PMax: Table Tops — Campaign Analysis

Summary

The account’s best PMax campaign and the only one still enabled. 4.12x blended ROAS across 22 days with 19.3 conversions. Channel data confirms PMax correctly allocated 98% of spend to Search, with negligible Display/YouTube waste. However, revenue dried up completely after Jan 28 (12 days of £0 revenue despite continued spend), which is likely attribution lag on a small-data campaign rather than a genuine performance collapse.

Campaign Configuration

SettingValue
Campaign ID77
NamePMax: Table Tops (SC)
TypePerformance Max
Bidding strategyMAXIMIZE_CONVERSION_VALUE
Daily budget£60
Target ROASNot captured (NULL)
StatusENABLED
First dataJan 19, 2026

Channel Breakdown (confirmed via segments.ad_network_type)

Channel% SpendImpressionsClicksCostCPCConvRevenueROAS
SEARCH98.3%32K801£689£0.8619.3£2,8924.20x
CONTENT (Display)0.9%2K20£6£0.310£00.00x
YOUTUBE0.5%4454£4£0.930£00.00x
SEARCH_PARTNERS0.2%3085£2£0.330£00.00x
GMAIL0.1%1651£1£0.520£00.00x
TOTAL100%35K831£701£0.8419.3£2,8924.12x

All 19.3 conversions came from Search. Non-Search channels consumed only 1.7% of budget (£12) — negligible waste compared to Knife Rack’s 15.2%.

Why Table Tops avoids the Display trap

MAXIMIZE_CONVERSION_VALUE (Table Tops) vs MAXIMIZE_CONVERSIONS (Knife Rack) makes a critical difference. MAXIMIZE_CONVERSION_VALUE optimises for revenue, not conversion count — so it has no incentive to buy cheap Display clicks that don’t produce revenue. MAXIMIZE_CONVERSIONS just wants more conversions, so cheap Display clicks that occasionally trigger micro-conversions look attractive to the algorithm.

The £60/day budget is also better calibrated. Natural Search demand roughly fills it, so there’s less leftover budget for the algorithm to dump into Display.

Performance Data

Daily Breakdown

DateChannelImprClicksCostConvRevenueROAS
Jan 19 SunSEARCH2,02053£411.0£1794.37x
Jan 20 MonSEARCH1,54139£362.0£2206.11x
Jan 21 TueSEARCH1,63248£412.0£2095.10x
Jan 22 WedSEARCH1,45337£290£00.00x
Jan 23 ThuSEARCH1,44840£332.0£2487.52x
Jan 24 FriSEARCH1,55935£302.0£2287.60x
Jan 25 SatSEARCH1,54937£281.0£2187.79x
Jan 26 SunSEARCH1,75747£412.0£3688.98x
Jan 27 MonSEARCH1,62942£302.0£42814.27x
Jan 28 TueSEARCH1,56646£413.3£59414.49x
Jan 29 WedSEARCH1,03129£281.0£00.00x
Jan 30–Feb 9SEARCH~16K~348£3111.0£00.00x

The “collapse” after Jan 28

The audit flagged Table Tops as “collapsing” — 6.57x ROAS in the Jan 19 week to 0.25x in Feb 2–8. The channel data shows this isn’t a channel-mix problem (98% Search throughout). Three factors explain it:

1. Low data volume. 19.3 total conversions in 22 days. That’s less than 1 per day. A few days without a conversion is statistically normal, not a signal of collapse.

2. Attribution lag. Google’s click-based conversion attribution can take 7–30 days. Clicks from Jan 30 onwards may not have finished attributing. The recent weeks always look worse than they really are — this is a known artefact with any conversion-optimised campaign.

3. The campaign is only 22 days old. PMax’s algorithm needs ~6 weeks to fully optimise. The first 3 weeks showed strong performance (19.3 conversions), and the algorithm was still in learning phase for the remaining period.

Conversion verification

All 19.3 conversions are confirmed purchases via “Analyzify - Purchase” conversion action (not add-to-carts). Average conversion value: £150 — consistent with table top pricing (£100–200 range). The ROAS figures are genuine.

Device Breakdown (30 days, Jan 11 – Feb 9)

DeviceSpend% SpendClicksCPCConvRevenueROAS
Desktop£41759.5%365£1.148.3£1,4953.59x
Mobile£26237.4%440£0.6010.0£1,3605.19x
Tablet£233.3%26£0.871.0£371.63x
Connected TV£0.03<0.1%00.0£00.00x

Key findings:

  • Mobile outperforms desktop on ROAS (5.19x vs 3.59x) at nearly half the CPC (£0.60 vs £1.14). For a ~£150 AOV product this is somewhat unexpected — table tops are a considered purchase you’d expect to convert better on a larger screen.
  • Mobile gets more clicks and more conversions (440 clicks, 10.0 conv) than desktop (365 clicks, 8.3 conv) despite lower spend. Mobile traffic is significantly cheaper and converts better for this product category.
  • Desktop still healthy — 3.59x ROAS is well above breakeven. Both devices are profitable.
  • Connected TV appeared — £0.03, negligible, but indicates PMax is testing video placements. Worth monitoring if this grows.

Caveat: 10 vs 8.3 conversions is a small sample. The mobile/desktop ROAS difference (5.19x vs 3.59x) could narrow with more data. Directionally useful, not conclusive.

Analysis

Strengths

  • Clean channel allocation: 98% Search is excellent for PMax. MAXIMIZE_CONVERSION_VALUE correctly avoids Display junk placements.
  • Strong CPC: £0.86 on Search — reasonable for table top product searches.
  • High AOV: £150 average conversion value gives healthy margin after ad cost (~£36 cost per conversion).
  • Budget efficiency: £60/day is well-calibrated to natural demand. No budget waste.

Concerns

  • Too few conversions to evaluate properly. 19.3 in 22 days is not enough data. Need 6–8 weeks minimum.
  • Revenue gap after Jan 28: 12 days of £0 revenue on Search. If this persists beyond attribution lag (i.e., by mid-February the late-Jan clicks still show £0), the campaign may have genuinely stalled.
  • Cannibalization limited but unknown: PMax has limited search term visibility through Google’s API — only a subset of queries are exposed. We can’t fully assess whether PMax Search is winning auctions that Shopping Catch All would have captured. Table tops are in the >£20 range, so they’re eligible for the Catch All campaign.

What we can’t see

QuestionWhy it mattersData needed
Full PMax search termsGoogle only exposes a subset — cannibalization with Shopping Catch All can’t be fully assessedAPI limitation (no fix)
Did attribution fill in for late Jan?Determines if “collapse” is real or lagRe-query in 2–3 weeks
Which table top products convert?Product mix affects sustainabilityProduct-level PMax data
Was a config change made?Could explain the performance shiftCampaign config history (Phase 1)

Verdict

Keep running, but monitor closely. The channel data is clean (98% Search), the ROAS on Search is strong (4.20x), and the AOV supports paid acquisition. The “collapse” is most likely attribution lag on a low-volume campaign.

Review criteria at 6 weeks (early March):

  • If cumulative ROAS remains above 3x with attribution caught up → healthy campaign, consider increasing budget
  • If late-Jan clicks never attribute conversions → genuine performance decline, investigate search terms and product mix
  • If search term data reveals heavy cannibalization of Shopping Catch All → consider whether PMax adds incremental value

Comparison with Knife Rack

MetricTable TopsKnife Rack
Bidding strategyMAXIMIZE_CONVERSION_VALUEMAXIMIZE_CONVERSIONS
Search % of spend98.3%84.8%
Display % of spend0.9%12.2%
Search ROAS4.20x2.38x
Display conversions00
AOV£150£53
Cost per conversion£36£26
Margin after ad cost~£100+~£10-15
VerdictViableNot viable (correctly paused)

The key differences: MAXIMIZE_CONVERSION_VALUE avoids Display waste, higher AOV provides margin buffer, and the budget is better calibrated.

  • ./pmax-knife-rack-analysis.md — Knife Rack PMax analysis (paused, Display waste pattern)
  • ../../shopify-app/docs/ad-platform/working-plan.md — Search term sync (Phase 2), campaign config history (Phase 1)
  • ../../../market-intelligence.md — PMax channel-level reporting availability, account-level placement exclusions