48	$4.09	$8.17	$1.48

Outlier Impact: OTTO Maximize Conversions has P50=$3.02 but P95=$44.46. The top 5% of campaigns drive the extreme averages seen in the raw CPC comparison. The IQR ($7.69) is comparable to Imported Max Conv ($5.50), confirming the median tells a very different story than the mean.

Head-to-Head: Strategy Performance (Imported Search Only)

Strategy	Campaigns	Median CPC	Median CTR	Median CVR	Median CPA	Total Spend
MAXIMIZE_CONV_VALUE	2,558	$0.60	6.55%	0.00%	$49.93	$44.2M
TARGET_SPEND	5,354	$1.27	5.74%	0.00%	$175.35	$74.5M
MANUAL_CPC	9,101	$1.44	4.08%	0.00%	$161.42	$62.7M
MAXIMIZE_CONVERSIONS	16,787	$2.38	5.46%	0.63%	$116.07	$743.2M
TARGET_CPA	3,806	$6.66	4.23%	0.00%	$284.60	$8.6M
TARGET_IMPRESSION_SHARE	2,167	$7.96	4.18%	0.11%	$302.01	$30.0M

Best Overall: Maximize Conversion Value delivers the lowest CPC ($0.60) and best CPA ($49.93) among imported Search campaigns. Maximize Conversions leads in volume ($743M spend, 16.8k campaigns) with the only meaningful median conversion rate (0.63%).

Worst Performers: Target CPA ($284.60 CPA) and Target Impression Share ($302.01 CPA) are the most expensive strategies. Both have low CTR (~4.2%) and near-zero median conversion rates, suggesting they attract low-intent traffic or are misconfigured.

Spend Velocity & Efficiency

Strategy	Origin	N	$/day	Median Spend	Days Active	Conv/1k$
MAXIMIZE_CONV_VALUE	Imported	3,226	$3.33	$1,842	777	17.42
MAXIMIZE_CONVERSIONS	OTTO	1,550	$2.19	$329	176	0.03
MAXIMIZE_CONVERSIONS	Imported	16,374	$1.92	$1,166	758	5.14
TARGET_IMPRESSION_SHARE	Imported	1,679	$1.67	$1,671	1,296	1.32
TARGET_SPEND	OTTO	295	$1.65	$233	174	0.00
TARGET_ROAS	Imported	481	$1.32	$836	588	21.31
MAXIMIZE_CONV_VALUE	OTTO	84	$1.28	$119	97	5.86
TARGET_SPEND	Imported	5,366	$0.75	$575	974	1.69
MANUAL_CPC	Imported	6,864	$0.31	$509	2,313	1.57
TARGET_CPA	Imported	3,964	$0.23	$401	1,709	1.09

Efficiency Winner: Target ROAS delivers the highest conversion efficiency (21.31 conv/$1k) despite moderate spend velocity. Maximize Conv Value follows at 17.42. OTTO Maximize Conversions has very low efficiency (0.03 conv/$1k) — likely because most OTTO campaigns are young (median 176 days) and haven't accumulated enough conversion data.

Cohort Survival Rate (Last 12 Months)

Strategy	Origin	Started	Still Enabled	Survival %
TARGET_IMPRESSION_SHARE	Imported	419	254	60.6%
TARGET_ROAS	Imported	215	108	50.2%
TARGET_CPA	Imported	164	69	42.1%
TARGET_IMPRESSION_SHARE	OTTO	29	10	34.5%
TARGET_SPEND	OTTO	863	242	28.0%
MAXIMIZE_CONV_VALUE	OTTO	208	51	24.5%
MAXIMIZE_CONVERSIONS	OTTO	3,066	733	23.9%
MAXIMIZE_CONV_VALUE	Imported	2,202	364	16.5%
MAXIMIZE_CONVERSIONS	Imported	13,175	1,664	12.6%
MANUAL_CPC	Imported	1,712	175	10.2%
TARGET_SPEND	Imported	5,813	415	7.1%
MANUAL_CPC	OTTO	99	0	0.0%

OTTO survival is 2x Imported for Maximize Conversions: 23.9% vs 12.6%. OTTO campaigns are more likely to remain active, suggesting better initial setup quality leads to longer campaign life. Manual CPC has the worst survival (10.2% Imported, 0% OTTO) — a dying strategy.

Strategy x Channel: Search Performance (Median)

Strategy	Origin	N	Median CPC	Median CTR	Median CVR	Median CPA	Total Spend
TARGET_IMPRESSION_SHARE	Imported	2,167	$7.96	4.18%	0.11%	$302.01	$30.0M
TARGET_CPA	Imported	3,806	$6.66	4.23%	0.00%	$284.60	$8.6M
MAXIMIZE_CONVERSIONS	OTTO	2,030	$3.02	6.32%	0.00%	$82.47	$107.4M
TARGET_SPEND	OTTO	422	$2.56	7.14%	0.00%	$99.80	$0.2M
MAXIMIZE_CONVERSIONS	Imported	16,787	$2.38	5.46%	0.63%	$116.07	$743.2M
MAXIMIZE_CONV_VALUE	OTTO	132	$1.29	6.88%	0.00%	$18.58	$0.2M
MANUAL_CPC	Imported	9,101	$1.44	4.08%	0.00%	$161.42	$62.7M
TARGET_SPEND	Imported	5,354	$1.27	5.74%	0.00%	$175.35	$74.5M
MAXIMIZE_CONV_VALUE	Imported	2,558	$0.60	6.55%	0.00%	$49.93	$44.2M
MANUAL_CPC	OTTO	87	$0.57	7.35%	0.00%	$44.86	$10K

OTTO CTR Advantage: Across every strategy, OTTO campaigns have higher median CTR (6-7%) than Imported (4-6%). This is consistent with OTTO's tighter keyword targeting (73% exact match) driving more relevant ad placements.

Smart Ads Campaign Analysis

Platform Split

Channel Mix by Spend

Key Highlights

OTTO Smart Ads: Total Performance Impact

Data Coverage

Campaign Structure Quality: OTTO vs Imported

Structural Scorecard: OTTO vs Imported

Keywords per Ad Group Distribution

Ad Groups per Campaign

Ad Extension Coverage

OTTO Created

Imported

RSA Ad Copy Coverage

Bidding Strategy Adoption

Tracking & Targeting Completeness

Composite Structure Score Distribution

Data Extraction Methodology

OTTO-Created vs Google-Created (Imported) Campaigns

Statistical Distribution (Campaigns with >$100 spend, >10 clicks)

Monthly CPC Trend: OTTO vs Imported (Search Only)

Monthly Conversion Rate: OTTO vs Imported (Search Only)

OTTO vs Imported by Channel Type

Performance by Quartile

Campaign Maturity: Performance Over Time

Search-Only Head-to-Head: OTTO vs Imported (USD-Normalized)

CTR Distribution: Search Campaigns Only

CPC Distribution (USD-Normalized)

Cost per Conversion Distribution (USD)

Key Takeaways (Corrected)

Why the Raw CPC Comparison Is Misleading

The Channel Mix Problem

Search-Only CPC Comparison (Fair)

CPC by Spend Threshold

Landing Page Type Utilization

Landing Page Type: Conversion Rate Comparison

Campaign Serving Funnel

Ad Group Serving Breakdown

Campaign Status Distribution (Never-Served OTTO Campaigns)

Top Error Messages on OTTO Campaigns

Low Search Volume Assessment

Spend Concentration by Decile

Account Health Detail

Search vs PMax: Quarterly Spend & CPC Trend

Search vs PMax: Conversion Rate Trend

Search Campaigns - Quarterly

Performance Max - Quarterly

All Channels: Q1 2026 Snapshot

Cost per Conversion by Industry

CPC by Industry (Highest to Lowest)

Conversion Rate by Industry (Highest to Lowest)

Industry Performance Detail (sorted by Cost/Conv)

Spend & Cost-per-Conversion by Budget Tier

Performance by Budget Tier

Negative Keyword Adoption (Served Campaigns)

Negative Keyword Sources

Estimated Wasted Click Savings from Negative Keywords

Negative Keyword Quality Audit

Bug #1 (CRITICAL): 926K Neg KWs Never Synced to Google — Root Cause Found

Bug #2 (CRITICAL): 46,303 Positive/Negative Keyword Conflicts

Bug #3 (MEDIUM): Match Type Rule Violations

Performance Impact: Neg KWs Are Working

Audit Table: 2.67M Conflict Records (Passive)

Code Architecture

Specific Code Changes Required

Fix 1 (P0): Set action=SEND_TO_ACCOUNT in analyzer

Fix 2 (P0): Chain analysis → push in Celery task

Fix 3 (P0): Strip match-type syntax in conflict check

Fix 4 (P1): Enforce PHRASE match for 3+ word terms

Fix 5 (P1): One-time cleanup of existing 46K conflicts

Fix 6 (P2): Retroactive conflict check on positive KW creation

Fix 7 (P2): Handle the 926K backlog carefully

Fix 8 (P2): Alert on audit_negativekeywordconflict

Additional Risks Identified by Gemini

Match Type Distribution

Conversion Rate by Match Type (Actual Platform Data)

OTTO vs Imported by Dominant Match Type

OTTO's Exact-Match Strategy: Conversion Rate

OTTO's Exact-Match Strategy: Click-Through Rate

OTTO vs Non-OTTO: Full Comparison

Fix 1 (P0): Set `action=SEND_TO_ACCOUNT` in analyzer