ClickCatalyst
Back to Tracking Mismatch Audit
Authority Layer · 7 metrics · 5 data sources

Inside the Tracking Mismatch Engine

Every metric in your Tracking Mismatch audit compares Google Ads against GA4 using documented formulas. This page explains how drift percentages, phantom signals, signal density, and bidding trust verdicts are calculated — with the exact thresholds that trigger each status.

Data Sources

read-only OAuth · no campaign changes ever

All data is read via Google Ads API and GA4. Derived intelligence tables are pre-computed in our pipeline before your audit runs.

Google Ads API — Campaign Performance

google_ads_campaign_performance

Click, VT, cross-device, all conversions with anti-fanout grain collapse per campaign per day

Google Ads API — Ads Conversions

google_ads_ads_conversions

Conversion action names, categories, click vs modeled split, external source

Google Ads API — Ad Group Conversions

google_ads_ad_groups_conversions

Conversion lag buckets by action category

GA4 — Page Performance

ga4_page_performance

Landing page sessions, conversions, bounce rates for tag health by page analysis

Derived Intelligence — Signal Health

signal_health_intelligence_daily

Pre-calculated drift (7d + 30d), signal score, signal status, conversion density per campaign

01

Absolute Ads ↔ GA4 Gap

The raw conversion count difference between platforms

Total Ads conversions (from campaign_performance) minus total GA4 conversions (from signal_health_intelligence_daily summing ga4_conversions_30d). The positive gap represents phantom conversions the algorithm uses to make bid decisions with no GA4 confirmation.

⚠️

Why this matters: If Ads reports 500 conversions and GA4 reports 300, the algorithm bids as if 500 conversions happened. The 200 phantom conversions inflate CPCs across the entire account — not just for one campaign.

💡

In plain English: We subtract GA4's conversion count from Google Ads' conversion count. The difference is phantom — conversions Google claims that your own analytics cannot verify. The algorithm treats these phantoms as real when deciding how much to bid.

Gap < 10%Normal attribution variance — acceptable
Gap 10–25%Meaningful discrepancy — investigate
Gap > 25%Systemic mismatch — algorithm compromised
02

Signal Density

Does the algorithm have enough data to learn from?

Conversions per click over a 7-day window from signal_health_intelligence_daily (conversion_density_7d field). Campaigns below 0.01 density have less than 1 conversion per 100 clicks — the algorithm cannot distinguish good traffic from bad at this signal level.

⚠️

Why this matters: Smart Bidding needs a minimum volume of conversion signal to work. Below the threshold, the algorithm is essentially guessing — but it still spends your money with full confidence. Low-density campaigns should use manual bidding.

💡

In plain English: We count how many conversions happen per 100 clicks over 7 days. If fewer than 1 in 100 clicks converts, the algorithm doesn't have enough data to learn. It's spending your money but flying blind.

BLIND< 1 conv / 100 clicks — switch to manual
LOWWeak signal — algorithm struggling
STRONG> 10% — reliable for Smart Bidding
03

7-Day vs 30-Day Drift Window

Is the mismatch recent or structural?

Compares attribution_drift_7d_pct against attribution_drift_30d_pct per campaign from signal_health_intelligence_daily. When 30-day drift exceeds 7-day, late-arriving conversions are accumulating in Ads but not GA4 (window mismatch). When 7-day exceeds 30-day, something changed recently (tag deployment, consent mode update).

⚠️

Why this matters: The fix depends on which window is worse. A structural window mismatch needs attribution model alignment. A recent spike needs a specific change to be identified and reversed. This comparison tells you which path to take.

💡

In plain English: We compare the tracking gap over the last 7 days vs the last 30 days. If the 30-day gap is bigger, the problem is structural (your platforms use different measurement windows). If the 7-day gap is bigger, something broke recently — check what changed this week.

30d > 7dStructural — align attribution windows
7d > 30dRecent change — find what broke
EqualConsistent — stable baseline
04

Phantom Signals (Modeled Gap)

Budget flowing to decisions based on unverifiable data

From campaign_performance with anti-fanout grain collapse (MAX per campaign per date): all_conversions minus click_conversions per campaign. Campaigns with >30% phantom ratio are flagged. The spend column shows how much budget flows to phantom-driven bid decisions.

⚠️

Why this matters: This combines two dimensions: HOW MUCH phantom data exists AND HOW MUCH BUDGET it controls. A campaign with 60% phantom but $50 spend is manageable. A campaign with 40% phantom and $5,000 spend is an emergency. Both dimensions matter.

💡

In plain English: We find campaigns where a big chunk of conversions are Google's estimates (not verified clicks). Then we show how much budget flows to these campaigns. High phantom + high spend = maximum damage to your bidding accuracy.

< 30%Acceptable modeled share
30–50%Significant phantom signal
> 50%Algorithm guessing — switch to manual
05

Signal Volatility by Campaign

Is the tracking quality consistent or wildly variable?

Standard deviation of attribution_drift_7d_pct per campaign over the reporting period, requiring at least 7 days of data. High volatility means some days the data is clean and other days it's wildly off — the algorithm cannot build a stable bidding model when signal quality oscillates.

⚠️

Why this matters: Average drift can look moderate while hiding extreme swings. A campaign with 20% average drift might swing from 5% to 45% day-to-day. On the 45% days, every bid decision is wrong. Volatility catches problems that averages miss.

💡

In plain English: We measure how much the tracking gap bounces around day-to-day. A stable gap (even if large) is manageable because it's predictable. A volatile gap (swinging from fine to terrible) is worse because the algorithm can never settle on a reliable bidding model.

Low volatilityConsistent gap — predictable and manageable
High volatilityErratic — likely intermittent tag failure
06

Bidding Trust Assessment

Should this campaign use Smart Bidding or manual CPC?

Each campaign evaluated on three criteria: signal_health_score (≥80 = trust), conversion_density_7d (sufficient volume), and attribution_drift_30d_pct (<15% = reliable). The verdict ranges from TRUST (safe for Smart Bidding) to DO NOT TRUST (switch to manual immediately).

⚠️

Why this matters: Most advertisers leave Smart Bidding on by default across all campaigns. This assessment identifies the specific campaigns where Smart Bidding is actively harmful — where the data quality is too low for the algorithm to make good decisions.

💡

In plain English: We score each campaign on three things: is the tracking data reliable, is there enough conversion volume, and do Ads and GA4 agree? If all three are good, Smart Bidding is safe. If any fails badly, switch to manual bidding until the tracking is fixed.

TRUSTScore ≥ 80, drift < 15% — automated bidding is reliable
CAUTIONScore 60–80 — automate with manual oversight
DO NOT TRUSTScore < 40 — switch to manual CPC
07

Tracking Risk Summary

Three structural risks quantified as account-level percentages

Three categories calculated from campaign_performance and signal_health_intelligence: (1) Phantom Conversion Spend — % of budget on campaigns with >30% modeled, (2) Low Signal Spend — % of budget on CRITICAL/LOW_VOLUME campaigns, (3) Modeled Conversion Inflation — gap between all_conversions and click_conversions as a ratio.

⚠️

Why this matters: This is the executive summary. If combined risk exceeds 30% of budget, the entire account's bidding model is compromised. Fixes in one campaign can't compensate for noise in dozens of others. This frames tracking as a measurement infrastructure project, not a campaign-level task.

💡

In plain English: We calculate three types of tracking risk and express each as a percentage of your total budget. Added together, they tell you what fraction of your money flows to decisions based on unreliable data. Above 30% combined = your tracking is a business-critical infrastructure problem.

Combined < 15%Manageable — campaign-level fixes sufficient
Combined 15–30%Significant — systematic tag audit needed
Combined > 30%Account-level crisis — infrastructure project required

Proprietary notice: The methodology, scoring models, waste classification logic, and recovery projections are proprietary to ClickCatalyst Digital and provided for informational purposes only. Results are subject to standard market volatility. Recovery projections are estimates, not guarantees. Contact us within 14 days if any metric appears inaccurate.

Tracking Mismatch Methodology — How ClickCatalyst Compares Ads vs GA4 Conversions