Google Ads for jewelry brands: the playbook that takes a €120k/month account to -38% CPA

Jewelry is one of the highest-margin verticals on Google Ads. It is also one of the easiest to get wrong.

Average order value sits between €280-€2,800 depending on the brand. Margin spread is wider than any other ecom vertical - silver studs at 30% margin live in the same catalogue as solid gold engagement rings at 70% margin. Auction signal is noisy because "gold ring" can mean €40 plated or €4,000 solid. And the GMC compliance filters around metal claims, conflict minerals, and authenticity are stricter than most categories.

After running jewelry brands across cold launches, multi-store operators, and accounts inherited from previous agencies, we have a deterministic playbook. This is the one that took a €120k/month jewelry brand from 1.8 ROAS and -41% CPA YoY to 2.6 ROAS and -38% CPA in 9 weeks. Zero new spend. Zero bid increases. Full case study here.

CPA

-38%

ROAS

+44%

Conversions

+63%

New spend

€0

Why jewelry breaks generic Google Ads playbooks

Most ecom Google Ads playbooks are written for the average vertical. Jewelry is structurally different in three ways.

1. Margin spread is wider than any other vertical. Most ecom categories have 10-20% margin spread across the catalogue. Jewelry has 30-40% spread (plated silver to solid gold). One blended Performance Max ROAS target is structurally wrong - it either starves the high-margin solid gold or over-bids the low-margin plated.

2. Search queries are intent-ambiguous. "Gold necklace" can mean €40 plated or €4,000 solid. The same query routes a buyer who can afford your brand and a buyer who cannot. Without title-level disambiguation (solid 14k, 18k, vermeil), Smart Bidding cannot tell the conversions apart.

3. GMC compliance filters are stricter. Metal claims (14k vs 14kt vs 14karat), gold-plated vs gold-filled vs vermeil distinctions, conflict-mineral declarations for diamond brands, authenticity flags for high-value pieces. Generic ecom feed setups fail these checks routinely.

The fix is vertical-aware structure across feed, asset groups, and custom labels.

The three feed changes that compound on a jewelry account

Every jewelry account we onboard gets the same three feed changes shipped in the first two weeks. These are the moves that compound across every other optimisation.

Feed change 1: title rewrites for the top 80 SKUs

Pull search-term report. Cluster queries by metal type, weight, length, and occasion. Rewrite the top 80 SKUs by impression share with specs first.

Before: "Madeleine - Solid Gold Necklace - 14k" After: "14k Solid Gold Necklace, Madeleine, 18 inch, 4.2g, Womens"

The before-title matches "gold necklace" (intent-ambiguous, 92% bounce). The after-title matches "18 inch solid gold necklace women" (intent-clean, converts at 4x). Same product. Different match surface.

Time investment: 6 hours of focused work for 80 SKUs. The single highest-leverage move on a jewelry Google Ads account.

Feed change 2: metal_type custom labels

The most valuable custom label in the vertical. Tag every SKU with metal_type: solid_14k, solid_18k, plated, silver, vermeil, gold_filled. Feed into Performance Max listing-group rules so the algorithm can bid by metal tier separately.

Solid gold listings get a 4.0x ROAS floor. Plated listings get a 2.5x ROAS floor. Smart Bidding stops mixing solid gold conversion signal with plated conversion signal. Auction quality lifts immediately.

Feed change 3: variant-pricing item-group IDs

Same ring in 5 sizes, 3 metal types, 2 stones = 30 variant SKUs in the default Shopify feed. GMC reads 30 separate products with 30 different prices and treats the spread as misleading.

The fix: item-group IDs compress them into one product with variant attributes. PMax sees one product with 30x the conversion signal of any single variant. Smart Bidding compounds the data. Disapprovals for misleading pricing disappear.

These three feed changes compound on each other. Title rewrites improve match. Custom labels improve bidding. Item-group IDs improve conversion signal. The full lift typically shows up within 21-30 days.

Performance Max structure for jewelry

The default PMax setup for jewelry is one big campaign with all products. This is structurally wrong. The fix is splitting by metal tier and margin band.

Tier A campaign: solid gold + high-value pieces. Top 20% by margin and velocity. Typically solid 14k/18k, high-AOV diamond pieces, signature collections. tROAS 400-450%. ~50-60% of spend.

Tier B campaign: silver + plated + mid-tier. Rest of the active catalogue. Standard ROAS targets. Broader asset groups. ~30-35% of spend.

Tier C campaign: tail. Low-margin, slow-velocity, dead SKUs. Either paused or moved to a floor-budget PMax. ~5-15% of spend.

Each tier gets its own asset group structure, its own creative pack, and its own ROAS target. This stops Smart Bidding from cross-subsidising your worst products with your best.

Underneath the PMax stack:

Standard Shopping: bottom-funnel for branded queries and signature collections. Manual CPC control.

Search: branded queries (excluded from PMax via account-level brand exclusions), bottom-funnel buying intent ("18k solid gold necklace women"), brand-defence against competitors bidding on your terms.

Demand Gen: visual research phase. Jewelry has long consideration windows - buyers research for days or weeks before purchasing. Demand Gen on YouTube + Discover captures the research phase that Performance Max misses.

The full stack typically lifts ROAS 17-25% vs single-PMax setups. The PMax best practices guide covers the cross-stack split in more depth.

Asset packs and lifestyle photography cadence

Jewelry asset packs compound longer than any other vertical. Done right, lifestyle photography on solid pieces holds for 12+ months without CTR decay.

The cadence:

Lifestyle hero shots: every 12 months for evergreen collections, every 4 months for trend-driven pieces (Y2K silver, layered gold, statement chokers).

Product close-ups: every 6 months. High-resolution macro shots that show weight, finish, and stone clarity. Critical for high-AOV pieces.

Seasonal pivot packs: Q4 gifting (October ship), Valentine's (early January ship), Mother's Day (early April ship), wedding/engagement season (early May ship). Each window gets a dedicated asset pack 4-6 weeks before the buying season opens.

User-generated content: monthly. UGC is the highest-converting asset class on jewelry because trust signals matter more in the vertical than any other. Pull from social, get rights, rotate into PMax asset groups.

The reason cadence matters: stale creative kills CTR which kills Smart Bidding's ability to scale. Jewelry forgives stale creative more than beauty tools (4-6 week half-life) but punishes it harder than furniture (6-month stability). Get the cadence wrong and the account dies quietly over a quarter.

	Default Setup	Optimal Setup
PMax campaigns	1 (all SKUs)	3 (tier A/B/C)
Custom labels	None	metal_type + margin_band + velocity_30d
Title structure	Brand-first	Specs-first (metal, weight, size)
Branded queries	Captured by PMax	Excluded -> dedicated Search
Demand Gen	Not running	On for research phase
Item-group IDs	Per variant	Compressed
Server-side tracking	Client-side only	Server-side + enhanced conv
Asset rotation	Quarterly across catalogue	Per-collection cadence

GMC compliance for jewelry brands

Jewelry hits GMC compliance filters more often than most verticals. The four common issues:

1. Metal claim consistency. "14k" vs "14kt" vs "14 karat" - GMC reads them as different. Standardise across feed, product page, and ad copy. Mismatches trip authenticity flags.

2. Gold-plated vs vermeil vs gold-filled. Each is a distinct GMC category with its own disclosure requirements. Vermeil specifically requires sterling silver base + minimum gold thickness disclosure. Filing the wrong category triggers misleading-listing flags.

3. Conflict-mineral declarations. Diamond brands need conflict-free certification visible on product pages. Without it, GMC flags listings for restricted-product violations.

4. Stone authenticity for high-AOV pieces. Lab-grown vs natural disclosure required. "Diamond" alone is not enough - GMC wants lab_grown or natural specified.

The autofix engine has a tagged jewelry playbook for each of these. Most disapprovals clear within 24-48 hours of detection. For active suspensions, the discovery call walks through the suspension reason and we tell you on the spot whether the account is recoverable.

What this means for your jewelry brand this quarter

If you run a jewelry brand on Google Ads in 2026, three structural moves matter more than anything else.

One: Split Performance Max by metal tier. One PMax for everything is the most common mistake on jewelry accounts. The fix is a 4-hour restructure that compounds for the next 12 months.

Two: Rewrite the top 80 SKU titles with metal + weight + size + occasion. 6 hours of focused work. Single highest-leverage move on a jewelry Google Ads account. The case study lift came from this plus two other moves.

Three: Install server-side tracking. Jewelry has long consideration windows and high-intent return visits. Client-side pixels miss 25%+ of attributed revenue. Smart Bidding cannot scale a jewelry account on broken data.

For the full playbook self-serve, the Google Ads eCom Lab on Skool covers the structure, the custom labels, and the GMC compliance work. 740+ ecom operators inside, several running their own jewelry brands.

For done-for-you on a jewelry brand at $5K-€500K/month spend, drop the store URL on WhatsApp. We pull the account up live, walk through the GMC compliance first, and tell you on the spot whether the agency math works for your scale.

To compare how the same engine flexes across the other six ecom verticals we run, the metal-tier work here mirrors the AOV-band work on home decor and the freight-margin gating on furniture - same structural discipline, different label set.

The bid strategy was the last thing we touched. The first three weeks were feed, labels, and structure. The algorithm did the rest.

The jewelry brands that compound on Google Ads in 2026 are the ones that treat the structural work as the lever and the bid strategy as the consequence. Most still do it backwards.