PMax: Knife Rack — Campaign Analysis
Summary
A PMax campaign targeting magnetic knife racks that was correctly paused on Feb 3 after three weeks of sub-1.2x ROAS. The campaign exhibited classic PMax behaviour: when it exhausted high-intent Shopping inventory, it shifted spend to low-quality Display/YouTube placements with massive impression spikes and zero conversions. The fundamental issue is that knife racks (~£40-50 AOV) may not support paid acquisition at any reasonable cost structure.
Campaign Configuration
| Setting | Value |
|---|---|
| Campaign ID | 76 |
| Name | PMax: Magnetic Knife Rack Only Shopping (SC) |
| Type | Performance Max |
| Bidding strategy | MAXIMIZE_CONVERSIONS |
| Daily budget | £20 |
| Status | PAUSED (last data Feb 3) |
Performance Data
Weekly Progression (51 days: Dec 12 – Feb 3)
| Week | Impr | Clicks | Cost | CPC | Conv | Revenue | ROAS |
|---|---|---|---|---|---|---|---|
| Dec 8–14 | 48K | 534 | £191 | £0.36 | 12.2 | £629 | 3.30x |
| Dec 15–21 | 220K | 1,720 | £342 | £0.20 | 17.0 | £807 | 2.36x |
| Dec 22–28 | 12K | 170 | £207 | £1.22 | 6.0 | £345 | 1.67x |
| Dec 29–Jan 4 | 25K | 454 | £334 | £0.74 | 19.3 | £1,181 | 3.54x |
| Jan 5–11 | 21K | 407 | £337 | £0.83 | 14.5 | £823 | 2.44x |
| Jan 12–18 | 40K | 718 | £357 | £0.50 | 5.2 | £280 | 0.79x |
| Jan 19–25 | 52K | 782 | £342 | £0.44 | 7.0 | £311 | 0.91x |
| Jan 26–Feb 1 | 162K | 1,041 | £142 | £0.14 | 4.2 | £162 | 1.14x |
| Feb 2–3 | 2K | 32 | £14 | £0.43 | 1.0 | £40 | 2.88x |
Totals: £2,265 cost | 5,858 clicks | £0.39 CPC | 86.4 conversions | £4,578 revenue | 2.02x ROAS
Operational Note
Jan 29–30 was affected by the GMC suspension (domain switch to hairpin.com). However, performance was already deteriorating for two weeks before this event. See ../../reference/operational-log.md.
Channel Breakdown (confirmed via segments.ad_network_type)
As of Feb 10, the system syncs PMax data segmented by ad network. This confirms what was previously inferred from CPC patterns.
| Channel | % Spend | Impressions | Clicks | Cost | CPC | Conv | Revenue | ROAS |
|---|---|---|---|---|---|---|---|---|
| SEARCH | 84.8% | 115K | 1,794 | £1,921 | £1.07 | 86.4 | £4,578 | 2.38x |
| CONTENT (Display) | 12.2% | 467K | 4,034 | £276 | £0.07 | 0 | £0 | 0.00x |
| YOUTUBE | 2.8% | 1K | 25 | £63 | £2.54 | 0 | £0 | 0.00x |
| SEARCH_PARTNERS | 0.2% | 221 | 5 | £5 | £1.02 | 0 | £0 | 0.00x |
| TOTAL | 100% | 583K | 5,858 | £2,265 | £0.39 | 86.4 | £4,578 | 2.02x |
Every single conversion came from Search. Display, YouTube, and Search Partners produced zero conversions combined while consuming 15.2% of budget (£344).
What this means
- Search-only ROAS would have been 2.38x instead of the blended 2.02x. Display waste dragged total performance down by 15%.
- Display bought 467K impressions at £0.07 CPC — remnant inventory on junk placements. Google’s PMax algorithm was opportunistically buying cheap Display clicks to fill the budget when it couldn’t find Search users likely to convert.
- YouTube was expensive and useless — £2.54 CPC (higher than Search) with zero conversions from 25 clicks. Small absolute spend but 0% return.
The junk placement pattern — now confirmed
Previously inferred from CPC patterns, now verified by channel data:
| Date | Impressions | Clicks | CPC | Conversions | What happened |
|---|---|---|---|---|---|
| Dec 18 | 153,640 | 809 | £0.06 | 0 | Junk Display placement burst |
| Jan 17 | 23,025 | 334 | £0.15 | 0 | Same pattern, smaller |
| Jan 19 | 13,360 | 282 | £0.17 | 2 | Same pattern |
| Jan 22 | 12,198 | 136 | £0.45 | 0 | Same pattern |
| Jan 27 | 32,797 | 264 | £0.07 | 1 | Large burst |
| Jan 28 | 108,600 | 558 | £0.04 | 0 | Largest burst — this caused the Jan 28 anomaly noted in the data audit (129K total account impressions that day) |
MAXIMIZE_CONVERSIONS bidding has no CPC floor, so Google fills the budget with the cheapest available impressions when it can’t find Search users likely to convert.
Device Breakdown (30 days, Jan 11 – Feb 9)
| Device | Spend | % Spend | Clicks | CPC | Conv | Revenue | ROAS |
|---|---|---|---|---|---|---|---|
| Mobile | £629 | 69.6% | 1,681 | £0.37 | 11.2 | £503 | 0.80x |
| Desktop | £244 | 27.0% | 185 | £1.32 | 7.2 | £330 | 1.35x |
| Tablet | £31 | 3.4% | 769 | £0.04 | 0.0 | £0 | 0.00x |
Key findings:
- Tablet is the smoking gun: 769 clicks at £0.04 CPC is unmistakably Display remnant inventory — normal Shopping/Search tablet clicks cost £0.50+. This is the device-level confirmation of the Display junk placement pattern identified in the channel breakdown. Google targeted tablets for its cheapest Display fill.
- Mobile dominates spend (70%) at a low £0.37 CPC — also heavily influenced by Display placements. Mobile is where PMax dumped most of its budget into cheap inventory that didn’t convert.
- Desktop is the only viable device: £1.32 CPC is a real Shopping/Search price, and 1.35x ROAS is the closest to breakeven. But 1.35x on a £53 AOV product still doesn’t cover COGS and shipping.
The AOV Problem
Average conversion value: £53 (£4,578 / 86.4 conversions). Knife racks retail at ~£40-50. At the campaign’s average cost per conversion of £26 (£2,265 / 86.4), the margin after COGS and shipping is negligible or negative.
Even the campaign’s best period (Dec 29–Jan 4, 3.54x ROAS) had a £17 cost per conversion. For a £40-50 product, that leaves ~£10-15 gross margin before COGS — likely unprofitable.
This is a product category problem, not a campaign optimisation problem. Low-AOV products with commodity margins generally can’t sustain paid Shopping/PMax acquisition.
Timeline of Decline
| Period | ROAS | What was happening |
|---|---|---|
| Dec 12–Jan 4 | 2.36-3.54x | Christmas demand, higher conversion rates |
| Jan 5–11 | 2.44x | Post-Christmas normalisation |
| Jan 12–25 | 0.79-0.91x | Collapse — PMax exhausted high-intent audience, shifted to Display fill |
| Jan 26–Feb 1 | 1.14x | Budget dropping but still loss-making |
| Feb 2–3 | Campaign paused | Correct decision |
Verdict
Correctly paused. The campaign had a window of viability during Christmas demand (higher conversion rates offset the low AOV) but couldn’t sustain performance once seasonal demand normalised. MAXIMIZE_CONVERSIONS on a low-AOV product with PMax’s tendency toward Display fill is a losing combination.
If knife racks are to be advertised again, they’d need:
- A Shopping-only campaign (no PMax Display/YouTube bleed) — or PMax with account-level Display/YouTube exclusions
- Target ROAS bidding with a realistic floor
- Acceptance that volume will be very low given the product’s price point
- Or bundling with higher-AOV products to increase average order value
Note: Search-only ROAS was 2.38x — marginal but potentially viable if Display waste is eliminated. The question is whether this holds without Christmas demand.