Is PMax Worth It?

Supersedes: pmax-display-waste-analysis.md (2026-02-10, merged into this document)

Summary

A cross-campaign, cross-time-period analysis of whether Performance Max campaigns can be justified over Standard Shopping for The Hairpin Leg Company. Uses 6 months of data (Aug 2025 — Feb 2026) covering three PMax campaigns (Knobs, Knife Rack, Table Tops) alongside the long-running Shopping Catch All campaign.

Verdict: No. PMax underperforms Shopping on every meaningful metric. The junk traffic (Display, YouTube, Search Partners) has never converted. Even isolating PMax’s Search-only traffic, it returns lower ROAS than Shopping at a fraction of the volume. The account would have performed better if PMax had never been turned on.

Important context: The two worst offenders (PMax Knobs and PMax Knife Rack) are already paused. Ongoing waste from the only active PMax campaign (Table Tops) is ~£12/month — negligible. The figures below are primarily historical.


1. Junk Traffic Is Indefensible

6-Month Totals (Aug 2025 — Feb 2026)

Across all campaigns, all time periods — zero conversions from non-Search networks:

NetworkImpressionsClicksCostConversions
CONTENT (Display)1,580,000+14,265£1,0890
YOUTUBEVariousVariousIncluded above0
SEARCH_PARTNERSVariousVariousIncluded above0

There is no time window, no campaign, no product category where Display, YouTube, or Search Partners produced a single tracked conversion.

60-Day Network Breakdown (Dec 12, 2025 — Feb 9, 2026)

More granular data from the 60-day window where all network types were syncing:

NetworkSpend% of totalImpressionsClicksCPCConvRevenueROAS
SEARCH£36,32796.8%1,368.6£118,6883.27x
CONTENT (Display)£9432.5%726K14,188£0.070£00.00x
YOUTUBE£2130.6%2K84£2.540£00.00x
SEARCH_PARTNERS£400.1%60£0.670£00.00x
GMAIL£2<0.1%4£0.520£00.00x
TOTAL£37,525100%1,368.6£118,6883.16x

100% of conversions and 100% of revenue came from Search. Non-Search channels produced nothing across 726K+ impressions and 14,300+ clicks.

Display Network Detail

  • 726K impressions at £0.07 average CPC across 14,188 clicks
  • Concentrated in burst patterns: single days with 100K+ impressions (Dec 18: 152K, Jan 28: 109K on Knife Rack alone)
  • Zero conversions across every campaign, every day, every burst
  • Classic remnant inventory pattern — Google filling budget with the cheapest available placements when Search demand is exhausted

YouTube Detail

  • £213 across 84 clicks at £2.54 CPC — more expensive than Search on some campaigns
  • Zero conversions
  • YouTube pre-roll/in-stream ads for furniture hardware don’t convert — this is an awareness channel being charged on a performance basis

2. PMax Search-Only Still Underperforms Shopping

Stripping out junk and comparing like-for-like (Search network only):

ChannelROASCPAClicksConversions
Shopping (all)3.99x£22.8837,3001,508
PMax Search-only3.66x£21.284,958237
PMax Blended (with junk)2.94x£26.5219,312237

PMax Search-only is 8% lower ROAS than Shopping. Blended is 26% lower. The CPA looks comparable, but PMax is running on a fraction of the volume — 237 conversions vs 1,508. At that sample size, the apparent similarity in CPA could easily be noise.


3. PMax Drags Account Performance Down

ScenarioAccount ROAS
Without PMax spend3.23x
With PMax spend3.13x
Delta-3.1%

Every pound shifted from Shopping to PMax has returned less.

From the 60-day window:

MetricActualWithout wasteDifference
Account spend£37,525£36,367-£1,158
Account revenue£118,688£118,688(unchanged)
Account ROAS3.16x3.26x+3.2%

The 3.2% ROAS improvement is modest in absolute terms. The more important finding is that Display/YouTube adds zero value for this account — not low value, not marginal value, literally zero conversions across the entire dataset.


4. The Incrementality Question Is Unanswerable

Google’s core justification for PMax is that it reaches audiences Shopping cannot. The data neither supports nor refutes this because:

  • Seasonal confounds: Nov 2025 (Shopping + PMax running) had 370 conversions vs Sep 2025 (Shopping only) at 296. But Nov is pre-Christmas — of course it’s higher.
  • Shopping wasn’t suppressed at the account level: Shopping Catch All held at 4.42x ROAS while PMax ran (vs 4.05x before). But the per-campaign analysis documents confirm product-level suppression — PMax Knobs specifically cannibalised Shopping for knob products, and it took 5 weeks to recover after pausing (see pmax-knobs-analysis.md).
  • No clean holdout test exists: Proving incrementality would require running Shopping-only and Shopping+PMax on identical product sets in identical time periods. That hasn’t happened.

5. The One Bright Spot Doesn’t Prove The Case

Table Tops PMax is the strongest performer: 4.12—4.25x ROAS, 98% Search allocation, £150 AOV. But this is explained by its configuration, not by PMax being a better channel:

  1. It uses MAXIMIZE_CONVERSION_VALUE bidding (not MAXIMIZE_CONVERSIONS), which naturally avoids Display waste.
  2. It targets a high-AOV product category with strong search demand.
  3. These same products already appear in Shopping Catch All and would very likely perform similarly there.

Table Tops proves that PMax can work when configured tightly, but it does not prove it works better than Shopping for the same products.


6. What PMax Is Supposed To Offer (And What Actually Happened)

Google’s PitchWhat The Data Shows
Broader reach via Display, YouTube, Gmail, DiscoverThese channels generated zero conversions across 6 months
AI-driven bidding across channelsThe AI allocated 17—31% of budget to Display for Knobs and Knife Rack. Wrong every time.
Audience signals beyond keyword matchingNo evidence this drove incremental conversions that Shopping missed
Simplified management (one campaign vs multiple)Real convenience benefit, but not a performance one

7. Waste By Campaign

CampaignStatusWaste spend% of budgetWaste channelsSearch-only ROAS
PMax: KnobsPAUSED£23030.9%Display (£190), YouTube (£41)0.54x
PMax: Knife RackPAUSED£34015.0%Display (£276), YouTube (£63)2.38x
PMax: Table TopsENABLED£121.7%Display (£6), YouTube (£4), Gmail (£1)4.20x
Shopping campaigns£00%N/A (Search only)
Search campaigns£00%N/A (Search only)

Pattern: Limited search demand = more Display waste

CampaignPMax AOVActual AOVDisplay % of spendSearch ROAS
Table Tops£150£1500.9%4.20x
Knife Rack£5312.2%2.38x
Knobs£35 (8 conv)£46—86 (order data)25.5%0.54x

PMax’s algorithm allocates more budget to Display when it can’t find profitable Search auctions. Knobs had the highest Display waste not because of low AOV (actual order AOV is £46—86, comparable to Knife Rack) but because search demand for the category is limited. With fewer search queries to bid on, PMax fills the budget with Display remnant inventory. This is a structural issue with PMax’s multi-channel allocation for niche product categories, not a product economics problem. See knobs-and-handles-analysis.md for the full analysis.


8. Campaign Detail

PMax Knobs (PAUSED Dec 23, 2025)

  • 12-day campaign, 0.37x blended ROAS
  • 31% of spend went to Display — zero conversions from it
  • Used MAXIMIZE_CONVERSIONS bidding (the worse option for waste control)
  • Suppressed Shopping for knob products; took 5 weeks to recover after pausing
  • Full analysis: pmax-knobs-analysis.md

PMax Knife Rack (PAUSED Feb 3, 2026)

  • 2.02x blended ROAS, 12.2% Display waste
  • Also used MAXIMIZE_CONVERSIONS bidding
  • £53 AOV too low for profitable PMax at this waste rate
  • Full analysis: pmax-knife-rack-analysis.md

PMax Table Tops (ENABLED)

  • 4.12x ROAS, only 2% Display waste
  • Uses MAXIMIZE_CONVERSION_VALUE bidding (the better option)
  • £150 AOV absorbs the small amount of waste
  • Only PMax campaign that could be considered break-even vs Shopping
  • Full analysis: pmax-table-tops-analysis.md

9. Recommendations

Consolidate into Standard Shopping

The data supports this over every alternative:

  1. Kill Display/YouTube/Partners allocation: Indefensible. Zero conversions, ever. If any PMax campaign is kept, use MAXIMIZE_CONVERSION_VALUE bidding and brand exclusions to force Search-only behaviour.

  2. PMax does not outperform Shopping: 3.66x vs 3.99x ROAS on Search traffic alone. The volume is too small for PMax to demonstrate any unique advantage.

  3. The suppression risk is real: Documented for Knobs — PMax cannibalised Shopping at the product level and took 5 weeks to recover after pausing. This hidden cost doesn’t appear in PMax’s own metrics.

  4. If you want to test PMax properly: Run a controlled incrementality test — identical products, identical time period, one in Shopping-only and one in Shopping+PMax. Without that, PMax’s claimed incremental value is unverifiable.

The pragmatic move: keep Standard Shopping (6 months of proven 3.99x ROAS at scale) and stop paying the ~£180/month Display tax that PMax imposes.

For any future PMax campaigns

If PMax is used again despite the above:

  • Apply Display/YouTube exclusions from day one (Google now supports account-level placement exclusions for PMax)
  • Use MAXIMIZE_CONVERSION_VALUE (not MAXIMIZE_CONVERSIONS) — this naturally reduces Display waste (Table Tops: 0.9% vs Knife Rack: 12.2%)
  • Set budget to match realistic Search demand, not higher — excess budget gets dumped into Display
  • Be aware that PMax suppresses Standard Shopping for the same products. PMax takes auction priority. For niche products, consider whether Standard Shopping alone would perform better.
  • Prefer Standard Shopping over PMax for niche/low-volume products. PMax needs enough conversion data to optimise (~15—20 conversions/month). Products with limited search demand can’t generate this, so PMax fills budget with Display waste. Standard Shopping is Search-only by definition and doesn’t have this problem.

Monitoring

With the channel data now syncing via segments.ad_network_type, build a dashboard alert for:

  • Any non-Search channel exceeding 5% of a campaign’s spend
  • Any Display spend exceeding £5/day on a single campaign
  • This catches Display waste early before it accumulates

Data Sources & Verification

  • Primary data: google_ads_raw_daily table with network_type column (added Feb 2026 via segments.ad_network_type)
  • Product-level data: google_ads_product_raw_daily table
  • 180-day backfill performed 2026-02-11 covering Aug 2025 — Feb 2026
  • All rows in google_ads_raw_daily have network_type populated — no unsegmented data exists
  • Conversion figures cross-checked against “Analyzify - Purchase” conversion action daily data — totals match exactly
  • All conversions are confirmed purchases (not add-to-cart or page views)
  • CSV export: pmax-junk-traffic-report.csv (shopify-app project root)

  • pmax-knobs-analysis.md — PMax Knobs campaign deep dive
  • pmax-knife-rack-analysis.md — PMax Knife Rack campaign deep dive
  • pmax-table-tops-analysis.md — PMax Table Tops campaign deep dive
  • knobs-and-handles-analysis.md — Knobs category search demand and economics
  • shopping-catch-all-analysis.md — Shopping Catch All baseline performance
  • ../../reference/operational-log.md — Timeline of campaign changes and events