Hairpin PPC

shopping-catch-all-vs-top-performers

44 min read

Shopping Catch All vs Top Performers: Consolidation Analysis

Context

Four non-brand Shopping campaigns have operated over a 26-month window:

  • Shopping Top Performers (id 67): Jan 2024 - present, ENABLED. Per-ad-group tROAS bidding across 78+ ad groups, narrowed over time to ~3 with meaningful spend. Covers 10-14 products (2025 Q4 onwards). Total: £86,953 spend, £246,572 revenue, 2.84x ROAS.
  • Shopping Testing (id 69): Feb 2024 - Jul 2025, PAUSED. Predecessor to Catch All. Total: £30,341 spend, £41,392 revenue, 1.36x ROAS. Never profitable at scale, though improved to 2.23x ROAS in its final seven months (Jan-Jul 2025).
  • Shopping Catch All >£20 (id 71): Created Dec 10 2024, active Jan 2025 - present. Campaign-level tROAS. Single ad group, 700+ products. Total: £29,396 spend, £132,388 revenue, 4.50x ROAS.
  • Shopping Catch All <£20 (id 78): Created Jan 26 2026 - present. Single ad group, 115 products. Total: £323 spend, £920 revenue, 2.85x ROAS. Very new, tiny spend.

Also relevant: three PMax campaigns tested in late 2025/early 2026 (Knobs Only, Magnetic Knife Rack Only, Table Tops) that compete with Shopping for some of the same products.

Questions: Should Top Performers (67) be consolidated into Catch All (71)? What is the cost of running both? Is the Catch All model working? What role does the <£20 campaign play?

Data sources: 26 months of campaign-level monthly metrics (Jan 2024 - Feb 2026), search term overlap data, product mix per campaign, impression share, device performance, change history from google_ads_change_events, and PMax context.

Broader Account Context

This analysis focuses on the Jan 2024 - Feb 2026 window where the key campaigns overlap, but three points from the wider account history (documented in full in the Account Review 2020-2026) are relevant:

Shopping is now the only non-brand acquisition channel. Both of SC’s generic search campaigns have been paused — £84k total spend across two attempts, both under 2x ROAS. The account currently runs Brand Search + Shopping/PMax only. This means Shopping carries 100% of non-brand paid acquisition. Getting the Shopping structure right is not an optimisation exercise — it is the entire non-brand paid strategy.

Campaign 67 has been declining since launch. Top Performers was SC’s day-1 Shopping campaign, launched alongside the agency transition on Aug 18 2023. Its lifetime total is £112k spend at 3.00x ROAS — but that average masks a steady decline. In its first five months (Aug-Dec 2023) it delivered ~3.55x ROAS while serving a broad product catalogue. From Jan 2024 onwards (the period analysed in this document) it fell to 2.84x as it narrowed to a handful of hairpin leg SKUs. Campaign 67 hasn’t just been outperformed by Campaign 71 — it has been deteriorating on its own terms for over two years.

CPC has tripled since the previous agency era. Account-wide CPC has risen from £0.45 (Adpulse, 2020-2023) to £1.26 (SC era), driven by market-wide inflation and the shift from Search to Shopping. Against that backdrop, Campaign 71’s £0.81 CPC is 36% below the SC-era account average — it is genuinely cheap traffic in a market where CPCs have tripled. Campaign 67 at £1.45 CPC is 15% above the account average. The gap between these two campaigns (£0.81 vs £1.45) is wider than market inflation alone would explain.

Answer

Consolidate Campaign 67 into Campaign 71 and exclude colour-based duplicate SKUs from the Shopping feed. The data is unambiguous: Campaign 71 outperforms on every metric, in every month, with less management overhead. After consolidation, redirect budget to push impression share from 67% toward 80%, and optimise the product feed to capture the long-tail growth that Shopping is currently missing.

Should Top Performers (67) be consolidated into Catch All (71)? Yes. The data is unambiguous across every meaningful metric.

Campaign 71 (Catch All) delivers 4.50x ROAS on £29,396 spend. Campaign 67 (Top Performers) delivers 2.84x ROAS on £86,953 spend. That is not a marginal difference — Catch All returns 59% more per pound spent. The gap is perfectly consistent: in every month since Catch All’s creation (Jan 2025 - Feb 2026), Catch All has beaten Top Performers on ROAS in all 14 months. Campaign 71 started at 5.78x ROAS in its first month — dramatically outperforming from day one — and has never dipped below 3.32x. There is no period where Top Performers outperforms Catch All.

Campaign 67 has also narrowed itself into irrelevance. It started 2024 with 111 products across 78+ ad groups. By 2026 Q1 it serves 10-14 products, dominated by 3-5 hairpin leg SKUs that account for 60-70% of its revenue. It has become a single-product campaign with the overhead of a complex multi-ad-group structure. Meanwhile, Catch All serves 700+ products with no single product above 7% of revenue, capturing long-tail demand that Top Performers cannot reach.

What does cannibalization cost? An estimated £12,784/year in duplicate spend on 42-142 overlapping search terms per month. The top overlap terms — “table legs”, “hairpin legs”, “metal table legs” — appear in both campaigns every month. Campaign 71 consistently wins lower CPCs on these overlap terms and in several months (e.g. May 2025) generates more revenue from the same search terms than campaign 67.

Is the Catch All model working? Decisively. Campaign 71’s simple structure — one ad group, campaign-level tROAS, all products >£20 — outperforms the complex per-ad-group tROAS structure on every dimension: ROAS (4.50x vs 2.84x), CPC (£0.81 avg vs £1.45 avg), management overhead (one tROAS target vs dozens), and resilience to disruptions (Catch All delivered 4.99x ROAS in March 2025 while Top Performers crashed to 0.70x under a Maximize Clicks experiment).

What about the <£20 campaign (78)? Too early for conclusions on a campaign with £323 in spend. Early signs are reasonable: 2.85x ROAS, clear price-based product segmentation working. The structural approach — separating products by price point for different tROAS targets — is sound. Monitor through Q1 2026.

Recommended action: Consolidate campaign 67 into campaign 71. Move the high-performing hairpin leg SKUs into 71’s product feed. The data shows 71 handles the same search terms at lower CPC. After consolidation, consider increasing 71’s budget — its impression share averages 67%, meaning a third of eligible impressions are being lost. Settle the tROAS target (it has yo-yoed through 7 changes in 4 months) and give the algorithm time to optimise.

Beyond consolidation: There is significant Shopping revenue being left on the table through product feed gaps. The generic head term “table legs” is collapsing in Shopping (-30% YoY) while specific long-tail terms (“dining table legs”, “oak table legs”, “white table legs”) are surging — but product titles don’t match these growing queries. Similarly, 38 products at desk height (71cm) exist in the catalogue but none include “desk” in their title, missing a market that has grown +123% above pre-COVID levels. Some feed optimisation work has been done historically, but not informed by this search term analysis. A targeted feed review using the data in the Shopping Category Gap Analysis could unlock significant incremental Shopping revenue at zero additional ad spend.

The rest of this document is the evidence behind these answers.

The Full Picture: 26-Month Monthly Performance

Campaign 67 — Shopping Top Performers

Per-ad-group tROAS bidding. Started with 78+ ad groups and 111 products; narrowed to ~3 ad groups and 10-14 products by 2026.

MonthImpressionsClicksCostConvRevenueROASCPCCVRIS
Jan 2024985,6736,189£6,508196£13,5152.08x£1.053.2%75.4%
Feb 2024835,7835,730£7,029151£15,4872.20x£1.232.6%74.4%
Mar 2024523,5903,971£4,162156£15,3593.69x£1.053.9%73.4%
Apr 2024685,9345,147£6,426197£21,2863.31x£1.253.8%73.6%
May 2024600,5564,275£6,688166£18,8832.82x£1.563.9%67.0%
Jun 2024346,7472,668£4,551111£12,8472.82x£1.714.2%72.6%
Jul 2024242,8841,884£2,97596£13,2404.45x£1.585.1%76.5%
Aug 2024257,3912,016£3,428107£10,9443.19x£1.705.3%78.5%
Sep 2024396,5912,998£5,042170£15,9393.16x£1.685.7%78.2%
Oct 2024390,8603,236£6,279161£19,8643.16x£1.945.0%79.5%
Nov 2024251,9292,185£3,888135£15,7144.04x£1.786.2%78.5%
Dec 2024252,0871,988£4,01497£9,0582.26x£2.024.9%77.1%
Jan 2025220,1881,892£2,65985£8,1813.08x£1.414.5%80.1%
Feb 202592,360802£1,12427£2,7112.41x£1.403.4%79.9%
Mar 2025615,0683,233£6,42054£4,4850.70x£1.991.7%74.7%
Apr 202591,811785£1,01331£3,0553.02x£1.293.9%76.0%
May 2025105,354941£1,18528£3,0992.62x£1.263.0%78.4%
Jun 2025116,7941,021£1,59148£4,5832.88x£1.564.7%79.6%
Jul 2025121,2891,070£1,61446£5,0233.11x£1.514.3%82.5%
Aug 2025159,3511,582£2,05584£8,2604.02x£1.305.3%86.3%
Sep 2025133,9951,312£1,65758£4,8022.90x£1.264.4%85.0%
Oct 202555,918682£74224£1,5142.04x£1.093.6%90.1%
Nov 2025134,4281,360£2,11268£6,8853.26x£1.555.0%87.0%
Dec 2025114,8331,023£1,64661£6,0123.65x£1.615.9%82.5%
Jan 2026112,8911,202£1,35053£4,0102.97x£1.124.4%88.0%
Feb 2026*52,188611£79519£1,8162.29x£1.303.1%84.4%

*Partial month (12 days)

2024 totals: £60,989 spend, 42,287 clicks, 1,743 conversions, £182,136 revenue, 2.99x ROAS, avg CPC £1.44.

2025 totals: £23,819 spend, 15,703 clicks, 615 conversions, £58,610 revenue, 2.46x ROAS, avg CPC £1.52.

2026 YTD: £2,145 spend, 1,813 clicks, 72 conversions, £5,826 revenue, 2.72x ROAS, avg CPC £1.18.

Lifetime totals: £86,953 spend, 59,803 clicks, 2,429 conversions, £246,572 revenue, 2.84x ROAS, avg CPC £1.45.

Campaign 69 — Shopping Testing (Predecessor to Catch All)

The first attempt at a broad Shopping campaign. Paused after Jul 2025.

PeriodSpendClicksConvRevenueROASAvg CPC
2024 (Feb-Dec)£22,35019,709301£23,5401.05x£1.13
2025 (Jan-Jul)£7,9906,879195£17,8512.23x£1.16
Lifetime£30,34126,588496£41,3921.36x£1.14

Shopping Testing was the right idea (broad product coverage) with the wrong execution in 2024, where it returned 1.05x ROAS and lost approximately £12,934 at 40% gross margins. Its dramatic improvement in 2025 (2.23x ROAS in its final seven months) validated the broad approach — Campaign 71 (Catch All) then proved it conclusively with 4.50x lifetime ROAS.

Campaign 71 — Shopping Catch All >£20

Created Dec 10 2024, activated Jan 2025. Single ad group, campaign-level tROAS, all products over £20.

MonthImpressionsClicksCostConvRevenueROASCPCCVRIS
Jan 2025127,0621,158£83348£4,8115.78x£0.724.2%61.2%
Feb 2025109,3101,095£74043£3,1474.25x£0.683.9%68.3%
Mar 2025179,9891,670£1,27270£6,3504.99x£0.764.2%59.4%
Apr 2025227,7552,378£2,06498£9,0434.38x£0.874.1%66.3%
May 2025179,2921,992£1,49089£8,2215.52x£0.754.5%64.9%
Jun 2025203,9601,996£1,69787£7,4304.38x£0.854.3%64.1%
Jul 2025310,8002,961£2,906118£11,2833.88x£0.984.0%68.6%
Aug 2025360,8623,851£2,889160£13,5234.68x£0.754.2%64.4%
Sep 2025308,8743,144£2,597156£13,7785.31x£0.835.0%67.5%
Oct 2025378,0293,899£3,299148£14,4924.39x£0.853.8%75.3%
Nov 2025278,2373,286£2,736134£12,9534.73x£0.834.1%71.7%
Dec 2025287,9543,256£2,692143£11,7724.37x£0.834.4%73.2%
Jan 2026406,7484,428£3,055126£11,8623.88x£0.692.9%69.9%
Feb 2026*105,0461,345£1,12333£3,7243.32x£0.842.5%66.9%

*Partial month (12 days)

2025 totals: £25,217 spend, 30,686 clicks, 1,295 conversions, £116,801 revenue, 4.63x ROAS, avg CPC £0.82.

2026 YTD: £4,179 spend, 5,773 clicks, 160 conversions, £15,587 revenue, 3.73x ROAS, avg CPC £0.72.

Lifetime totals: £29,396 spend, 36,459 clicks, 1,454 conversions, £132,388 revenue, 4.50x ROAS, avg CPC £0.81.

Key observation: Campaign 71 has never posted a month below 3.32x ROAS. Its worst month (Feb 2026, 3.32x) is higher than Campaign 67’s lifetime average (2.84x). Its best month (Sep 2025, 5.31x) exceeded Campaign 67’s best month since Jan 2025 by 32% (67’s best in the overlap period was Aug 2025 at 4.02x).

Campaign 78 — Shopping Catch All <£20

Created Jan 26 2026. Price-based segmentation of products under £20.

MonthImpressionsClicksCostConvRevenueROAS
Jan 202614,648139£935£2662.85x
Feb 2026*38,635376£23014£6542.85x

*Partial month (12 days). Campaign started Jan 26 2026 (6 days of January data).

Too early for conclusions. £323 total spend, 515 clicks, 19 conversions. The 2.85x ROAS across both months is consistent but based on minimal data. The structural logic is sound — low-AOV products need different tROAS targets than high-AOV products. See the Cannibalization section for evidence that price-based segmentation between 71 and 78 is working at the product routing level.

Combined Non-Brand Shopping Performance (All Campaigns)

MonthClicksCostConvRevenueROAS
Jan 20246,189£6,508196£13,5152.08x
Feb 20247,394£8,080166£16,6142.06x
Mar 20246,831£6,270179£16,4942.63x
Apr 20247,458£8,537237£24,6902.89x
May 20246,173£8,816197£22,0442.50x
Jun 20245,223£6,679150£15,3332.30x
Jul 20243,763£5,054134£15,8983.15x
Aug 20243,820£6,163138£13,8562.25x
Sep 20245,090£7,739214£19,1042.47x
Oct 20244,417£8,585175£21,0562.45x
Nov 20243,033£5,561150£17,0793.07x
Dec 20243,161£5,774135£12,0532.09x
2024 Total61,552£83,7662,071£207,7362.48x
Jan 20254,686£5,257203£18,0233.43x
Feb 20255,241£4,690189£15,7303.35x
Mar 20259,151£12,283301£29,0402.36x
Apr 20257,150£7,332261£24,2363.31x
May 20256,919£6,523268£25,9113.97x
Jun 20257,117£7,708265£25,5433.31x
Jul 20256,923£7,757282£28,2643.64x
Aug 20257,367£6,717318£28,6284.26x
Sep 20256,467£6,278296£26,1274.16x
Oct 20255,852£5,247211£20,0473.82x
Nov 20256,111£6,216276£26,4194.25x
Dec 20255,666£5,566265£23,2674.18x
2025 Total78,650£81,5743,135£291,2353.57x
Jan 20267,076£5,530222£19,6073.55x
Feb 2026*2,427£2,24272£6,5582.93x
2026 YTD9,503£7,772294£26,1653.37x

*Partial month

The restructuring is already working: 40% more revenue on 3% less spend. This is not a story of cutting spend to improve efficiency — it is a story of better campaign structure producing more revenue from the same investment.

The combined Shopping performance improved dramatically from 2024 to 2025. Total ROAS rose from 2.48x to 3.57x while spend dropped only 3% (£83,766 to £81,574) — and revenue ROSE 40% (£207,736 to £291,235). The account is spending virtually the same and getting dramatically more back. This is not a story of cutting spend to improve efficiency at the expense of volume — it is a story of restructuring (replacing Shopping Testing with Catch All, reallocating budget from Top Performers) producing more revenue from the same investment.

Head-to-Head: Campaign 67 vs Campaign 71

Month-by-Month Since Catch All’s Creation

This is the central comparison. Both campaigns ran simultaneously from Jan 2025 onward, competing for the same Google Shopping auctions.

MonthC67 ROASC71 ROASC67 CPCC71 CPCC67 CostC71 CostC67 RevenueC71 Revenue
Jan 20253.08x5.78x£1.41£0.72£2,659£833£8,181£4,811
Feb 20252.41x4.25x£1.40£0.68£1,124£740£2,711£3,147
Mar 20250.70x4.99x£1.99£0.76£6,420£1,272£4,485£6,350
Apr 20253.02x4.38x£1.29£0.87£1,013£2,064£3,055£9,043
May 20252.62x5.52x£1.26£0.75£1,185£1,490£3,099£8,221
Jun 20252.88x4.38x£1.56£0.85£1,591£1,697£4,583£7,430
Jul 20253.11x3.88x£1.51£0.98£1,614£2,906£5,023£11,283
Aug 20254.02x4.68x£1.30£0.75£2,055£2,889£8,260£13,523
Sep 20252.90x5.31x£1.26£0.83£1,657£2,597£4,802£13,778
Oct 20252.04x4.39x£1.09£0.85£742£3,299£1,514£14,492
Nov 20253.26x4.73x£1.55£0.83£2,112£2,736£6,885£12,953
Dec 20253.65x4.37x£1.61£0.83£1,646£2,692£6,012£11,772
Jan 20262.97x3.88x£1.12£0.69£1,350£3,055£4,010£11,862
Feb 2026*2.29x3.32x£1.30£0.84£795£1,123£1,816£3,724

Campaign 71 won on ROAS in all 14 months. Its worst month (3.32x) is higher than Campaign 67’s lifetime average (2.84x). The entire performance gap comes from CPC — Campaign 71 buys the same quality click for 45% less.

Campaign 71 won on ROAS in all 14 months. There are no exceptions. From its very first month (5.78x vs 3.08x in Jan 2025) through the most recent partial month (3.32x vs 2.29x in Feb 2026), Campaign 71 has delivered higher ROAS in every single period.

Campaign 71 had lower CPC in all 14 months. Weighted average CPC across the overlap period: campaign 67 at £1.48, campaign 71 at £0.81. Campaign 71 is 45% cheaper per click.

Campaign 71 generated more revenue in 13 of 14 months despite a lower budget for most of the period. The only exception is Jan 2025 when campaign 67 was still heavily funded (£2,659 vs £833 spend).

March 2025: Campaign 67 at 0.70x ROAS vs Campaign 71 at 4.99x — a 7:1 efficiency ratio on the same auctions. This is the clearest demonstration that campaign-level tROAS is structurally more resilient than manual bidding experiments.

The March 2025 disaster illustrates the structural difference. Campaign 67 was switched to Maximize Clicks during the Feb 2025 conversion tracking break. It spent £6,420 and returned £4,485 — a 0.70x ROAS, the worst month in the entire dataset. Campaign 71, using campaign-level tROAS throughout, delivered 4.99x ROAS in the same month. The contrast is even starker than a simple comparison suggests: Catch All generated £6,350 in revenue on £1,272 in spend while Top Performers burned through five times the budget for less revenue. The tROAS algorithm was conservative when conversion data was unreliable; Maximize Clicks just bought clicks regardless of quality.

Summary Metrics (Jan 2025 - Feb 2026)

MetricCampaign 67Campaign 71Advantage
Total spend£25,964£29,396—
Total revenue£64,436£132,38871 (+105%)
ROAS2.48x4.50x71 (+81%)
Total clicks17,51636,45971 (+108%)
Avg CPC£1.48£0.8171 (-45%)
Total conversions6861,45471 (+112%)
Revenue per click£3.68£3.63~Equal
Avg CVR3.9%4.0%~Equal
Cost per conversion£37.85£20.2171 (-47%)
Avg IS82.5%67.0%67 (narrow product set)

Campaign 71 delivers more than double the revenue on similar spend. It wins on efficiency (CPC, ROAS, cost per conversion) and on volume (clicks, conversions). The two campaigns have virtually identical revenue per click (~£3.65) and conversion rates (~4%) — the traffic quality is equivalent. The entire performance gap comes from CPC: campaign 71 buys the same quality click for 45% less, so every pound of spend reaches twice as far.

The only metric where campaign 67 leads is impression share — and that is a function of its narrow product set, not superior performance. When you serve 10 products, winning 82-90% of auctions for those 10 products is expected. When you serve 700+ products, a 67% IS across a much larger product space is actually more inventory.

What Happened to Campaign 67? A Timeline of Decline

Campaign 67 did not always underperform. In 2024, it was the primary Shopping campaign and delivered acceptable results. Its decline is a story of narrowing focus, increasing complexity, and budget reallocation as Catch All proved its worth.

Phase 1: Full-Width Campaign (Jan-Jun 2024)

111 products, 78+ ad groups, per-ad-group tROAS targets ranging from 150% to 650%.

MetricQ1 2024Q2 2024
Products served11133
Monthly avg spend£5,900£5,888
Monthly avg revenue£14,787£17,672
Avg ROAS2.51x3.00x
Avg CPC£1.11£1.46

The campaign was broad and moderately effective. ROAS was acceptable at 2.51-3.00x. Product count was already narrowing from 111 to 33 as the algorithm focused spend on winning products.

Phase 2: Contraction (Jul-Dec 2024)

Product count dropped further. CPC rose as the campaign concentrated on fewer, more competitive products.

MetricQ3 2024Q4 2024
Products served1819
Monthly avg spend£3,815£4,727
Monthly avg revenue£13,374£14,879
Avg ROAS3.51x3.15x
Avg CPC£1.66£1.91

ROAS improved as the campaign dropped unprofitable products, but CPC rose to £1.91 avg in Q4. The campaign was becoming a specialist — good at selling hairpin legs, expensive on everything else, and no longer capable of capturing long-tail demand.

Phase 3: Managed Decline (2025)

Catch All launched in Jan 2025. Budget was progressively reallocated away from campaign 67.

QuarterProductsMonthly Avg SpendMonthly Avg RevenueROASCPC
Q1 202540£3,401£5,1261.51x£1.72
Q2 202518£1,263£3,5792.83x£1.38
Q3 202515£1,775£6,0283.40x£1.34
Q4 202510£1,500£4,8043.20x£1.47

March 2025 (0.70x ROAS, Maximize Clicks experiment) drags Q1 down. After that, performance stabilised at 2.62-4.02x ROAS but on decreasing spend. By Q4 2025, campaign 67 was spending £1,500/month — less than a quarter of what Catch All spent.

YoY Comparison: Campaign 67 in 2024 vs 2025

Metric20242025Change
Total spend£60,989£23,819-61%
Total revenue£182,136£58,610-68%
ROAS2.99x2.46x-18%
Total clicks42,28715,703-63%
Total conversions1,743615-65%
Avg CPC£1.44£1.52+6%
Products served111→1940→10Continued narrowing

The decline in spend and revenue was primarily due to budget reallocation to campaign 71, not performance degradation. CPC actually rose slightly from £1.44 to £1.52 as the campaign narrowed. ROAS dropped from 2.99x to 2.46x, dragged down by the catastrophic March 2025 (0.70x).

Excluding March 2025: campaign 67’s 2025 ROAS was 3.11x on £17,399 spend and £54,125 revenue — a slight improvement over 2024’s 2.99x. The campaign is not broken; it is redundant. Its best products perform adequately, but campaign 71 serves those same products at higher ROAS and lower CPC.

The Bidding Complexity Problem

Campaign 67’s per-ad-group tROAS structure created significant management overhead for diminishing returns:

  • Feb 17, 2025: Mass reset — 22 ad groups all set to tROAS 320% on the same day.
  • Mar 3, 2025: Switched to Maximize Clicks (the decision that produced the 0.70x March).
  • Mar 31, 2025: Back to tROAS 200%, budget cut from £215 to £150.
  • Apr-May 2025: Individual ad group tROAS adjustments across multiple groups (280→250, 410→370, etc.).
  • May 27, 2025: Priority changed to 2 (high priority).
  • Jul 21-23, 2025: Budget cut from £150 to £90 (reallocation to Catch All).
  • Sep 9, 2025: Budget cut further to £80.
  • Nov 20, 2025: tROAS lowered to 175% via Maximize Clicks toggle.

This is a campaign being actively managed by humans trying to outperform an algorithm. Every intervention represents time and attention that could be spent elsewhere. Meanwhile, campaign 71 required a single tROAS target and produced better results.

Cannibalization Between Campaigns

Campaign 67 vs Campaign 71

Both campaigns bid on Shopping ads in the same auctions. The overlap is substantial and measurable.

MonthOverlap TermsC67 Cost on OverlapC67 Rev on OverlapC71 Cost on OverlapC71 Rev on Overlap
Jan 202563£941£2,823£159£692
Feb 202551£341£712£143£350
Mar 2025116£1,271£1,216£309£2,018
Apr 202577£432£1,550£373£751
May 202578£400£612£316£2,725
Jun 202574£588£1,958£276£1,276
Jul 202596£643£2,492£568£2,010
Aug 2025142£1,070£4,945£568£2,234
Sep 2025107£790£2,372£426£2,042
Oct 202586£412£735£547£2,327
Nov 2025114£870£3,102£550£1,721
Dec 202591£798£3,518£408£2,526
Jan 2026117£824£2,504£478£1,746
Feb 202642£281£1,151£132£435
MetricAnnual Estimate
Estimated annual overlap cost (campaign 67)~£8,281/year
Estimated annual overlap cost (campaign 71)~£4,503/year
Total annual cannibalization spend~£12,784/year

£12,784/year in duplicate spend on search terms both campaigns bid on. Campaign 71 consistently wins lower CPCs on these overlap terms. Every pound Campaign 67 spends competing with Campaign 71 on “table legs” is a pound that could fund Campaign 71 bidding on long-tail terms it currently misses due to budget constraints.

Top Overlap Terms

The same search terms appear in both campaigns every month. These are not edge cases — they are core product terms.

TermFrequencyNotes
”table legs”Every monthCore product category
”hairpin legs”Every monthCore product + brand
”metal table legs”Every monthHigh-volume product term
”desk legs”Most monthsSecondary category
”bench legs”Most monthsSecondary category

Campaign 71 Wins on Overlap Terms

On the shared search terms, campaign 71 consistently achieves lower CPC. In several months, it also generates more revenue from the same terms:

May 2025 example (overlap terms only):

  • Campaign 67: £612 revenue on overlap terms
  • Campaign 71: £2,725 revenue on overlap terms
  • Catch All generated 4.5x more revenue from the same search queries.

This is the strongest evidence for consolidation. When both campaigns bid on “table legs”, campaign 71 wins the click more cheaply and converts it more often. Running both campaigns on these terms does not improve outcomes — it increases cost.

Campaign 71 vs Campaign 78

The <£20 product segmentation between these two campaigns is working as intended.

MonthOverlap TermsC71 Clicks on OverlapC78 Clicks on OverlapStatus
Jan 202620176168Initial routing
Feb 202641515Clean separation

The overlap between campaigns 71 and 78 is shrinking rapidly — from 20 overlapping terms in January to just 4 in February. This is the price-based segmentation working correctly, with products cleanly routing to their designated campaigns.

Product Mix: The Fundamental Structural Difference

Campaign 67 — Specialist

Campaign 67 has narrowed to a handful of hairpin leg SKUs. The top 3 products account for 60-70% of its revenue. By 2026 Q1, only 10-14 unique products receive meaningful spend.

PeriodUnique ProductsConcentrationDominant SKUs
Q1 2024111LowDiverse
Q2 202433ModerateHairpin legs emerging
Q3 202418HighHairpin legs dominant
Q4 202419HighTop 3 = ~60% of revenue
Q1 202540Very highshopify_gb_5190925058182, 28-boxw-sq-bl, 16-box-sq-bl
Q4 202510ExtremeTop 3 = ~70% of revenue
Q1 202614ExtremeSame SKUs

This concentration is a side effect of the per-ad-group tROAS structure. Each ad group optimises toward its target, and the algorithm naturally funnels spend toward the products with the best conversion rates. The result is a campaign that is functionally a “Hairpin Legs Only” campaign with the label “Top Performers” and the overhead of a multi-ad-group structure.

Campaign 71 — Generalist

Campaign 71 serves 700+ unique products per quarter. No single product exceeds 7% of revenue. The product mix includes:

  • Hairpin legs (all sizes and finishes)
  • Desk legs and frames
  • Table frames and bases
  • Bench legs
  • Breakfast bar legs
  • Oak and wooden legs
  • Outdoor table legs
  • Knife racks
  • Hooks and brackets
  • Shelving components
QuarterUnique Products
Q1 2025756
Q2 2025725
Q3 2025722
Q4 2025725
Q1 2026698

This is the long-tail capture that campaign 67 abandoned as it narrowed. Catch All discovers demand across the full product catalogue. Products that campaign 67 dropped as “low performers” may simply have needed different tROAS targets or more patience — campaign 71’s single tROAS accommodates all of them.

Campaign 69 — Broad but Poorly Executed

Campaign 69 (Shopping Testing) also attempted broad coverage, and its product counts tell the story of an expanding then contracting experiment:

QuarterUnique Products
Q1 202430
Q2 2024639
Q3 2024652
Q4 202421
Q1 2025190
Q2 2025167
Q3 2025157

Campaign 78 — Price-Based Segment

QuarterUnique Products
Q1 2026115

Why This Matters for Consolidation

Campaign 67’s remaining hairpin leg SKUs can be absorbed into campaign 71 without disruption. Campaign 71 already bids on the same search terms for the same products — it just does so at lower CPC. The few SKUs where campaign 67 has a strong track record (the top 3 hairpin leg variants) are not products that require bespoke tROAS targets. They are mainstream products with high conversion rates that any well-configured Shopping campaign would prioritise.

The risk that consolidation would “lose” campaign 67’s learned bidding intelligence is overstated. Campaign 71 has 14 months of conversion data across 700+ products including hairpin legs. Its algorithm already knows how to bid on these products. Removing campaign 67 from the auction eliminates the internal competition that currently inflates CPCs on those products.

Duplicate Hairpin Leg SKUs: Size-Based vs Colour-Based Listings

The Duplicate Structure

Hairpin legs exist in two different product listing formats on Shopify, and both are submitted to the Google Shopping feed:

  • Size-based listings (e.g. “71cm Hairpin Legs - Desk & Dining Table”): Organised by height, with all colours and designs as variants within a single product page. SKUs like 28-3-sk-or, 4-2-sk-bl. Landing pages include dropdowns for every size, colour, and design — effectively a single entry point for the entire hairpin range.
  • Colour-based listings (e.g. “Black Hairpin Legs”): Organised by colour, with sizes as variants. SKUs suffixed with -colour (e.g. 28-3-sk-or-colour, 4-2-sk-bl-colour). Landing pages only show that single colour — switching to copper or grey requires navigating back to the collection.

Campaign 67’s shopify_gb_* format SKUs (e.g. shopify_gb_5190925058182) are also size-based listings, just with Shopify’s internal product ID as the SKU identifier rather than the clean format.

Performance Comparison (Campaign 71, 2025)

MetricSize-basedColour-basedGap
SKUs160173+13 more colour SKUs for the same products
Impressions521,857416,245Size gets 25% more
Clicks3,6592,834Size gets 29% more
Cost£3,186£2,488—
Revenue£20,136£12,215Size generates 65% more
ROAS6.32x4.91xSize wins by 29%
CVR6.1%5.0%Size wins by 22%
CPC£0.87£0.88Identical

Same physical products, identical CPC, but colour-based listings convert worse and generate less revenue per pound spent.

159 Duplicate Pairs in the Feed

Of the 174 unique base hairpin SKUs in Campaign 71, 159 exist in both formats — both a size-based and colour-based listing competing in the same auctions. Only 14 colour-based listings have no size equivalent, and only 1 size-based listing has no colour equivalent.

Google is already learning which format converts better. In the duplicate pairs, 73% of impressions go to size-based listings (338,141 vs 126,073). But the colour-based listings still absorb 126,000 impressions and £2,488 in spend at a lower ROAS.

Head-to-Head: Selected Duplicate Pairs

Base SKUSize ImpsColour ImpsSize ROASColour ROAS
28-3-sk-or (71cm Raw Steel)73,3008768.16x0x
28-3-sk-gr (71cm Grey)52,84485813.33x0x
28-3-sk-brs (71cm Brass)16,3771,72612.35x2.21x
28-3-hd-re (71cm Heavy Duty)15,5246756.97x0x
14-3-sk-or (36cm Raw Steel)23,51114,6883.89x5.80x
4-2-sk-or (10cm Raw Steel)46,02956,9284.97x2.09x
4-2-sk-bl (10cm Black)10,87215,7053.15x4.03x

A handful of colour-based listings outperform their size equivalent (e.g. 14-3-sk-or, 4-2-sk-bl), but these are exceptions in a dataset of 159 pairs. The pattern is clear: size-based listings deliver higher ROAS in the majority of cases, and the most extreme performance gaps (8x, 12x, 13x ROAS vs 0x) are always in favour of size-based.

Why Size-Based Listings Convert Better

The conversion rate gap (6.1% vs 5.0%) is explained by the landing page experience:

  • Size-based page (e.g. /4-x-hairpin-legs-28inch-71cm-desk-dining-table): Customer lands on the product, sees dropdown menus for all sizes (10cm to 102cm), all colours (raw steel, black, brass, copper, grey), and all designs (standard, heavy duty). One page serves the entire range. A customer searching “hairpin legs” who lands here can find exactly what they want without leaving the page.
  • Colour-based page (e.g. /black-hairpin-legs): Customer lands on the product, sees only black hairpin legs in various sizes. No option to switch to copper or grey without navigating out to the collection and back. Higher friction, higher bounce probability.

Impact on Conversion Signals

159 duplicate pairs split conversion history across two competing SKUs for the same physical product. This dilutes the algorithm’s learning signal and wastes £2,488/year on a listing format that converts 22% worse.

The duplicate listings create two problems beyond the direct performance gap:

  1. Split conversion history: Google’s algorithm learns from conversion data per SKU. When the same physical product has two SKUs, conversion events are split between them — neither SKU builds as strong a conversion signal as a single consolidated entry would.
  2. Auction self-competition: Both SKUs can appear in the same auction. While Google generally shows only one per advertiser, the algorithm must choose between them — and sometimes picks the colour-based listing, which then converts at a lower rate.

Impression Share Analysis

Campaign2024 Avg IS2025 Avg IS2026 Q1 ISTrend
Campaign 6775.4%81.3%86.2%Rising (narrowing product set)
Campaign 71—66.8%68.4%Stable (wide product set)

Campaign 67: High IS Is a Symptom, Not a Strength

Campaign 67’s impression share has risen from 73-79% (2024) to 84-90% (late 2025). This looks like improvement but is actually a consequence of its narrowing product set. When a campaign serves only 10 products, winning 90% of auctions for those 10 products is expected. A campaign serving 700 products at 67% IS is capturing far more total impression volume.

Campaign 71: Room to Grow

Campaign 71 at 67% IS is losing a third of eligible impressions. Budget increases in Oct 2025 pushed IS to 75%, confirming this is budget-constrained, not structural. After consolidation, redirecting C67’s budget should push IS toward 75-80%.

Campaign 71’s IS ranges from 59-75%, averaging ~67%. This means a third of eligible impressions are being lost, primarily due to budget constraints.

The IS is not declining — the fluctuations reflect budget changes and seasonal demand shifts:

PeriodIS RangeContext
Jan-Mar 202559-68%Initial ramp-up, small budget
Apr-Jun 202564-66%Stable
Jul-Sep 202564-68%Budget raised (Jul 21: £83→£163)
Oct-Dec 202571-75%Budget raised further (Oct 10: £223)
Jan-Feb 202667-70%Budget cut and tROAS tightened (Jan 4: £200→£150, 260%→400%)

The late 2025 budget increases pushed IS toward 75%, confirming that IS is budget-constrained, not structurally limited. After consolidation, redirecting campaign 67’s budget to campaign 71 would provide room to push IS toward 75-80%.

Budget and Bidding Decisions: How They Correlate With Performance

The change history reveals a pattern: every budget reallocation from 67 to 71 improved overall Shopping performance. Every bidding experiment on 67 produced mixed or negative results.

Key Budget Reallocations

DateCampaign 67Campaign 71Effect on Combined ROAS
Jul 21-23, 2025£150→£90/day£83→£163/dayCombined ROAS improved: 3.64x (Jul) → 4.26x (Aug)
Sep 9, 2025£90→£80/day£163→£183/dayCombined ROAS stable: 4.16x (Sep)
Oct 10, 2025(unchanged)£183→£223/day, tROAS 290%→250%Combined ROAS: 3.82x (Oct) — tROAS tightened, IS improved

The July 2025 reallocation is the clearest signal. Campaign 67’s budget was cut 40% (£150→£90) and campaign 71’s budget was nearly doubled (£83→£163). The result: combined Shopping ROAS jumped from 3.64x in July to 4.26x in August — one of the best months in the dataset. Campaign 71 used the additional budget at 4.68x ROAS and £0.75 CPC.

The tROAS Yo-Yo Problem

Campaign 71’s tROAS has been changed 7 times in 4 months:

DatetROAS ChangeBudgetROAS Next Month
Jun 11, 2025300%→290%£833.88x (Jul)
Oct 10, 2025290%→250%£2234.73x (Nov)
Nov 3, 2025250%→200%£180—
Nov 5, 2025200%→250%£180—
Nov 9, 2025250%→280%£180—
Dec 11, 2025280%→260%£2004.37x (Dec)
Jan 4, 2026260%→400%£1503.88x (Jan)
Feb 10, 2026400%→350%£150—

tROAS is a learning algorithm. It needs weeks of stable data to optimise. Three changes in one week (Nov 3-9) and seven changes in four months prevent the algorithm from converging on an optimal bidding strategy. Each change resets the learning window.

The tightening from 260%→400% in January 2026 was particularly aggressive. ROAS held at 3.88x but clicks rose (the campaign was finding more qualifying traffic at the higher threshold) and revenue held at £11,862. The algorithm is hitting the tROAS target while being selective — it is refusing auctions it would have won at 260% because they do not meet the 400% threshold.

Campaign 67’s Bidding Experiments

DateChangeResult
Feb 17, 2025Mass tROAS reset: 22 ad groups all set to 320%Feb ROAS: 2.41x (below average)
Mar 3, 2025Switch to Maximize ClicksMar ROAS: 0.70x — worst month in dataset
Mar 31, 2025Back to tROAS 200%, budget cut £215→£150Apr ROAS: 3.02x (recovery)
Nov 20, 2025tROAS lowered to 175%Nov ROAS: 3.26x (adequate)

The March 2025 Maximize Clicks experiment is the clearest example of what goes wrong when bidding strategy and conversion tracking are misaligned. Maximize Clicks was activated on March 3 during the aftermath of the February conversion tracking break. The algorithm bought 3,233 clicks at £6,420 — more clicks than any month since May 2024 — but with no conversion signal to guide it, only 54 converted. The £4,485 in revenue was a 0.70x return on £6,420 in spend. Campaign 71, running campaign-level tROAS through the same period, delivered 4.99x ROAS. The tROAS algorithm pulled back when conversion data was unreliable. Maximize Clicks did not.

Device Performance

Both campaigns serve similar device mixes. The data confirms campaign 71’s advantage holds across both desktop and mobile.

Campaign 67 (2025+ Quarterly)

QuarterDesktop CPCDesktop ROASMobile CPCMobile ROAS
Q1 2025£2.152.36x£1.611.17x
Q2 2025£2.293.29x£1.032.49x
Q3 2025£2.162.71x£0.994.15x
Q4 2025£2.242.50x£1.103.94x
Q1 2026£1.743.08x£0.972.47x

Campaign 71 (2025+ Quarterly)

QuarterDesktop CPCDesktop ROASMobile CPCMobile ROAS
Q1 2025£1.085.75x£0.604.65x
Q2 2025£1.185.13x£0.704.43x
Q3 2025£1.244.75x£0.704.54x
Q4 2025£1.174.85x£0.714.20x
Q1 2026£1.063.61x£0.593.68x

Campaign 71 is cheaper and more efficient on both desktop and mobile in every quarter. The desktop CPC gap is notable: campaign 67 averages £2.12 on desktop while campaign 71 averages £1.15 — nearly half the cost. Mobile CPC is similarly split: £1.14 vs £0.66. Tablet volume is negligible in both campaigns. All traffic is on the Search network (Google Shopping ads in search results) — no Display, Video, or other networks.

The PMax Question

Three Performance Max campaigns were tested in late 2025/early 2026. PMax campaigns compete with Shopping for the same products in the same auctions — they are relevant context for any Shopping restructuring.

CampaignPeriodSpendROASStatus
Knobs Only Shopping (75)Oct-Dec 2025£2,461Mixed (0.35x in Dec)Paused
Magnetic Knife Rack Only (76)Nov 2025-Jan 2026£3,1101.93-3.49xPaused
Table Tops (77)Jan-Feb 2026£720+5.83x→0.79xActive — monitoring

Knobs Only and Magnetic Knife Rack have been paused by agreement. Neither performed well enough to justify the overlap with Shopping. Knobs Only ended at 0.35x ROAS in December — effectively zero return. Even excluding Display waste (which PMax cannot control), the Knobs Search-only performance was only 0.54x ROAS. Magnetic Knife Rack held 1.93-3.49x ROAS but underperformed Catch All’s 4.50x and was trending downward. Both campaigns are documented in detail in the PMax Knobs Analysis and Knobs & Handles Analysis.

Table Tops is the remaining active PMax campaign. It swung from 5.83x to 0.79x ROAS in its first two months on minimal spend — too early to judge. This needs close monitoring through Q1 2026 before making a keep/pause decision. The volatility is consistent with PMax on a niche category: insufficient data volume for the algorithm to stabilise.

The broader lesson: PMax on niche product categories with low search volume gives the algorithm too little data to optimise, and the inability to exclude Display placements means a portion of spend is always wasted on non-converting inventory. Standard Shopping campaigns provide more control and, for this account, consistently better ROAS.

How Campaign 67 Handled Disruptions vs Campaign 71

February 2025: Conversion Tracking Break

A Shopify checkout upgrade broke conversion tracking around Feb 5-10.

MetricCampaign 67Campaign 71
Feb 2025 ROAS2.41x4.25x
Mar 2025 ROAS0.70x4.99x
Recovery monthApril (3.02x)Never needed — stable throughout
Root cause of C67 crashSwitched to Maximize Clicks on Mar 3tROAS self-regulated

Campaign 71’s tROAS bidding pulled back spending when conversion signals weakened. Campaign 67 was switched to Maximize Clicks — an unbounded strategy that buys clicks regardless of conversion probability. The result was a month of undirected spending. The contrast in March is the most extreme in the dataset: 4.99x vs 0.70x — a 7:1 efficiency ratio on the same set of auctions.

Budget Reallocation (Jul 2025)

MetricCampaign 67Campaign 71
Budget change£150→£90/day£83→£163/day
ROAS before (Jun)2.88x4.38x
ROAS after (Aug)4.02x4.68x

Both campaigns improved after the reallocation. Campaign 67 improved because reduced budget forced the algorithm to be more selective. Campaign 71 improved because additional budget let it enter more auctions at the same efficiency.

tROAS Tightening (Jan 2026)

MetricCampaign 67Campaign 71
Change(No major change)tROAS 260%→400%, budget £200→£150
Jan 2026 ROAS2.97x3.88x
Jan 2026 revenue£4,010£11,862
Jan 2026 clicks1,2024,428

Campaign 71 maintained 3.88x ROAS even after aggressive tROAS tightening. The tROAS constraint is doing its job — filtering out less profitable auctions — but the 400% target may be unnecessarily conservative given the campaign’s historical ability to maintain 4.0-5.0x at 250-290% targets.

Key Findings

1. Campaign 71 Outperforms Campaign 67 on Every Efficiency Metric

Over the 14-month overlap period (Jan 2025 - Feb 2026):

MetricCampaign 67Campaign 71Difference
ROAS2.48x4.50x+81%
CPC£1.48£0.81-45%
CVR3.9%4.0%~Equal
Revenue/click£3.68£3.63~Equal
Cost/conversion£37.85£20.21-47%

Campaign 71 is 45% cheaper per click, converts at an equivalent rate, generates equivalent revenue per click, and returns 81% more per pound spent. The performance gap is entirely driven by CPC — campaign 71 buys the same quality traffic for nearly half the price. There is no metric where campaign 67 demonstrates a structural advantage.

2. Campaign 71 Won ROAS in All 14 Months

Not 13 of 14. Not “most months.” All of them. From its first month (5.78x vs 3.08x) to the most recent (3.32x vs 2.29x), campaign 71 has delivered higher ROAS in every single period of co-existence. Campaign 71 has never posted a month below 3.32x ROAS — its worst month is higher than campaign 67’s lifetime average (2.84x).

3. The Complex Ad Group Structure Provides No Demonstrable Benefit

Campaign 67’s per-ad-group tROAS was designed to let the algorithm optimise each product category independently. In practice:

  • It narrowed to 3 ad groups with meaningful spend, making the per-group structure redundant
  • It required constant human intervention (7+ bidding changes in 2025 alone)
  • It produced a catastrophic failure when one group’s strategy was changed experimentally (March 2025)
  • It delivered worse ROAS than campaign 71’s single campaign-level tROAS

Campaign 71’s simplicity is not a compromise — it is an advantage. One tROAS target means one learning algorithm operating on the full dataset. Per-ad-group tROAS fragments the data, giving each group’s algorithm fewer signals to work with.

4. Cannibalization Costs ~£12,784/Year With No Offsetting Benefit

The two campaigns share 42-142 search terms per month. The estimated annual cannibalization spend is £12,784 — £8,281 on campaign 67 and £4,503 on campaign 71. This is not a case where internal competition drives better auction outcomes. Campaign 71 consistently wins lower CPCs on overlap terms and often generates more revenue from the same queries.

Every pound campaign 67 spends competing with campaign 71 on “table legs” is a pound that could fund campaign 71 bidding on a long-tail term it currently cannot reach due to budget constraints.

5. Campaign 67 Has Narrowed to Functional Irrelevance

From 111 products (Q1 2024) to 10-14 products (Q1 2026). From 78+ ad groups to 3 with meaningful spend. From £6,500/month spend to £1,500/month. The campaign’s product catalogue is a subset of campaign 71’s — every product campaign 67 serves, campaign 71 also serves.

Campaign 67 is not adding reach. It is not testing new products. It is not discovering demand. It is bidding on the same hairpin leg SKUs that campaign 71 would bid on anyway, at higher CPC, with more management overhead.

6. Campaign 71’s Impression Share Has Room to Grow

At 67% average IS, campaign 71 loses a third of eligible impressions. The data shows IS is budget-constrained: when budget was increased in Oct 2025 (£183→£223/day), IS rose from 68% to 75%. Redirecting campaign 67’s budget (~£80/day) to campaign 71 after consolidation would provide headroom to push IS toward 75-80% without increasing total Shopping spend.

7. The tROAS Yo-Yo Is Hurting Campaign 71’s Algorithm

Seven tROAS changes in four months (Jun 2025 - Feb 2026). Each change resets the algorithm’s learning window. The November 2025 sequence — 250%→200%→250%→280% in six days — is counterproductive. Google’s tROAS algorithm needs 2-4 weeks of stable conditions to optimise. Changing the target every few days ensures the algorithm never converges.

Despite this, campaign 71 has maintained above-3.32x ROAS throughout. This is evidence of the underlying campaign’s strength — it performs well even when actively handicapped by frequent bidding changes.

8. March 2025 Proves tROAS Is the Right Strategy for Shopping

The Maximize Clicks experiment on campaign 67 produced 0.70x ROAS — the worst month in the dataset. Campaign 71’s tROAS delivered 4.99x in the same month with the same conversion tracking disruption. The mechanism is clear: tROAS restrains spending when conversion signals degrade; Maximize Clicks does not.

This is not an edge case. It is the expected behaviour of these two strategies under signal loss. tROAS has a built-in feedback loop (only spend where expected ROAS exceeds target). Maximize Clicks has no feedback loop for conversions at all.

9. Price-Based Segmentation (Campaigns 71 + 78) Is Structurally Sound

The <£20 campaign (78) cleanly separates low-AOV products from campaign 71. The overlap is shrinking rapidly (20 terms in January to 4 in February). The segmentation logic works.

This is a fundamentally different approach from campaign 67’s product-specific segmentation. Price-based splits create clean, non-overlapping product sets with different tROAS targets appropriate to their AOV. Product-specific splits (campaign 67’s approach) create overlapping auction competition and management complexity.

10. Combined Shopping Revenue Rose 40% on Flat Spend

Combined non-brand Shopping revenue: £207,736 (2024) → £291,235 (2025). A 40% increase on 3% less spend. The Catch All model is working — completing the consolidation is the logical next step.

The most powerful finding is at the portfolio level. From 2024 to 2025, combined non-brand Shopping revenue rose from £207,736 to £291,235 — a 40% increase — while spend dropped only 3% (£83,766 to £81,574). This improvement was driven by the Catch All model replacing Shopping Testing and absorbing budget from Top Performers. The restructuring did not sacrifice volume for efficiency — it delivered both.

11. Impression Decline Is Part Market, Part Product Feed Gap

Total Shopping impressions dropped 38% YoY despite stable impression share. Hairpin legs account for much of this (-48%, matching the structural market decline). But other categories are declining in Shopping while growing in search — pointing to product feed gaps, not market weakness.

Total Shopping impressions dropped 38% YoY (9.6M in 2024 → 5.9M in 2025) despite stable impression share (70-76%). This is not a competition issue — the total eligible auction pool is shrinking.

However, the decline is not uniform across product categories. Hairpin legs impressions dropped 48% — consistent with the -57% structural decline in “hairpin legs” search volume documented in the SEO trend data. But other categories are also declining in Shopping despite growing in organic search:

CategoryShopping Impression ChangeMarket (SEO) TrendGap Cause
Desk legs-10%+123% growth38 products at 71cm height exist but none include “desk” in title
Handles/knobs-91%+58-97% growthAll 76 SKUs priced <£20; no Shopping campaign served them until C78 launched Jan 2026
Table legs (long-tail)Head term -30%Long-tail surgingProduct titles say “Square Industrial Legs | 71cm Table” — don’t match “dining table legs”, “oak table legs”
Sofa/cabinet legs-55%+53-66% growthServed by hairpin variants at 10/20cm heights, but titles don’t include these search terms

Some feed optimisation work has been done historically, but not informed by this search term vs product title analysis. Three immediate actions require no additional ad spend:

  1. Feed title optimisation for table legs: Update titles to include the specific material/style terms that are growing (“oak”, “white”, “dining”, “trapezium”) alongside the existing product description. Google Shopping matches search terms to product titles — if the title doesn’t contain the term, the product won’t appear.
  2. Add “desk” to 71cm product titles: 38 products / 207 SKUs at desk height with zero “desk” coverage. Can be done via Merchant Center feed rules without touching Shopify.
  3. Monitor handles/knobs via Campaign 78: Now that the <£20 campaign exists, these SKUs have a campaign. Early signs are reasonable (C78 at 2.85x ROAS) but handles/knobs specifically need monitoring — a previous PMax test returned only 0.54x ROAS even on Search-only, suggesting the challenge may extend beyond campaign structure to competition, offer positioning, or on-site conversion.

A detailed analysis with specific product-to-search-term mapping is in the Shopping Category Gap Analysis.

12. Duplicate Hairpin Leg SKUs Are Diluting Conversion Signals

159 hairpin leg products exist in two formats in the Shopping feed: size-based listings (6.32x ROAS, 6.1% CVR) and colour-based listings (4.91x ROAS, 5.0% CVR). The colour-based listings absorb 416,245 impressions and £2,488 in spend while converting 22% worse than their size-based equivalents. Google is already preferring size-based (73% of duplicate-pair impressions), but the colour listings still consume budget and split conversion history across two SKUs for the same physical product. The performance gap is explained by landing page quality: size-based pages have full size/colour/design navigation, while colour-based pages show only one colour with no cross-navigation.

Excluding the -colour suffix SKUs from the Shopping feed would consolidate conversion signals, eliminate the weaker listing format, and redirect all hairpin leg impressions through the higher-converting product pages.

Change Events Timeline (Non-Brand Shopping)

Complete bidding strategy and budget changes relevant to these campaigns.

DateCampaignChangeEffect
Dec 10, 202471 (Catch All)Created: £40/day, tROAS 4.5x—
Dec 16, 202471Ad group tROAS set to 4.5xActivated
Feb 5, 202567Budget reallocation during conversion tracking break—
Feb 17, 20256722 ad groups all reset to tROAS 320%Mass standardisation
Mar 3, 202567Switched to Maximize Clicks (from tROAS 280%)0.70x ROAS in March
Mar 31, 202567Back to tROAS 200%, budget cut £215→£150Recovery
Apr-May 202567Multiple ad group tROAS adjustmentsOngoing tuning
May 27, 202567Priority changed to 2 (high priority)—
Jun 11, 202571tROAS 300%→290%Slight loosening
Jul 21-23, 202567, 71C67: £150→£90; C71: £83→£163Key reallocation — Aug combined ROAS hits 4.26x
Sep 9, 202567, 71C67: £90→£80; C71: £163→£183Further reallocation
Oct 10, 202571Budget to £223, tROAS 290%→250%IS pushed toward 75%
Nov 3, 202571tROAS 250%→200%Loosened
Nov 5, 202571tROAS 200%→250%Reversed 2 days later
Nov 9, 202571tROAS 250%→280%Third change in 6 days
Nov 9, 202571Budget £223→£180Cut alongside tROAS changes
Nov 20, 202567tROAS lowered to 175%—
Dec 11, 202571tROAS 280%→260%; budget £180→£200—
Dec 23-27, 2025AllChristmas pause/resume—
Jan 4, 202671Budget £200→£150; tROAS 260%→400%Aggressive tightening
Jan 26, 202678 (<£20)Campaign createdPrice-based segmentation
Feb 10, 202671tROAS 400%→350%Partial loosening

Recommendations

Immediate: Consolidate Campaign 67 into Campaign 71

Action: Pause campaign 67. Ensure the hairpin leg SKUs it serves are present in campaign 71’s product feed (they almost certainly already are, given 71 serves 700+ products). Simultaneously, exclude the -colour suffix SKUs from the Shopping feed (see below).

Rationale: Campaign 71 outperforms campaign 67 on ROAS (4.50x vs 2.84x), CPC (£0.81 vs £1.45), and resilience to disruptions. The two campaigns share 42-142 search terms monthly, costing ~£12,784/year in cannibalization. Campaign 67 serves only 10-14 products — a subset of campaign 71’s 700+ — with more management overhead. Campaign 71 has won ROAS in all 14 months of co-existence.

Risk: Minimal. Campaign 71 already bids on the same products and search terms. The algorithm will absorb the incremental volume. Monitor hairpin leg SKU performance in campaign 71 for 4 weeks post-consolidation to confirm they maintain conversion rates.

Budget: Redirect campaign 67’s current budget (~£80/day) to campaign 71. This increases campaign 71’s daily budget from £150 to £230 and should push IS from ~67% toward 75-80%.

Immediate: Exclude Colour-Based Hairpin Leg SKUs from the Shopping Feed

Action: Exclude all SKUs ending in -colour from the Google Shopping feed. This is a feed-level change in Google Merchant Center (or the feed app) — no Shopify product changes needed. The colour-based product pages remain on the website for SEO and browse-by-colour navigation; they are only removed from paid Shopping.

Rationale: 159 hairpin leg products exist in both size-based and colour-based formats in the feed. The colour-based listings deliver 4.91x ROAS vs 6.32x for size-based — a 29% performance gap explained by inferior landing page UX (no cross-colour navigation). They absorb £2,488/year in spend and 416,245 impressions while splitting conversion history across two SKUs per product. Google already sends 73% of duplicate-pair impressions to size-based listings, confirming it has learned the preference — but the colour listings still consume budget and dilute signals. Excluding them consolidates all conversion data onto the higher-converting format and eliminates the split.

Timing: Do this alongside the Campaign 67 consolidation so Campaign 71 gets a clean, deduplicated hairpin leg SKU set from day one.

Short-term: Stabilise Campaign 71’s tROAS

Action: Set campaign 71’s tROAS to 300% and leave it for at least 6 weeks.

Rationale: Seven tROAS changes in four months — including three in one week — prevent the algorithm from converging. The campaign has historically delivered 3.9-5.8x ROAS at 250-290% targets, consistently exceeding whatever target was set. A 300% target is conservative enough to maintain efficiency while permissive enough to capture volume. The current 350% (set Feb 10) was preceded by 400% (Jan 4) — at that level, the algorithm was refusing auctions it would have won profitably at 250-290%. Loosening to 300% gives the algorithm room to bid on more of the long-tail inventory where Catch All has historically performed best, while still maintaining a floor well above breakeven (~200% at 40% margins). The algorithm needs a stable target to optimise around — 6 weeks minimum before reassessing.

Short-term: Keep Paused PMax Campaigns Paused; Monitor Table Tops

Action: Knobs Only and Magnetic Knife Rack remain paused. Monitor Table Tops closely through Q1 2026 — review at £2,000 cumulative spend or by end of March, whichever comes first.

Rationale: The Knobs and Knife Rack PMax experiments have been analysed and the decision to pause them is agreed. Even excluding Display waste, Knobs Search-only returned only 0.54x ROAS. Table Tops is the remaining active PMax and is too early to judge — the 5.83x→0.79x swing across its first two months reflects insufficient data, not a conclusive verdict. If Table Tops stabilises above 3.0x ROAS on meaningful spend, it may justify continued operation. If it remains volatile or falls below Catch All’s efficiency, it should be paused and the products left to Campaign 71.

Short-term: Monitor Campaign 78 (<£20)

Action: Continue campaign 78 at current budget through Q1 2026. Review at £2,000 cumulative spend.

Rationale: At £323 total spend, it is too early to evaluate. The structural approach (price-based segmentation) is sound, and the overlap data shows products routing correctly and the overlap shrinking. But 2.85x ROAS on £323 spend is not statistically meaningful. Minimum 8-10 weeks of data before making structural decisions.

Short-term: Product Feed Optimisation

Action: Review and update product feed titles to align with the search terms that are actually growing. Prioritise three changes:

  1. Add “desk” to 71cm product titles — 38 products / 207 SKUs at desk height exist in the catalogue with no “desk” coverage. This can be done via Merchant Center feed rules without touching Shopify product data.
  2. Enrich table leg titles with material/style terms — growing search terms include “oak table legs”, “white table legs”, “dining table legs”, “trapezium table legs”. Current product titles like “Square Industrial Legs | 71cm Table” don’t include these terms. Google Shopping matches queries to product titles — if the word isn’t in the title, the product won’t show.
  3. Verify handles/knobs coverage in Campaign 78 — now that the <£20 campaign exists, the 76 handle/knob SKUs have a campaign for the first time since Shopping Testing was paused. Confirm they are active in the feed and appearing in auctions.

Rationale: Some feed optimisation work has been done historically, but not informed by this search term analysis. The data shows the generic head term “table legs” is collapsing in Shopping (-30% YoY) while specific long-tail terms are surging by 80-900%. The products exist — the titles just don’t match the queries. This is the highest-leverage action available because it requires zero additional ad spend and could unlock significant incremental impressions across the growing search terms. See the Shopping Category Gap Analysis for the full product-to-search-term mapping.

Projected Upside from Consolidation + Feed Optimisation

The combined effect of these recommendations is not just efficiency — it is growth:

  • Cannibalization elimination: ~£12,784/year in duplicate spend redirected to new auctions
  • Impression share improvement: Redirecting C67’s budget (~£80/day) to C71 could push IS from 67% toward 75-80%. At current conversion rates (4.0%) and revenue per conversion (~£91), each additional 1% of IS represents roughly £300-500/month in incremental revenue
  • Feed optimisation: The desk legs gap alone represents a market that has grown +123% above pre-COVID levels. Capturing even a fraction of this with feed title changes could add meaningful impressions at zero cost
  • Algorithm stability: A single stable tROAS target on one consolidated campaign will produce more consistent, predictable performance — reducing the management overhead that currently demands 7+ bidding interventions per quarter

These are conservative estimates. The actual upside depends on auction dynamics and marginal ROAS at higher IS levels. But the direction is clear: consolidation saves money, and that money can be redirected into growing inventory that the current structure cannot reach.

Medium-term: Consider Category-Based Segmentation

Action: After consolidation is stable (4-6 weeks), evaluate whether category-based segmentation (table legs, desk accessories, hooks) would outperform the current price-based segmentation (>£20 / <£20).

Rationale: Price-based splits assume low-AOV and high-AOV products need different tROAS targets. This is partially true, but category-based splits might capture a more meaningful difference — a customer searching “desk legs” has different intent and conversion behaviour than a customer searching “J-hooks”. Category segmentation would allow different tROAS targets per product category rather than per price band. This is not urgent — the current structure works — but worth evaluating once the consolidation dust settles.

Medium-term: Increase Campaign 71’s Budget After Consolidation

Action: If IS remains below 75% after absorbing campaign 67’s budget, consider further budget increases in £20/day increments, waiting 2 weeks between each increase.

Rationale: Campaign 71 at 67% IS is leaving a third of eligible impressions on the table. At 4.50x ROAS, additional spend at the margin should be profitable even if marginal ROAS drops to 3.0-3.5x. Budget increases should be gradual — sudden jumps give the algorithm room to bid more aggressively rather than more broadly. The Nov 2025 budget jump (£223) that coincided with IS improvement to 75% suggests the campaign can absorb more budget efficiently.

Summary

The Consolidation Case in Numbers

MetricCampaign 67 (Top Performers)Campaign 71 (Catch All)
Lifetime ROAS2.84x4.50x
Lifetime CPC£1.45£0.81
Products served10-14700+
Ad groups78→3 (complex)1 (simple)
Bidding strategyPer-group tROAS (constant tuning)Campaign-level tROAS (stable)
Management overheadHigh (7+ interventions in 2025)Low (set tROAS, adjust budget)
Resilience to disruptionsPoor (0.70x in Mar 2025)Strong (4.99x in same month)
Head-to-head ROAS wins0 of 14 months14 of 14 months
Impression share82-90% (narrow set)67% (wide set, room to grow)
Cannibalization cost~£8,281/year~£4,503/year

Campaign 67 was the right campaign for 2023-2024. It was SC’s day-1 Shopping campaign, launched during the agency transition, and served a broad product catalogue at ~3.55x ROAS in its first five months. But its gradual narrowing — from 111 products to 10, from 78 ad groups to 3, from primary campaign to secondary — means it no longer serves its original purpose. Campaign 71 does everything campaign 67 does, plus everything campaign 67 stopped doing, at lower cost and higher return.

The combined non-brand Shopping programme has already improved dramatically: total ROAS rose from 2.48x (2024) to 3.57x (2025), while spend dropped only 3% and revenue ROSE 40% (£207,736 to £291,235). This improvement was driven primarily by the Catch All model replacing Shopping Testing and absorbing budget from Top Performers. Completing the consolidation — fully retiring campaign 67 — is the logical next step.

The central lesson from this data: a single well-configured campaign with campaign-level tROAS outperforms a complex multi-ad-group structure with per-group tROAS. Simplicity is not a compromise. In Google Shopping, it is a competitive advantage. Consolidation and duplicate SKU cleanup are the immediate actions; feed optimisation is the growth lever.

Data Sources

  • Campaign-level monthly metrics: 26 months (Jan 2024 - Feb 2026), all four non-brand Shopping campaigns
  • Search term overlap data: Monthly, Jan 2025 - Feb 2026, campaigns 67 vs 71 and 71 vs 78
  • Product mix data: Quarterly unique product counts and concentration metrics
  • Change history: From google_ads_change_events table, Feb 2024 - present
  • Impression share: Monthly, all campaigns
  • Device performance: Quarterly breakdown, campaigns 67 and 71
  • PMax campaign metrics: Oct 2025 - Feb 2026, three campaigns. Detailed analyses: PMax Knobs Analysis, Knobs & Handles Analysis
  • Product feed gap analysis: Shopping Category Gap Analysis
  • Duplicate SKU analysis: Size-based vs colour-based hairpin leg listings in Campaign 71, 2025 data, 159 duplicate pairs compared
  • Broader account history: Account Review 2020-2026 — full 6-year account history, agency comparison, CPC inflation analysis, conversion tracking disaster
  • Network type: Confirmed all-Search for all Shopping campaigns