Why SC’s Search Campaigns Fail: A CPC Decomposition

SC has spent £84,403 on three generic search campaigns over 2.5 years, all paused with ROAS under 2x. AP ran profitable search campaigns (7-8x ROAS) for the preceding 3 years. The same search terms, the same products, the same landing pages. What’s different?

Data sources: google_ads_raw_daily (107k rows), google_ads_search_term_daily (237k rows), ad_campaigns. All spend in GBP.

Related documents:


Executive Summary

The 4x CPC gap between AP and SC on the same search terms is NOT caused by quality score, account history, ad copy, or landing pages. SC’s quality scores are 9, CTR is comparable or better, CVR is consistently higher, and CPCs were near-identical at the transition point.

The gap has two components:

  • Market inflation (~2-2.5x): Unavoidable. Shopping CPC on the same terms doubled over the same period.
  • Maximise Conversions bid strategy amplification (~1.5x on top): Fixable. Both SC search campaigns experienced a specific trigger event that caused Maximise Conversions to permanently ratchet CPCs upward.

The two trigger events are fully documented below:

  • Feb 2024: A Shopping budget surge shifted conversion attribution away from search, causing Maximise Conversions to panic-bid
  • Feb 2025: Shopify’s checkout broke conversion tracking, feeding bad data to Maximise Conversions

A Manual CPC approach at £1.20-1.50 caps could realistically halve the bid-strategy component, bringing search to ~2.4x ROAS — marginal but potentially viable if incremental to Shopping.


The Numbers, Side by Side

AP Text Search (all campaigns excl. Brand, pre-May 2023):

Half-yearSpendClicksConv.ROASCPCCVR
2020-H1£32,79578,2292,4106.67x£0.423.08%
2020-H2£57,809129,5224,1976.92x£0.453.24%
2021-H1£41,868113,9403,5888.12x£0.373.15%
2021-H2£27,97454,5452,0707.55x£0.513.79%
2022-H1£19,95742,4001,4958.32x£0.473.53%
2022-H2£14,80425,6181,1288.35x£0.584.40%
2023-H1£8,25014,2734285.26x£0.583.00%

SC Generic Search (all three campaigns, full lifetimes):

CampaignPeriodSpendClicksConv.ROASCPCCVR
(OLD) General Search Top PerformersAug 23 - Mar 24£39,34516,8997621.74x£2.334.51%
(OLD) General Search TestingJan 24 - Mar 24£5,7681,646300.54x£3.501.82%
General Search (SC)Oct 24 - Jul 25£29,27214,5214952.01x£2.023.41%

Factor 1: CPC Is 4x Higher

The same search term, “hairpin legs”, shows the shift clearly:

PeriodAgencySpendClicksCPCROAS
2020AP£13,47032,103£0.427.24x
2021AP£13,04025,226£0.526.87x
2022AP£7,29810,677£0.686.58x
2023-Q3/Q4SC£2,9993,180£0.946.13x
2024SC£7,1424,441£1.613.49x
2025-H1SC£9,5323,665£2.601.79x
2025-H2SC£1,163412£2.821.51x

AP was paying £0.42-0.68/click for “hairpin legs” in its operational years. SC is paying £1.80-2.82 for the same term. That’s a 3-4x increase.

Quality score and account history can be ruled out. SC’s search ads achieved quality scores of 9. SC’s CTR on the same terms is comparable or better than AP’s (23.6% vs 22.1% on “hairpin legs”). SC’s CVR is also better (5.2% vs 4.0%). And crucially, at the exact transition point (Aug 2023), CPCs were near-identical — AP’s final 6 months averaged £0.77 on “hairpin legs”, SC’s first 6 months averaged £0.95. There was no starting penalty.

The CPC gap opened AFTER the transition, driven by two factors:

  • Market-wide CPC inflation (~2x): Shopping CPC on the same “hairpin legs” term doubled from ~£0.50 (2023) to ~£1.05-1.30 (2024-2025)
  • Maximise Conversions bid strategy amplification (~1.5x on top): Both SC generic search campaigns used Maximise Conversions, which ratcheted CPCs upward during demand spikes and never reverted

Factor 2: Same Search Terms, Better CTR and CVR

Head-to-head comparison on text search campaigns (excluding Shopping/PMax, excluding corrupted May-Jul 2023):

Search TermAP CPCAP CTRAP CVRSC CPCSC CTRSC CVRCPC Multiple
hairpin legs£0.5122.1%4.0%£1.8223.6%5.2%3.6x
hairpin table legs£0.4921.3%3.7%£1.8519.1%5.5%3.8x
hairpin leg£0.3135.8%6.5%£2.4446.8%6.7%7.9x
metal table legs£0.4010.1%1.3%£1.808.4%2.6%4.5x
hairpin desk legs£0.7136.6%3.2%£2.9348.1%12.7%4.1x

SC’s CTR is comparable or better on every term. SC’s CVR is consistently higher — often significantly so (12.7% vs 3.2% on “hairpin desk legs”). The ads are good. The landing pages are good. The conversion path is better than it was under AP. The problem is purely CPC.

At £0.51/click, “hairpin legs” with a 4% CVR and ~£60 AOV delivers healthy ROAS. At £1.82/click with the same economics, ROAS drops by two-thirds. SC’s better CVR partially compensates, but not enough to overcome a 3.6x CPC increase.


Factor 3: Device Mix and Mobile CPC

AP SearchSC Generic Search
Mobile spend£116,479 (57%)£53,342 (63%)
Mobile CPC£0.39£1.93
Desktop spend£73,838 (36%)£29,108 (35%)
Desktop CPC£0.57£3.10
Desktop ROAS11.01x2.48x
Mobile ROAS5.39x1.42x

SC’s mobile CPC is 5x AP’s. Desktop CPC is also 5.4x higher. This isn’t a device mix issue — the CPCs are inflated across all devices. SC’s mobile traffic at £1.93/click and 1.42x ROAS is clearly unprofitable.


Factor 4: The CPC Gap Wasn’t There at Launch — It Opened Afterwards

At the agency transition, text search CPCs were near-identical:

PeriodAgencyCPC on “hairpin legs”CTRClicks
Mar-Aug 2023AP (final months)£0.7720.6%2,834
Aug 2023-Jan 2024SC (first months)£0.9520.7%4,045

SC launched at parity. The CPC gap opened inside SC’s campaigns over time, through two distinct “ratcheting” events — both visible in the weekly CPC data for “hairpin legs” on the Top Performers campaign:

WeekCPCImpressionsEvent
W39-41 (Sep-Oct)£0.64-0.90883-939Campaign launch
W42-50 (Oct-Dec)£0.91-1.24674-989Q4 seasonal rise
W51-52 (Christmas)£0.66-1.06522-636Holiday dip
W01-03 (Jan 1-19)£0.70-0.92752-1,034Post-holiday recovery
W04-05 (Jan 22-Feb 4)£1.12-1.49928-939CPC starts climbing
W06 (Feb 5-11)£1.591,022Shopping budget surge → conversion attribution shift
W07-08 (Feb 12-25)£2.20-2.23653-849CPC doubles, doesn’t revert
W09-11 (Feb 26-Mar 15)£2.55-2.82364-615Death spiral → campaign paused

The CPC went from ~£0.90 to ~£2.50 in three weeks (W04-W07) and never recovered.

What triggered the February 2024 CPC doubling?

The root cause is identifiable in the daily spend data: a deliberate Shopping budget surge in W06 (Feb 5-11) starved the search campaign of conversion attribution, triggering Maximise Conversions to permanently ratchet CPCs upward.

Background: Testing campaign launched January 11. SC added a second search campaign — (OLD) General Search Testing - SC — on TARGET_SPEND (Maximize Clicks) at £70/day. It used phrase match on competitor/retailer queries (screwfix, ikea, b&q, amazon). Over its 3-month life it spent £5,768 for ~15 conversions (£384/conversion). This was a background drain but not the immediate trigger.

The trigger: Shopping Top Performers budget was massively increased for W06. The daily spend data is unambiguous:

DateShopping TP Daily SpendSearch TP Daily SpendTotal Account
Feb 3£237£165£557
Feb 4£274£182£635
Feb 5£325£193£674
Feb 6£437£173£747
Feb 7£563£146£837
Feb 8£734£175£1,040
Feb 9£571£151£934
Feb 10£454£161£794
Feb 11£528£191£860
Feb 12£503£117£797
Feb 13£474£148£772
Feb 14£199£223£583
Feb 15£56£122£383

Shopping Top Performers went from ~£200/day to £734/day — a 3-4x increase. Its current budget is £80/day, so the budget was explicitly raised (likely to ~£400/day or higher) during this period. Then on Feb 15, when Shopping Testing launched, Shopping Top Performers was cut back dramatically to £40-156/day.

The conversion attribution shift is the smoking gun:

WeekSearch TP SpendSearch TP ConvsShopping TP SpendShopping TP ConvsAccount Total Convs
W04£70020.5£1,49644.6~92
W05£1,16826.2£1,28834.4~99
W06£1,1907.6£3,61254.5~90
W07£99516.1£1,43032.2~88
W08£72214.9£73723.4~76

In W06, Shopping gained ~20 conversions while search lost ~19. Total account conversions barely changed (99 → 90). The conversions didn’t disappear — they were reattributed from search to Shopping via last-click attribution, because Shopping was bidding 3x more aggressively and capturing the final click before purchase.

The chain reaction:

  1. Shopping budget surge (Feb 5-13) → Shopping captures last-click attribution on conversions that were previously attributed to search
  2. Search conversions collapse 71% in one week (26.2 → 7.6) while search spend stays flat at £1,190
  3. Maximise Conversions sees £157/conversion (vs normal ~£45) and aggressively raises bids to try to recover
  4. Shopping budget returns to normal (Feb 15: Shopping Testing launches, Shopping TP cut to £56/day) — some conversions return to search
  5. But search CPC doesn’t revert — Maximise Conversions has learned the higher bid level. CPC goes from £1.59 (W06) to £2.23 (W07) and stays above £2.50 for the rest of the campaign’s life
  6. Impression collapse follows — weekly impressions on “hairpin legs” fell from 1,034 (W03) to 364 (W11) as fewer auctions were profitable at the inflated CPC
  7. Campaign paused March 15 at £2.82/click — 3x its launch CPC

This is a well-documented Google Ads pitfall: Maximise Conversions campaigns are fragile to sudden changes in the conversion pool. When a parallel campaign (Shopping) temporarily absorbs conversions via attribution, Maximise Conversions interprets the drop as “my bids are too low” and raises them — when the real cause is attribution shift, not bid competitiveness. The correct response would have been to pause the search campaign during the Shopping budget test, or switch it to Manual CPC temporarily.

The Testing campaign compounded the damage. Its top keywords by spend:

KeywordMatchSpendClicksConversions
screwfix hairpin legsPhrase£3131021.0
short hairpin legsPhrase£204581.0
ikea hairpin legsPhrase£178530.0
black metal hairpin table legsPhrase£149462.0
b&q hairpin legsPhrase£89320.0
stool hairpin legsPhrase£106220.0
hairpin legs amazonPhrase£6260.3

Almost all spend went to competitor-brand queries (screwfix, ikea, b&q, amazon) that predictably don’t convert — people searching “hairpin legs screwfix” want to buy from Screwfix. The campaign also shared overlapping terms with Top Performers, adding internal auction noise.

The second attempt repeated the pattern

General Search (SC) (launched Oct 2024) also used Maximise Conversions at £120/day. Its CPC trajectory:

MonthCPC on “hairpin legs”Notes
Oct 2024£1.62Launch (higher starting point than 2023 — market has moved)
Nov 2024£1.49Stable
Dec 2024£1.68Q4 seasonal rise
Jan 2025£1.39Post-holiday dip
Feb 2025£2.95Shopify checkout broke conversion tracking (documented in brand-search-vs-shopping.md)
Mar-Jul 2025£2.82-3.01CPC stuck at elevated level — same ratcheting pattern

The Feb 2025 CPC doubling has a clearer root cause: Shopify’s checkout broke conversion tracking, feeding bad data to Maximise Conversions. But the outcome was identical — CPC doubled and never reverted. Campaign paused July 2025 at 2.01x blended ROAS.


Factor 5: Quantifying the CPC Gap — Market Inflation vs Bid Strategy

The 4x CPC gap between AP and SC is NOT a single cause. Using Shopping CPC as a control for market inflation (same queries, different auction, different bidding), we can decompose it:

Shopping CPC on “hairpin legs” over time (market inflation benchmark):

PeriodShopping CPCText Search CPCRatio
2021 Q4£0.40£0.67 (AP)1.7x
2023 Q3£0.65£0.83 (SC launch)1.3x
2024 H1£0.85£2.01 (SC)2.4x
2024 H2£1.30£1.69 (SC campaign 2)1.3x
2025 H1£1.05£2.96 (SC)2.8x

Shopping CPC roughly doubled from £0.40 to £1.00-1.30 over the SC era — genuine market inflation of ~2-2.5x. This is unavoidable and consistent across the furniture/hardware vertical.

Text search CPC went from £0.83 to £2.96 — approximately 3.5x. The gap between Shopping inflation (2-2.5x) and text search inflation (3.5x) represents the bid strategy amplification component.

The decomposition:

FactorMagnitudeFixable?
Quality Score / account historyNone — QS is 9, CTR is comparable, CPCs matched at transitionN/A
Ad copy / landing page qualityNone — CVR is better under SCN/A
Market-wide CPC inflation~2-2.5x (evidenced by Shopping CPC trend)No
Maximise Conversions amplification~1.5x on top (two documented ratcheting events)Yes

The AP-era £0.51/click on “hairpin legs” is gone — market inflation has made it a £1.00-1.30 keyword. But the £2.50-3.00 CPCs SC experienced are NOT the market price — they’re the market price amplified by an automated bid strategy that ratchets up and doesn’t self-correct.

Evidence for the bid strategy component: Brand Search ran on Manual CPC for 22 months under SC (Aug 2023 - Nov 2025) and maintained CPCs of £0.40-0.85 normally. It experienced temporary spikes during specific incidents but always reverted. The generic search campaigns on Maximise Conversions showed the opposite pattern — spikes that never reverted.


The Verdict: Should SC Try Search Again?

The data suggests search is viable, but only with a fundamentally different approach:

  1. The CPC gap is partly fixable. About 2-2.5x of the gap is permanent market inflation (unavoidable). But the remaining ~1.5x is bid strategy amplification from Maximise Conversions — and that IS fixable. Both SC generic search campaigns used Maximise Conversions and both experienced the same pattern: stable CPCs for 3-4 months, a demand shock or tracking disruption, CPC doubling, and failure to revert. A different bid strategy could realistically achieve £1.20-1.50/click on “hairpin legs” instead of £2.50-3.00.

  2. Manual CPC is the obvious test. Brand Search ran on Manual CPC for 22 months at £0.40-0.85 with temporary spikes that self-corrected. Running generic search on Manual CPC with a £1.20-1.50 max CPC cap would give cost control that Maximise Conversions fundamentally lacks on thin-volume campaigns. At £1.30/click with SC’s demonstrated 5.2% CVR and ~£60 AOV, ROAS would be ~2.4x — not great, but far better than the sub-2x that Maximise Conversions delivered.

  3. The Shopping campaigns ARE serving the same search terms. “Hairpin legs” generates £4,521 in Shopping Top Performers spend at £1.05 CPC (vs £1.80+ in SC search). Shopping covers the same demand at a lower CPC. The core question is incrementality: does search capture conversions Shopping wouldn’t have? This can only be answered with a controlled test.

  4. If attempted, start narrow and stay on Manual CPC. Exact match on the top 5-10 converting terms only. Hard CPC cap at £1.20-1.50. No Maximise Conversions until the campaign has at least 30 conversions/month (the minimum for Smart Bidding to function reliably). Measure incrementality by comparing total account conversions with search on vs off.

  5. Do NOT repeat the Testing campaign mistake. SC’s (OLD) General Search Testing - SC spent £5,768 on competitor/retailer queries (screwfix, ikea, b&q, amazon) and generated ~15 conversions. Phrase match on competitor brand names is pure waste in this category — people searching “hairpin legs screwfix” want to buy from Screwfix, not from a DTC brand they haven’t heard of. Any future search campaign should use negative keywords to exclude competitor names from day one.


Brand Search: The Match Type Problem

A side finding from this analysis: the Brand Search campaign is leaking non-brand spend. In 2025, £1,837 (7.2% of Brand Search spend) went to non-brand terms like “hairpin legs”, “hairpin bench legs”, “wicked hairpin legs”, and “desk hairpin legs” at an average CPC of £4.31 (vs £1.79 for actual brand terms). These non-brand terms converted at 0.85x ROAS — a net loss.

The campaign is using broad match or broad match modifier for brand keywords, allowing Google to serve ads on loosely related queries. This should be tightened to phrase or exact match on core brand terms (hairpin leg company, hairpin leg co, the hairpin leg company, and common misspellings).

The brand CPC trend is also concerning: “hairpin leg company” (the top term) went from £0.58/click in 2023 to £1.28/click in 2025. This suggests either increased competitor bidding on the brand or quality score degradation. Either way, tightening match types would reduce wasted spend and improve the campaign’s ROAS from 5.38x (brand terms only) vs the blended 3.51x it’s showing in the 90-day view.


Appendix: Data Quality Notes

  • Search term data: google_ads_search_term_daily table, joined to ad_campaigns on ad_campaign_id for campaign name and bid strategy info
  • Campaign-level data: google_ads_raw_daily at level = 'campaign', joined to ad_campaigns on ad_campaign_id
  • Corrupted period: May 1 - July 31, 2023 conversion data excluded from all AP vs SC comparisons
  • Agency classification: AP campaigns by AP or AP| prefix; SC campaigns by - SC or (SC) suffix
  • Quality score data: google_ads_keywords table — only 7 keywords have QS data (all Brand Search). SC generic search campaigns have NULL QS (paused campaigns lose QS data). User confirms QS was 9 on SC generic search keywords when active.
  • Bid strategies: Confirmed via ad_campaigns.bidding_strategy(OLD) General Search Top Performers - SC: MAXIMIZE_CONVERSIONS; (OLD) General Search Testing - SC: TARGET_SPEND; General Search (SC): MAXIMIZE_CONVERSIONS; Brand Search - SC: TARGET_SPEND (Manual CPC until Nov 2025)
  • Change events: google_ads_change_events only covers Jan-Feb 2026 (34 rows) — cannot verify historical budget or bid strategy changes prior to this