Key takeaways
- Set brand exclusions at the account level before launch so PMax cannot cannibalize branded search traffic.
- Structure asset groups by audience intent, not creative inventory, and give each group distinct signals and creative.
- Conversion value rules are how you encode business logic into the algorithm. Without them, PMax optimizes for volume.
- Define a test end date and decision criteria before the campaign starts, not after budget has already been spent.
- Document every guardrail decision in writing so automation and new team members do not undo it later.
Performance Max guardrails are the structural constraints you set before the campaign launches so the algorithm optimizes toward your actual business goals rather than the proxy metrics it finds easiest to hit. Set them up front and the automation works with your strategy. Skip them and you will spend the next three months explaining to a client why branded search cannibalized prospecting budget.
This is a working checklist, not a philosophy piece. Go through it in order before you press launch.
Brand exclusions first, always
PMax will harvest branded search traffic if you let it. That traffic converts well, which makes the campaign look efficient, which gives the algorithm more confidence to keep doing it. The result is an ROAS number that impresses nobody who actually understands the account.
Set brand exclusions at the account level, not just the campaign level. Account-level negative keyword lists apply automatically to new campaigns, so a future PMax campaign does not undo the work. Include:
- Your brand name and common misspellings
- Your domain name and product names
- Executive names if they generate branded queries in your category
- Any co-branded partner terms you are buying separately
If you are running a dedicated branded search campaign alongside PMax, this separation is what keeps the two from competing against each other and inflating your blended CPA.
One practical note: Google does not allow negative keywords inside PMax campaigns directly. You apply them through account-level lists or by requesting brand exclusions through your rep. Verify the current method in your account before launch, as Google has changed the interface more than once.
Asset group structure reflects audience intent, not creative inventory
The most common PMax mistake is building one asset group and throwing every creative into it. The algorithm then mixes signals from audiences with very different intent and optimizes toward whatever converts fastest, which is usually the bottom of the funnel.
Structure asset groups around distinct audience signals and commercial intent levels:
- Prospecting group: signals from in-market audiences, competitor URLs in the audience signal, interest categories. Creative that introduces the product.
- Retargeting group: signals from site visitors, cart abandoners, CRM lists. Creative that closes.
- Customer list group (if your goal is upsell or cross-sell): signals from existing customers only. Creative that assumes familiarity.
Each group gets its own headlines, descriptions, images and video. Do not reuse the same assets across groups and expect differentiated performance. The algorithm will find the path of least resistance regardless of what you intended.
Keep the number of asset groups manageable. Three to five focused groups with strong signals outperform ten groups with thin signals every time.
Conversion value rules encode your business logic
PMax optimizes toward conversion value by default. If all conversions carry equal value in your account, the algorithm has no way to know that a demo request from a $50,000 ACV prospect is worth more than a newsletter signup.
Conversion value rules let you apply multipliers based on audience membership, location, or device. Practical applications:
- Apply a higher value multiplier to conversions from users on your CRM high-value segment
- Apply a lower multiplier to conversions from geographies that historically churn faster
- Downweight mobile conversions if your data shows mobile leads close at a significantly lower rate
This is not a workaround. It is the correct way to tell the algorithm what your business actually values. Without it, you are optimizing for volume and calling it performance.
Set these rules before the campaign accumulates significant spend. Changing them mid-flight shifts the optimization target and can destabilize a campaign that has already exited the learning phase.
Placement exclusions: run the report before you scale
PMax serves across Search, Shopping, Display, YouTube, Discover, Gmail and Maps. Display and YouTube placements in particular can generate impressions and clicks on inventory you would not consciously choose to buy.
Google provides a placement report for PMax, though it shows a subset of placements rather than a complete list. Run it weekly during the first month. Look for:
- Mobile app categories (children's apps are a common source of low-quality traffic)
- Parked domain traffic
- Placements that generate clicks but zero conversions across a meaningful spend threshold
You can exclude placements at the account level through the content exclusions settings. Build a running exclusion list and apply it to every PMax campaign in the account. This is one of those decisions that should live in your account's institutional memory, not in someone's head, so that the next person managing the account does not have to rediscover the same bad placements.
Set a test end date before the campaign starts
PMax campaigns have a learning phase. Google suggests waiting four to six weeks before making major changes, and that guidance is reasonable. The problem is that without a defined end date, tests run indefinitely. Budget keeps flowing, the team keeps watching, and nobody makes a call.
Before launch, write down:
- The date you will pull the first performance report
- The minimum spend threshold that makes the data meaningful (this depends on your CPA or ROAS target, but a rough estimate is enough conversions to achieve statistical confidence, typically 30 to 50 conversions per asset group)
- The specific metrics that will determine whether the campaign continues, scales, or gets paused
This sounds obvious. It rarely happens in practice. The result is that campaigns accumulate spend without a clear decision framework, and the account carries dead weight for quarters.
If you have run PMax tests before and killed them, document why. What was the failure mode: creative quality, audience signal, conversion value setup, or something structural? That record is what prevents the team from running the same test again in six months.
Budget allocation and campaign interaction
PMax and standard Search campaigns compete for the same queries in some cases. Google's stated priority is that exact match keywords in Search campaigns take precedence over PMax for those specific queries, but the interaction is more complicated in practice, particularly for phrase and broad match.
A few structural decisions that reduce unwanted competition:
- Keep your highest-value branded and non-branded exact match terms in Search campaigns with sufficient budget so they are not budget-constrained. A constrained Search campaign creates space for PMax to fill.
- Set PMax budget based on incremental goals, not as a share of total budget. If your Search campaigns are already capturing your core demand efficiently, PMax budget should be sized against the incremental volume you expect from new audiences and placements.
- Monitor Search Impression Share on your core terms after PMax launches. A drop in impression share on terms you care about is a signal that the campaigns are interfering.
Encode decisions so automation does not undo them
The structural problem with PMax, and with most automation-heavy campaigns, is that the decisions that shaped the account are rarely visible to the algorithm or to the next person on the account. The algorithm finds what converts. If what converts conflicts with a business decision made in a leadership meeting last month, the algorithm does not know that.
This is where most teams lose ground over time. A campaign gets set up correctly. Someone makes a business decision: a product line is being sunset, a geographic market is being deprioritized, a margin threshold has changed. That decision travels through Slack and a meeting recording and maybe a brief. It rarely makes it into the campaign settings or the documentation that an AI tool or a new team member would actually read.
The practical fix is to treat account decisions as documentation artifacts, not just verbal agreements. Every time a guardrail is set or changed, write down what it is, why it was set, and what would need to be true to change it. Keep that record somewhere the account team can actually find it.
Where Almond fits
Almond's Business Context Engine is built specifically for this problem. It reads your ad accounts, Slack, meeting notes, Notion, email and CRM, and captures decisions as they happen, versioned and ranked by relevance and recency. When your team decides on Monday to deprioritize a product line, that decision is in the context layer before the algorithm has a chance to contradict it.
For PMax specifically, this means the guardrails you set are not just in campaign settings that someone might override. They are part of the account's institutional memory, surfaced when a workflow or an AI recommendation touches that campaign. The Workflow Builder's 30 prebuilt Google and Meta automations run on that context layer, so automations act on current business judgment rather than stale triggers.
Almond 2.0 early access is open at almond.ai. Every member gets a detailed 1:1 Google and Meta ad account audit with the CEO, recorded.
Frequently asked questions
- What are the most important Performance Max guardrails to set before launch?
- Brand exclusions at the account level, conversion value rules that reflect actual business value, asset groups structured by audience intent, and a defined test end date with clear decision criteria. These four decisions prevent the algorithm from optimizing toward the wrong goals from day one.
- How do you add negative keywords to a Performance Max campaign?
- PMax does not support negative keywords directly at the campaign level the way Search campaigns do. You apply them through account-level negative keyword lists or by requesting brand exclusions through your Google rep. Build your exclusion list before launch and apply it account-wide so future campaigns inherit it automatically.
- How long should you run a Performance Max test before making a decision?
- Google recommends four to six weeks for the learning phase. More practically, wait until each asset group has accumulated 30 to 50 conversions. Set the end date and decision criteria before the campaign starts, not after. Campaigns without a defined end date tend to run indefinitely without a clear verdict.
- Will Performance Max cannibalize your branded search campaigns?
- It can, if brand terms are not excluded. PMax will harvest branded queries because they convert well, inflating ROAS while adding little incremental value. Set account-level brand exclusions before launch and monitor Search Impression Share on your core terms after PMax goes live to catch any interference early.
- How do conversion value rules work in Performance Max?
- Conversion value rules let you apply multipliers to conversions based on audience membership, location, or device. Use them to tell the algorithm that a lead from a high-value CRM segment is worth more than a generic signup. Set them before the campaign exits the learning phase to avoid destabilizing optimization.
Almond 2.0 early access
Get a 1:1 Meta and Google audit with the CEO.
Every waitlist member gets a 1:1 audit of their ad accounts. Work email only.