Inside this resource.
9 sectionsA map of the published document. Choose a section to start reading.
- 01Plan the seasonal campaign work
- 02Historical BFCM Account Example
- 03Pillar 1: Multi-Feed Testing Architecture
- 04Pillar 2: RSA Geo-Segmentation System
- 05Pillar 3: PMax Segmentation Matrix
View 4 more sections
Prepare the layers before the promotion.
This is a prospective preparation sequence. Adapt the calendar and campaign mix to the account; the historical example is a separate record.
- Check the base
Review tracking, catalog readiness, stock and the actual offer.
- Plan the lanes
Define feed, market and campaign roles before splitting traffic.
- Prepare work
Build the message, assets and relevant destinations.
- Stage the launch
Verify the prepared work and agree the monitoring plan.
Read the full source: Pillar 1: Multi-Feed Testing Architecture · Pillar 2: RSA Geo-Segmentation System · Pillar 3: PMax Segmentation Matrix · The 30-Day Implementation Timeline
Jump to a section
Plan the seasonal campaign work
This guide includes a historical seasonal account example in its original context. The reported profit, spend and campaign counts describe that example, not a forecast. Compare consistent dates, attribution and commercial definitions before adapting a decision to your own account.
Historical BFCM Account Example
The original BFCM 2025 account review reported 39 campaigns and €44,321.96 in its profit field. The report's accounting inputs are not reproduced here, so do not read that field as independently verified net business profit. Campaign-level observations below retain their original labels; they do not isolate the effect of segmentation or establish a benchmark for other brands.
The M.R.P. method organizes three testable planning choices: multi-feed title testing, geographic RSA segmentation and a PMax segmentation matrix. More campaigns are not inherently better; retain enough budget and signal to interpret each segment.
Pillar 1: Multi-Feed Testing Architecture
The Problem: One product feed → one Shopping/PMax campaign → one set of titles → one type of search intent captured.
The Solution: Multiple feed variants testing different angles.
Here's what the account structure looked like:
| Feed Strategy | Campaign Example | Key Metric |
|---|---|---|
| Standard titles | Google Shopping (main) | 1.99% CR |
| Longtail-optimized titles | US B Feed Longtail | 7.41% CR |
| Brand-forward angle | PMax - WW - Brand | 7.83 ROAS |
| Geo + category split | PMax - IT/UK - All Categories | 2.59 ROAS |
The longtail-optimized feed - "US B Feed Longtail" in the account - delivered 3.7x higher conversion rate than the standard Shopping feed.
Same products. Different title structure.
Why This Works
Different search intents need different feed optimization:
- "Hammock" → Main feed captures this (broad intent)
- "Ultralight hammock for backpacking under 1lb" → Longtail feed captures this (specific intent)
These queries express different specificity. Review actual intent and product fit; the wording alone predicts neither conversion rate nor CPC.
How to Implement
Step 1: Prepare a private working copy of product data; retain valid product identity and review a supported experiment before changing Merchant Center Step 2: Rewrite titles with long-tail keyword patterns
- Front-load specific features
- Include use-case language
- Add specifications (weight, size, material)
Step 3: Choose an eligible comparison design with clear campaign roles; do not upload duplicate offers merely to create extra exposure Step 4: Predefine sample, conversion-lag allowance and risk limits; compare CR, ROAS and contribution under a suitable control
Test outcome: Measure conversion rate, contribution margin and uncertainty against comparable traffic; no 2x improvement is assumed.
Pillar 2: RSA Geo-Segmentation System
The Problem: One "Brand" Search campaign worldwide → different markets competing for the same budget → high-CPC markets (US) eat budget meant for efficient markets.
The Solution: Dedicated Search campaigns per geography.
Here's the actual performance breakdown:
| Market | Campaign | CTR | ROAS | Profit |
|---|---|---|---|---|
| US | Search - US - Brand | 23.03% | 8.36 | €14,590.69 |
| Worldwide | Search - WW - Brand | 24.03% | 7.83 | €7,252.05 |
| UK | Search - UK - Brand | 25.59% | 4.80 | €2,028.92 |
| Australia | Search - AU - Brand | 27.21% | 4.33 | €769.92 |
| Italy | Search - IT - Brand | 24.75% | 4.20 | €958.31 |
| France | Search - FR - Brand | 26.76% | 5.83 | €847.69 |
Every single geo-segmented brand campaign delivers 23-27% CTR.
These are reported observations for this account and period. Brand demand, markets, bids and other differences may contribute; no causal or reproducible CTR lift is established.
Why This Works
-
Markets can have different costs and economics. Compare dated observed CPC, margin and acquisition outcomes; no fixed US/UK/Italy cost ordering follows from the country name.
-
Market-specific bid control. Segmentation lets you set aggressive bids where ROAS is high and conservative bids where it's not.
-
Localized ad copy. Different currencies, shipping promises, local offers. RSA headlines can be tailored per market.
How to Implement
Step 1: Identify your top 3-5 markets by revenue Step 2: Create separate Brand Search campaigns per market Step 3: Set location targeting exclusively (don't use "people in or interested in") Step 4: Customize RSA headlines per market:
- Local currency
- Shipping timeframes
- Market-specific offers
Step 5: Set initial budgets proportional to historical revenue, then optimize based on ROAS
Evaluation: Set targets from the account's margins, brand demand and comparable historical data; the example is not a universal benchmark.
Pillar 3: PMax Segmentation Matrix
Most accounts I audit have one PMax campaign running the entire catalog. The pitch sounds reasonable: "Let Google figure it out."
The constraint is campaign-level budget control: asset groups share a campaign budget, although asset-group metrics are available. Separate campaigns only where independent budgets, goals or settings justify them.
When CPCs spike during BFCM, you're flying blind. (Review actual seasonal CPC patterns; a spike is not inevitable.)
The fix: PMax campaigns segmented by geography AND category/angle.
Here's the account structure:
| Campaign | Purchases | ROAS | Profit |
|---|---|---|---|
| PMax - US - All Categories | 53 | 2.40 | €378.88 |
| PMax - WW - Brand | 22 | 7.83 | €493.58 |
| PMax - AU - All Categories | 11 | 5.15 | €363.25 |
| PMax - UK - All Categories | 4 | 2.59 | €383.37 |
| PMax - IT - All Categories | 8 | 2.59 | €280.38 |
Notice: Not one mega PMax. Multiple segmented campaigns with clear naming conventions.
Why This Works
-
Geo-segmented PMax respects market dynamics. Australian customers behave differently than Italian customers. Separate campaigns = separate optimization.
-
"All Categories" vs "Brand" separation. Testing intent angles. Brand-focused PMax (WW - Brand) delivers 7.83 ROAS vs 2.40 for general categories.
-
Historical testing variants. The account shows OLD-P19, OLD-P21 campaign prefixes - evidence of ongoing testing, not set-and-forget.
How to Implement
Step 1: Map your current PMax structure Step 2: Split geographies only where business controls and interpretable signal justify separate campaigns; no universal US/UK/WW minimum Step 3: Consider intent splits:
- "All Categories" → general discovery
- "Brand" → brand-aware searchers
- "Best Sellers" → proven converters only
Step 4: Set asset groups per campaign to match the angle Step 5: Evaluate after the predefined sample and conversion-lag allowance, subject to risk guardrails; two weeks is not universal evidence sufficiency
Test hypothesis: Investigate whether segmentation improves the chosen business outcome after accounting for traffic mix and lag.
The 30-Day Implementation Timeline
You don't need to build all three M.R.P. components simultaneously. Here's a phased approach:
Week 1-2: Foundation
- Export current product feed
- Create longtail feed variant (rewrite top 20% of products first)
- Set up longtail Shopping/PMax campaign
- Document baseline metrics for comparison
Week 2-3: RSA Segmentation
- Identify top 3 geographic markets
- Create geo-segmented Brand Search campaigns
- Write market-specific RSA headlines
- Set initial budgets and bids
Week 3-4: PMax Matrix
- Split existing PMax by geography
- Consider brand vs category separation
- Set up asset groups per campaign
- Authorize the transition only after eligibility, tracking and settings readback, with a baseline and rollback plan
Week 4+: Optimization
- Compare longtail feed CR vs main feed
- Analyze geo-segmented ROAS differences
- Reallocate budget to highest performers
- Document learnings for next iteration
Your Next Seasonal Test
Use the historical campaign tables to frame questions, not expected results. Select one segmentation hypothesis, preserve a comparable baseline/control, define an evaluation period suited to volume and lag, and inspect contribution margin as well as platform ROAS. Verify Merchant Center product-ID/feed setup before creating alternate versions; duplicating feeds is not itself a valid experiment or a license to duplicate offers.
The 30-day sequence is an illustrative preparation schedule. Plan backward from the next actual seasonal event and available access/assets rather than treating the original 2025 timing as current.
What We Got Wrong
The M.R.P. Method came from BFCM failures. (Ours, specifically.)
Mistake: Building structure too late. First year, we started segmentation in November. Not enough time for Google to learn.
Campaigns underperformed during the critical window. Now we start building architecture in September. Two months of data before peak.
Mistake: Over-segmenting small catalogs. Applied the full M.R.P. structure to a 50-SKU account. Campaigns starved for data.
Fragmentation killed performance. Now we right-size the structure - small catalogs get fewer segments until volume justifies splitting.
Mistake: Ignoring geo CPC differences. First geo-segmented campaigns used the same bids everywhere. US ate 80% of budget while UK and AU starved.
We learned to set geo-specific budgets proportional to efficiency, not volume.
Mistake: Panicking during CPC spikes. When CPCs rise, review the actual magnitude and cause. First year, we raised tROAS targets reactively.
Campaigns got conservative and volume tanked. Now we hold targets steady and let the structure absorb the CPC increases.
The M.R.P. Method exists because we learned what doesn't work during the most expensive time of year. Follow it and skip our BFCM tuition.
Want the BFCM Prep Checklist?
We put together a week-by-week checklist for building this structure before peak season. Start in September, launch ready in November.
Continue with the resource library for related checklists.
Choose who handles the next step.
Use the reference yourself, book a focused paid working session, or have Tegra handle the ongoing channel work.
Search, Shopping, Performance Max and Demand Gen campaigns, plus the research, product feeds, static and video creative, and landing, offer and pre-sale pages behind them.
Explore managed Google Ads → · See paid working sessions →