Shopping Catch All >£20 — Campaign Analysis
Summary
The account’s best-performing campaign by volume. Consistently profitable with 3.90x ROAS across 57 days and 237 conversions. The audit correctly identified it as the star, but the trend lines show gradual softening that warrants monitoring.
Campaign Configuration
| Setting | Value |
|---|---|
| Campaign ID | 71 |
| Name | Shopping Catch All - Over £20 - SC |
| Type | Shopping |
| Bidding strategy | TARGET_ROAS |
| Target ROAS | Not captured (NULL) |
| Daily budget | £150 |
| Serving status | SERVING |
Performance Data
Weekly Progression (57 days: Dec 12 – Feb 9)
| Week | Impr | Clicks | Cost | CPC | Conv | Revenue | ROAS | IS% | £/conv |
|---|---|---|---|---|---|---|---|---|---|
| Dec 8–14 | 27K | 289 | £240 | £0.83 | 12.9 | £1,074 | 4.48x | 72.8% | £19 |
| Dec 15–21 | 56K | 613 | £526 | £0.86 | 36.6 | £2,864 | 5.44x | 73.9% | £14 |
| Dec 22–28 | 30K | 395 | £225 | £0.57 | 13.9 | £996 | 4.44x | 75.8% | £16 |
| Dec 29–Jan 4 | 134K | 1,525 | £1,078 | £0.71 | 47.4 | £4,043 | 3.75x | 75.3% | £23 |
| Jan 5–11 | 104K | 1,055 | £736 | £0.70 | 38.0 | £2,851 | 3.87x | 72.4% | £19 |
| Jan 12–18 | 88K | 916 | £649 | £0.71 | 26.6 | £2,213 | 3.41x | 71.5% | £24 |
| Jan 19–25 | 84K | 938 | £628 | £0.67 | 24.1 | £2,614 | 4.16x | 67.0% | £26 |
| Jan 26–Feb 1 | 56K | 689 | £500 | £0.73 | 13.0 | £1,946 | 3.89x | 66.7% | £38 |
| Feb 2–8 | 68K | 891 | £786 | £0.88 | 21.7 | £2,456 | 3.12x | 68.3% | £36 |
Totals: £5,464 cost | 7,443 clicks | £0.73 CPC | 237 conversions | £21,305 revenue | 3.90x ROAS
Operational Note
Jan 29–30 impressions cratered to 4,252 and 1,323 (vs ~10K normal). This was caused by a GMC suspension during the hairpin.com domain switch — not a campaign performance issue. See ../../reference/operational-log.md.
Impression Variation Analysis
Daily impressions range from 1,323 to 23,605 (mean 11,512, std dev 4,476). The user flagged this as “huge variation” but noted CPC remains fairly constant. This is the key diagnostic: constant CPC with variable impressions = demand fluctuation, not a campaign problem.
Compare with PMax Knife Rack, where CPC swung from £0.04 to £1.20+ — that signalled Google mixing junk Display placements (penny clicks) with real Shopping clicks. Here, CPC stays in a tight £0.55–£1.24 band across all 57 days regardless of impression volume. This is pure Shopping traffic with natural demand elasticity.
What drives the variation
Seasonal demand (Dec 28 – Jan 4): Impressions surged to 14K–23K daily. Post-Christmas / New Year is peak home improvement browsing — gift money, renovation planning, furniture research. Organic demand spike, not a campaign setting change.
Operational events: Dec 23 dropped to 1,995 (Christmas Eve — nobody buying table legs). Jan 29–30 cratered to 4,252/1,323 (GMC suspension during domain switch — see ../../reference/operational-log.md).
Day-of-week pattern: Sundays average ~14,133 impressions vs weekdays at ~10,500–11,000. Weekend browsers have more time for considered-purchase research (table legs, shelving brackets, etc). This is a standard e-commerce pattern for non-impulse products.
Impression distribution
| Bucket | Days | % of total |
|---|---|---|
| Under 5K | 3 | 5% (all explained: Christmas Eve, GMC suspension) |
| 5K–15K | 44 | 77% (normal operating range) |
| Over 15K | 10 | 18% (mostly Dec 28–Jan 7 seasonal surge) |
The real concern: impression share, not impression volume
The absolute impression count fluctuating is demand-side — more people searching some days than others. That’s fine. The declining impression share (75.8% → 66.7%) is the problem. It means the campaign is missing a growing proportion of eligible impressions. Possible causes:
- TARGET_ROAS constraint — Google voluntarily declining auctions to maintain the ROAS target, even when those auctions might be profitable at a slightly lower return
- Budget capping on peak days — £150/day limit reached before all eligible searches served
- Competitive pressure — competitors bidding more aggressively, requiring higher bids to maintain share
The question for the Google Ads manager: would loosening the target ROAS slightly capture profitable impressions currently being left on the table? We can’t answer this without knowing the current target ROAS value (currently NULL in our data — Phase 1 fix).
Device Breakdown (30 days, Jan 11 – Feb 9)
| Device | Spend | % Spend | Clicks | CPC | Conv | Revenue | ROAS |
|---|---|---|---|---|---|---|---|
| Mobile | £1,519 | 54.8% | 2,599 | £0.58 | 61.6 | £5,234 | 3.45x |
| Desktop | £1,219 | 44.0% | 1,116 | £1.09 | 32.0 | £4,348 | 3.57x |
| Tablet | £36 | 1.3% | 68 | £0.53 | 1.0 | £229 | 6.40x |
Key findings:
- Mobile and desktop perform similarly on ROAS (3.45x vs 3.57x). No clear winner — both devices are healthy.
- Desktop CPC is nearly double mobile (£1.09 vs £0.58), but this is offset by a higher conversion rate per click on desktop.
- Mobile drives more volume: 61.6 conversions (66% of total) at a lower CPC. Mobile is this campaign’s workhorse.
- Tablet ROAS looks extraordinary (6.40x) but is based on a single conversion — statistically meaningless.
- No device requires bid adjustment. Both mobile and desktop are well within profitable range.
Search Term Cannibalization with Shopping Top Performers
Search term data reveals 105 terms appearing in both Shopping Catch All >£20 and Shopping Top Performers simultaneously. Top overlapping terms (30 days):
| Search Term | Top Performers | Catch All >£20 | ||
|---|---|---|---|---|
| Clicks / Cost | Conv | Clicks / Cost | Conv | |
| table legs | 6 / £10.67 | 1.0 | 5 / £11.22 | 1.0 |
| hairpin legs | 8 / £15.82 | 1.1 | 2 / £2.74 | 0.0 |
| desk legs | 3 / £5.09 | 1.0 | 4 / £4.73 | 0.0 |
| metal table legs | 2 / £6.03 | 1.0 | 2 / £5.36 | 0.0 |
| hairpin table legs | 4 / £9.03 | 1.0 | 2 / £1.11 | 1.0 |
| chrome table legs | 2 / £7.92 | 1.0 | 2 / £1.33 | 0.0 |
| breakfast bar legs | 2 / £3.33 | 0.0 | 3 / £7.20 | 1.0 |
| dining table legs | 2 / £2.43 | 0.3 | 4 / £5.00 | 0.3 |
Combined overlap spend: £190 (Top Performers: £97 + Catch All: £93)
The overlap is surprisingly even — neither campaign dominates the shared terms. Top Performers wins more conversions on hairpin-specific terms (hairpin legs, hairpin table legs), while Catch All picks up broader product terms (breakfast bar legs, dining table legs). This suggests the two campaigns are genuinely splitting auctions rather than one always winning.
Impact assessment: £190 overlap on £8,247 combined spend (2.3%) is not alarming. The cannibalization exists but isn’t material. If the campaigns were consolidated, the algorithm would simply bid on these terms once instead of twice — but the savings would be modest.
Analysis
Strengths
- Consistent profitability: Never had a full week below 3x ROAS. The floor is high.
- Volume: 237 conversions in ~8 weeks (~4.2/day). This is the account’s primary revenue driver.
- Efficient CPC: £0.73 average — the cheapest traffic of any campaign. TARGET_ROAS bidding is working well here.
- Budget utilisation: Average daily spend varies but regularly hits £100-140 of the £150 budget. Not wasteful, not constrained.
Concerns
Impression share declining: 75.8% (Dec 22 week) → 66.7% (Jan 26 week). The account is missing a third of available Shopping impressions for products >£20. The audit framed 69% IS as “room to scale” — but the direction is losing share, not just having headroom. This could mean:
- Competitors increasing bids
- The TARGET_ROAS target being set too high (conservative), causing Google to bid less aggressively
- Budget constraints on peak days limiting impression delivery
ROAS softening: 5.44x (Dec 15 week) → 3.12x (Feb 2 week). Still healthy, but a clear downward trend. Likely a combination of post-Christmas demand normalising and increased competition in January.
Cost per conversion doubling: £14-19 in Dec/early Jan → £36-38 in late Jan/Feb. This is the ROAS decline expressed differently. Acquiring each conversion is getting more expensive.
CPC creeping up in Feb: £0.67-0.73 in Jan → £0.88 in the Feb 2-8 week. Not dramatic yet, but if this continues alongside declining IS, it suggests competitive pressure.
What We Can’t See
- Target ROAS value: NULL in our data. We don’t know what the campaign is optimising toward, so we can’t assess whether it’s set appropriately. (Phase 1 fix in
shopify-app/docs/ad-platform/working-plan.md) - Product-level performance: Which products in the >£20 range are driving conversions vs wasting spend? Product daily metrics are now syncing (29K rows) but haven’t been analysed for this campaign yet.
PMax campaigns suppress this campaign’s product coverage
When PMax campaigns target the same products, PMax takes auction priority over Standard Shopping (Google’s documented behaviour). This has been confirmed with knobs/handles data: during the 12 days PMax: Knobs Only ran (Dec 12–23), Shopping Catch All could only show knob products at 1–5 impressions/day with zero clicks. After PMax was paused, Shopping took 5 weeks to ramp up before converting on knob terms (3 conversions at 7.25x ROAS from Jan 26).
This means any active PMax campaign is silently suppressing this campaign’s ability to serve for the same products. For high-volume products (table legs), this matters less because PMax allocates 98%+ to Search. For niche products (knobs, knife racks), PMax wastes budget on Display/YouTube while preventing Shopping from learning and optimising.
The current active PMax (Table Tops) has minimal overlap with this campaign’s core products. But if new PMax campaigns are created for niche product categories, this campaign’s performance for those products will be suppressed for the duration of the PMax campaign and for several weeks afterwards. See ./knobs-and-handles-analysis.md for the full analysis of this effect.
Verdict
The best campaign in the account by a clear margin. No immediate action needed, but the gradual erosion of impression share and ROAS should be tracked week-over-week. If IS drops below 60% or ROAS drops below 2.5x, the target ROAS and budget settings need review.
The “room to scale” claim from the audit is questionable — you’re losing share, not choosing not to spend. Scaling would require either higher bids (lower target ROAS) or higher budget, both of which cost more per conversion.