Case StudiesPMax Algorithm Audit
PMax AuditEcommerce · ApparelLast 30 Days · Generated 3/16/2026

How an 87% Waste Rate Was Hiding Inside a "Healthy" PMax Campaign

ACME Marketing was spending $1,200/month on Performance Max. Their campaigns were live, bids were optimizing, and the algorithm showed no errors. The ClickCatalyst PMax audit found that $1,000 of that spend was generating zero return — and identified the exact structural causes.

VERIFIEDAll figures extracted directly from ClickCatalyst audit output · No interpolation

Audit Snapshot

$1,200
Total PMax Spend
Last 30 Days
60 / 100
Algorithm Control Index
Moderate Leakage (50–79)
$1,000
Confirmed Waste
87.4% of spend
$713
Recovery Potential
70% capture rate applied against confirmed waste

What the Audit Said: Do This First

01

Exclude Zombie SKU: "Premium Leather Jacket"

Stops $1,239 in active bleeding

02

Block Mobile App placements

41,087 junk impressions — zero conversions

03

Fix conversion tags

27% attribution drift detected

Finding 1 — Attribution Drift

Critical

27% Attribution Drift — Smart Bidding Was Optimizing Blind

When Google Ads reports 59 conversions and GA4 records zero, Smart Bidding is optimizing for phantom signals. It inflates CPCs, misallocates budget, and corrupts every ROAS figure in the account.

CampaignAds Conv.GA4 Conv.Status
Clean Water General Keyword in Exact – Jan2026590CRITICAL

Detected by: Integrity Monitor — 3-layer audit (HTML scan + Google Ads API probe + BigQuery analysis)

Finding 2 — Zombie Products

$1,400 Spent Training the Algorithm to Fail

Two products consumed $1,400 with zero units sold. Every click they received was teaching PMax's algorithm to find more audiences like people who don't buy — contaminating the entire account's signal quality.

ProductClicksWastedSales
Premium Leather Jacket(Apparel)486$1,2000
Winter Beanie(Accessories)54$129.800

Detected by: PMax Zombie Product Detection — spend above currency-scaled guardrail with zero units sold

Finding 3 — Margin Destruction

One Product Was Losing $1,800 Net — PMax Kept Spending on It

Basic Cotton T-Shirt had a 2.2K ad spend against $405 gross profit. Net result: −$1,800. ROAS looked acceptable on the surface. Unit economics revealed a product actively destroying margin.

ProductAd SpendGross ProfitNet Profit
Designer Sunglasses$419.44$2,300+$1,900
Basic Cotton T-Shirt$2,200$405−$1,800
Canvas Sneakers$654.55$1,100+$425.45

Finding 4 — Junk Placement Bleed

41,087 Impressions on Toddler Games, Solitaire & Toy Unboxing Videos

With Display at 59% of impression share, PMax was defaulting to cheap inventory. The Context Engine isolated the specific placements: mobile game apps comprised 63% of all junk placements. Zero conversions across all of them.

63%
Mobile Apps
25%
YouTube Channels
12%
Websites
PlacementTypeImpressions
Toddler Games FreeMobile App23,768
Toy Unboxing VideosYouTube Channel16,326
Classic SolitaireMobile App10,885
Yahoo NewsWebsite8,103
Candy Match 3Mobile App6,434

Finding 5 — What Was Actually Working

Healthy

Exploration Efficiency Was in the Healthy Range

EER 0.3–0.8 = healthy. The algorithm is paying a reasonable learning tax. Primary issue is where it is learning — not how much.

$33.62
Exploit CPA
$77.82
Explore CPA
0.43
EER Score

How This Was Calculated

01

Algorithm Control Index (ACI)

Composite score (0–100) from waste leakage rate and maturity distribution. Below 50 = algorithm operating without structural constraints.

02

Exploration Efficiency Ratio (EER)

Measures whether PMax's learning tax is spent wisely. Compares top-spending asset CPA (exploitation) vs. all others (exploration). Below 0.3 = junk inventory discovery.

03

Zombie Product Detection

Products spending above the currency-scaled guardrail with zero units sold. They corrupt Shopping signal quality, training the algorithm on non-converting segments.

04

Signal Health & Attribution Drift

When Ads conversions diverge from GA4 by >20%, the algorithm is bidding on phantom conversions — inflating CPCs and suppressing Quality Scores account-wide.

05

Recovery Potential (70% Capture Rate)

Projections apply a conservative 70% capture rate against confirmed waste, accounting for market volatility and imperfect spend remapping.

The methodology, scoring models, waste classification logic, ACI formula, EER calculation, and recovery projections are proprietary to ClickCatalyst Digital. Results are subject to standard market volatility. Recovery projections are estimates, not guarantees.

Full Report

PMax Black Box & Algorithm Audit

6-page PDF · Includes all charts, placement data, margin analysis & methodology

↓ Download PDF
Case Study: 87% Waste Rate Hiding Inside a "Healthy" PMax Campaign