Service · BOI®
Performance Max doesn't know what your business needs.
PMax can be extremely effective. It can also spend a spectacular amount of money while making the account increasingly difficult to understand.
Google view / Business view
- Conversions
- Growing
- Revenue
- On target
- Contribution
- Unknown
- Stock pressure
- Unknown
Automation needs commercial context.
01 / Mechanism, not verdict
PMax isn't good or bad. It's a mechanism.
Performance Max decides where ads appear, which products receive budget, which audiences to prioritise, which creative combinations to use and how individual auctions are bid. That can be incredibly powerful. But greater execution freedom makes the quality of the objective more important.
The problem is rarely too much automation. It's too little commercial context.
GOOGLE KNOWS
Clicks. Conversions. Revenue. Conversion probability. Audience behaviour. Auction signals. Product data.
GOOGLE DOESN'T
Margin. Contribution. Stock pressure. Lifetime value. Working capital. Product lifecycle. Buying-team priorities. Or whether finance needs £250,000 of ageing stock turned back into cash.
Every product shouldn't enter PMax under the same objective.
Five products all report 5.0 ROAS. One has 65% margin. One has 25%. One acquires exceptional repeat customers. One has six days of stock left. One has ten months of ageing inventory. Inside a blended campaign, Google can treat all five as equally attractive revenue opportunities. Commercially they are completely different. That's why we classify products before deciding how they enter automation.
PMax and margin
Two products, £50,000 revenue each at 5.0 ROAS, but one produces twice the contribution. Google sees similar conversion value efficiency. Finance does not.
PMax and stock
Google largely understands available or not available. A retailer understands five days of cover, nine months, incoming replacements, broken size runs and end-of-life. Stock status is not a binary commercial signal.
PMax and customer acquisition
Existing customers convert more easily. If repeat demand carries the campaign it can look extremely efficient while genuine acquisition is expensive. We look at new customer share, nCAC, first-order contribution, repeat rate and lifetime value.
PMax and brand demand
Capturing people already searching for you improves reported efficiency substantially. The commercial question is what would have happened without the ad. We use brand exclusions, campaign separation, search controls and holdout testing where possible.
PMax and creative
Broader inventory means creative matters, but ecommerce creative shouldn't be disconnected from intent. Asset groups should represent something commercially coherent, not act as a filing cabinet with seventeen lifestyle images.
Tracking matters too
PMax has no independent ability to realise you've given it broken conversion data. Duplicate transactions, incorrect values, brand-heavy conversions: it will cheerfully optimise. Machine learning doesn't make bad measurement less dangerous, it makes it scalable.
02 / Control model
When we relax control, and when we don't
Automation is expanded or restricted on evidence, not on a default setting.
When we move products into PMax
When the products have proven demand, their economics are understood, the feed is strong, conversion data is reliable, the commercial objective is clear, and broader automation offers a realistic opportunity to improve scale or efficiency.
Control can be relaxed when the evidence supports relaxing it.
When we keep more control
When the data is immature, tracking is uncertain, product economics vary significantly, we need clean query evidence, brand demand is distorting the picture, product testing is underway, stock is constrained, or the objective simply requires tighter control.
More automation isn't automatically maturity. Sometimes maturity means knowing when not to use it.
How we manage Performance Max
Establish commercial reality
Margin, contribution, stock, customer economics, product priorities.
Validate measurement
Tracking, revenue, new customer signals, conversion quality, ecommerce reconciliation.
Audit the feed
Product data, coverage, identifiers, attributes, labels, Merchant Center.
Understand existing allocation
What is PMax spending against, and what role do brand and existing customers play?
Classify products and define the role
Scale, Profit, Gateway, Protect, Recovery or none. Then decide exactly what we're asking the campaign to do.
Measure and reallocate
Contribution, POAS, nCAC, incrementality where testable, stock movement, marginal return. Then respond as the economics change.
We don't make changes because it's Monday, rebuild asset groups so the report looks occupied, or accept every recommendation because Google placed a blue tick beside it. And "the algorithm is learning" may be technically relevant, but it isn't an explanation. Automation doesn't remove accountability.
40%
Display waste identified per audit
Typical share of PMax spend going to low-intent Display placements.
+94%
Contribution margin, Thermos
After restructure and deduplicated conversions. See the Thermos case study.
75+
Ecommerce brands managed
Cross-sector PMax accounts under active management.
£18k
Avg. monthly spend redirected to higher-job SKUs
Reallocated from low-margin to high-margin SKUs after BOI restructure.
03 / Why PMax destroys margin by default
Performance Max doesn't know your commercial intent. By default, it optimises blind.
Google's Performance Max bidder takes the conversion value you feed it and scales spend toward whichever SKUs produce that value cheapest. If the conversion value is gross revenue, the bidder finds your loss-leaders. If you've never set value rules, every SKU is treated the same regardless of margin, stock position, or customer-acquisition role. That's the structural problem with PMax: not that it's a black box, but that it's a black box optimising against the wrong commercial signal.
Failure 1
Over-scaling on loss-leaders
PMax sees high revenue per click on discount SKUs and scales spend toward them. Revenue grows. Contribution margin collapses. The account looks healthier on the surface and worse underneath. Without commercial intent at the bid layer, this is mathematically inevitable.
Failure 2
Under-investment in Gateway SKUs
The SKU that brings a customer in for the first time is often lower margin than the SKU they buy second. PMax can't see the second purchase, so it under-bids the first. You lose customer acquisition velocity to competitors who do see the full LTV picture.
Failure 3
Bidding hard on Recovery SKUs
A product running low on stock or with squeezed margins shouldn't be aggressively bid on - it should be carefully harvested. PMax keeps pushing budget at it because nothing tells the bidder to ease off. You burn spend on sales you can't ship profitably or at all.
04 / BOI® - Bid On Intent
Every SKU has one commercial job. PMax bids against that job.
BOI® - Bid On Intent - is JudeLuxe's proprietary methodology. Every SKU in your catalogue is assigned one of five commercial jobs based on contribution margin, stock position, cash impact, and customer-acquisition role. PMax is then structured so the bidder receives commercial intent, not blended targets.
- Scale
- Profit
- Protect
- Recovery
- Gateway
Each job is implemented through PMax's actual controls: asset group segmentation by job type, value rules carrying the commercial intent into the bidder, brand exclusion preventing cannibalisation, custom labels in the feed for granular targeting, and weekly job reassignment as stock and margin shift. On complex platforms: especially Adobe Commerce PPC accounts with configurable variants and B2B price tiers: those custom labels have to be applied at the variant level, not the parent product.
It's not a layer on top of PMax. It's how PMax should have been structured in the first place.
Side by side
PMax vs Shopping - and why you almost certainly need both
Most agencies run one or the other and call it strategy. The honest answer: Standard Shopping and Performance Max do different jobs, and most ecommerce accounts need both, structured to complement rather than compete for the same auctions.
Job
High-intent purchase queries
Standard Shopping
✓
Performance Max
-
Why
Direct control over bidding by query intent vs. limited query-level visibility in PMax.
Job
Brand defence on Shopping placements
Standard Shopping
✓
Performance Max
-
Why
Clean brand isolation possible; PMax can cannibalise without strict exclusions.
Job
Specific SKU promotion / hero product push
Standard Shopping
✓
Performance Max
-
Why
Granular SKU-level bidding control; PMax bidder may divert spend elsewhere.
Job
Cross-network reach (Display + YouTube + Discovery)
Standard Shopping
-
Performance Max
✓
Why
Multi-network in one campaign; Shopping is Shopping placements only.
Job
Audience expansion to lookalike buyers
Standard Shopping
-
Performance Max
✓
Why
Strong audience signal architecture; Shopping has a limited signal layer.
Job
Holiday / peak demand coverage
Standard Shopping
-
Performance Max
✓
Why
Adapts faster to demand surges; Shopping manageable but more manual.
Job
New customer acquisition (Gateway SKUs)
Standard Shopping
-
Performance Max
✓
Why
Better at finding new audiences; Shopping is slower at this.
The right answer for most ecommerce accounts running £10k+/month is: Standard Shopping for high-intent + brand defence, Performance Max for cross-network expansion + new customer acquisition. Both structured under the BOI job framework so they're bidding against commercial intent, not duplicating effort.
What changes in your account
Week 1, month 1, ongoing
Week 1
- 01Commercial review: a 30-minute commercial read of where PMax spend is losing contribution
- 02Map contribution margin to every SKU in the feed
- 03Assign initial BOI job to each SKU (typical mix: 12% Scale, 51% Profit, 23% Protect, 11% Recovery, 3% Gateway - varies by catalogue)
- 04Implement brand exclusions if missing
- 05Set up cannibalisation monitoring
Month 1
- 01Restructure asset groups by BOI job
- 02Implement value rules carrying commercial intent into the bidder
- 03Add custom labels to feed for granular targeting
- 04Resolve Consent Mode v2 under-reporting if applicable (typical lift: 15–30% conversions recovered)
- 05Weekly job reassignment cadence live
Ongoing
- 01Weekly BOI job review based on stock, margin, and cash signals
- 02Monthly deep-dive on SKU-level performance and margin drift
- 03Quarterly audit of asset group performance and feed quality
- 04Always-on cannibalisation monitoring
- 05Quarterly platform update audit (e.g., new PMax controls, Consent Mode changes)
05 / Pricing transparency
What it costs
PMax audits at JudeLuxe are free. Not a 30-minute call with a templated PowerPoint - a full 5–7 day audit by a senior practitioner, with a written PDF report and a 60-minute walkthrough. You keep the report whether you work with us afterwards or not.
Retained PMax management is on a fixed monthly fee, scoped to your catalogue complexity and spend level. No percentage of spend, so we have no financial reason to push your budget higher than your margins can support. When we recommend more spend, the SKU data backs it.
Questions
Common Performance Max questions
How we decide
BOI®
BOI® (Bid On Intent) gives every SKU one commercial job at a time. Bidding follows that job rather than a blended target that quietly averages winners and leakers.
- Scale
- Profit
- Protect
- Recovery
- Gateway
Verified case study
Wilsons Pet Food
+212% revenue with subscription-led PMax
PMax rebuilt around subscription intent and lifetime value rather than blended ROAS. Asset groups split by buyer intent, feed re-engineered to match how owners actually search, signals fed from the P&L.
Read the full case study+212%
Revenue growth
+112%
Contribution margin
+83%
Customer LTV
Period, comparison basis and source are stated on the case study and evidence register.
Continue reading
Related services and reading
Next step
Find out what PMax is actually costing you.
A 30-minute commercial review. No pitch, no preparation required. If we are the right fit, we will set out what a full diagnostic would involve. If we are not, we will usually point you somewhere better.
No lock-in: 3-month initial engagement, then month-to-month with 30-day notice.