Choco Media has been running Google Performance Max campaigns for clients across e-commerce, B2B lead generation, and local services since it became the default campaign type in 2022. In that time, we have developed a clear picture of where PMax works well, where it leaks budget quietly, and how to structure it so the algorithm does what you actually want. This post is google performance max field notes — not a Google tutorial. It is what we have observed in real accounts, with real budgets, and the adjustments we make before handing a PMax campaign to the machine.
If you are running paid media in 2026 and have any Google spend, you are almost certainly dealing with Performance Max whether you chose it deliberately or not. Smart Shopping migrated to it automatically. Local campaigns did too. The question is no longer whether to use it — it is how to make it generate actual pipeline rather than impressions that look good in a dashboard but do not close.
This is for marketers and in-house teams who manage or oversee Google Ads and want an honest view from someone running PMax in the field. By the end you will have a clear sense of the structural decisions that actually matter, the signals to monitor, and the failure modes to avoid before they eat a month of budget.
What Performance Max actually is (and what it is not)
Performance Max is Google’s fully automated, cross-channel campaign type. A single PMax campaign can serve ads across Search, Display, YouTube, Gmail, Maps, and Discover — all from one set of assets and one budget. The pitch is straightforward: feed Google your creative assets and conversion goals, and its machine learning finds the best combination of channel, audience, and ad format.
In practice, PMax is a signal-amplifier. If your conversion tracking is clean and your conversion events represent real business value, it amplifies good signals. If your tracking is loose or you are optimising toward low-quality micro-conversions, it amplifies those instead — and you will not understand why spend is up but pipeline is flat.
- PMax does not replace campaign strategy — it executes on the signals you give it
- It works across six Google networks simultaneously from a single budget
- Asset quality, audience signals, and conversion event quality are the three main levers you actually control
- It is not a black box — it is a system that reflects your inputs back at scale
The biggest misconception we encounter
Most teams think that setting up PMax means building asset groups, choosing a budget, and launching. What it actually requires is thoughtful conversion architecture before the campaign goes live. The campaign is only as good as the signal it trains on.
How we structure asset groups (and why most accounts get this wrong)
The asset group is the fundamental building block of a PMax campaign. Each asset group is a bundle of headlines, descriptions, images, videos, and audience signals that Google uses to construct ads across its networks. The common mistake is building one large asset group and throwing everything in. We almost never do that.
Our standard structure groups by product category or offer type — not by audience. This is counterintuitive but important. Google handles audience distribution based on your signals and conversion data. What it cannot do well is serve contextually relevant creative to someone searching for a specific product if all your headlines and images are blended into a single group.
- One asset group per distinct product or service line — this keeps messaging coherent and makes performance data readable
- Three to five headlines per asset group that share a clear value theme, not a random mix
- Separate brand and non-brand asset groups where brand volume is significant
- Video assets matter even for non-video campaigns — without them, Google auto-generates videos from your images, and the output is usually poor
The brand asset group question
Whether to include brand terms in PMax is a recurring client conversation. Our position: if you have meaningful brand search volume and care about branded CPC, create a separate brand asset group with tight audience signals and monitor closely. Alternatively, run a parallel Standard Search campaign for branded terms and use campaign-level brand exclusions in PMax — this is now possible and worth setting up if brand spend is material.
Audience signals: the input most teams underinvest in
PMax does not use audience targeting in the traditional sense — it uses audience signals as starting hints. You are telling the algorithm “start looking here” rather than “only show to these people.” That distinction changes how you should think about building your signal lists.
The most valuable audience signals are first-party: customer lists, converters, high-value page visitors, CRM segments. If you have these, load them before launch. Campaigns seeded with strong first-party signals find efficient conversion patterns faster than those starting cold with interest-based audiences.
- Upload your existing customer list as a positive signal — the algorithm uses it to find similar users
- Segment audience signals by value tier if your customers have meaningfully different lifetime values
- Use the optimised targeting default and resist the urge to restrict it early — let it explore before you tighten
- Seasonal first-party data is worth re-uploading before high-spend periods like campaign launches or Q4
In client work we have found that accounts with clean, segmented first-party signals reach their target CPA roughly 30-40% faster than accounts starting with interest categories only. The difference is not marginal.
The asset quality problem most teams ignore
Google rates each asset on a scale and labels them Low, Good, or Best. Most teams treat this as a vanity metric. We treat it as a performance signal.
Asset ratings correlate with impression share on high-intent placements. If your headlines are generic or your images are low resolution, Google will de-prioritise your asset group on premium placements in favour of competitors with stronger creative. This is especially visible on YouTube and the Display network where visual quality matters.
We inherited a PMax account where the client had four headlines, two images — both text overlays on solid colour backgrounds — and no video. The campaign was spending at a cost-per-lead 60% above target. Within six weeks of adding proper creative: three lifestyle images, a 30-second brand video, eight headlines with distinct value angles — CPL dropped to target and impressions on competitive search terms increased noticeably. The budget did not change. The assets did.
- Aim for “Best” rating on at least two of your three primary asset groups
- Minimum: five headlines, two descriptions, three landscape images, three square images, one logo
- Add a real video — even a simple 30-second brand or product video outperforms auto-generated alternatives consistently
- Rotate fresh creative every six to eight weeks; asset fatigue is real even in automated campaigns
Conversion tracking is not optional — it is the whole game
This is where Performance Max either works or quietly burns budget. The conversion events you set as primary are what the algorithm optimises toward. If those events do not represent genuine business value, the campaign will optimise its way to a metric that looks good but does not move revenue.
We audit conversion setup before touching anything else when we take over an existing PMax account. The most common issues we find:
- Page-view or scroll-depth events set as primary conversions — these inflate volume but carry no business signal
- Phone call conversions with a 10-second minimum — almost any accidental dial counts
- Duplicate conversion events from both the Google tag and an imported GA4 goal
- Lead form submissions that include test submissions or internal traffic
For most clients we aim for primary conversions to be the highest-value action in the funnel — booked call, purchase, or qualified form submission — and we set everything else as secondary. This means the campaign optimises toward outcomes, not activity. For the technical side of proper tracking setup, our post on GA4 and server-side tracking attribution covers the implementation details we use in production.
Value-based bidding: when to use it
If your conversions have meaningfully different values — a booked call versus a newsletter signup, for instance — value-based bidding is worth implementing. Assign values that reflect relative business worth, even if approximate, and switch to Maximise conversion value rather than Maximise conversions. In our experience this shifts spend toward higher-value conversion paths over two to three weeks as the algorithm recalibrates.
Budget and bidding: the decisions that actually matter
PMax requires a minimum learning period of roughly two weeks and around 30-50 conversions before its bidding stabilises. This has real implications for budget setting. Accounts that launch with too-low a daily budget, or that apply aggressive target CPA constraints from day one, struggle to exit the learning phase and often get stuck in a loop of low volume and poor efficiency.
Our general approach with new PMax campaigns:
- Launch with Maximise conversions and no target CPA for the first three to four weeks
- Set a daily budget that allows for at least 10 conversion events per week — below this, learning is very slow
- Only introduce a target CPA after the campaign has 50+ conversions and you have a baseline to work from
- Tighten targets gradually — a 10-15% reduction at a time, with a two-week observation window between changes
If you are running PMax alongside standard Search campaigns, watch for cannibalisation carefully. PMax has priority over most other campaign types on branded and generic terms. If your existing Search campaigns start losing impression share after launching PMax, that is the likely cause.
What to monitor and what to ignore
The PMax reporting interface is deliberately limited. You cannot see search term reports the same way you can in Search campaigns, placement-level data is aggregated, and breakdown by channel requires jumping between Asset Group and Campaign-level views. This opacity frustrates most advertisers. We have learned to focus on a small set of signals rather than chasing data the interface does not surface.
Signals we monitor weekly:
- Cost per primary conversion — the only metric that tells you if the campaign is working
- Asset performance ratings — declining ratings across an asset group often predict performance drops before they appear in CPA data
- Search term insight categories — PMax provides bucketed intent categories; a high proportion of branded spend is a warning sign that the campaign is capturing cheap conversions you would have gotten from Search anyway
- Impression share on competitor terms — visible in Auction Insights, useful for understanding where budget is actually going
What we largely ignore: raw impression counts, reach metrics, and Display placement volume. These inflate dashboards without reflecting business outcomes. For a broader view of how we structure account reviews, our post on auditing a paid media account in 90 minutes covers the full framework we apply — PMax analysis is one section of that.
The failure modes we see most often
After running PMax across accounts ranging from small monthly budgets to significant spend, we have a clear picture of the patterns that cause campaigns to underperform:
- Launching before conversion tracking is verified — always confirm conversion events are firing correctly before spending
- Over-segmenting asset groups by audience — PMax handles audience allocation; asset groups should reflect product or offer logic
- Applying target CPA too aggressively too early — the campaign needs conversion data before efficiency targets make sense
- Ignoring asset quality ratings — poor assets are served less; this compounds quietly over time
- Not using brand exclusions when needed — if PMax is eating branded terms you were already capturing cheaply via Search, the campaign looks efficient but is not adding incremental value
- Treating PMax as set-and-forget — it needs creative refreshes, audience signal updates, and bidding adjustments every few weeks
The last point is worth emphasising. PMax is not a fully autonomous system. It needs fresh inputs at regular intervals — updated first-party data, new creative assets, and occasional bidding reviews — to maintain performance as market conditions shift. Teams that treat it as passive tend to see performance decay over three to four months without a clear explanation.
Is Performance Max right for every account?
Not always. There are account profiles where standard Search campaigns outperform PMax, and some where the two should run together with clear separation of role.
PMax tends to work best when:
- You have 30+ conversions per month — enough signal for the algorithm to train on
- Your product or service has clear transactional intent and meaningful search volume
- You have first-party audience data to seed the campaign
- Creative resources are available to produce images and video
We are more cautious with PMax when:
- Monthly conversion volume is below 20 — the algorithm does not have enough data to be useful
- The business is highly brand-dependent and protecting branded CPC is a priority
- The conversion funnel has a long sales cycle with no reliable mid-funnel conversion proxy
Our paid media service includes a structured account architecture review before any PMax campaign goes live — this is where we make the call on how PMax sits alongside existing campaigns rather than discovering conflicts after budget has been spent. If you are managing Google Ads and want a clear read on your current setup, reach out and we can take a look together.