Optimizing Performance Max doesn't start with more buttons
PMax rarely fails because of one button
A business owner looks at Performance Max and sees mostly fog. A few charts, some asset scores, a ROAS that fluctuates week to week, and a Google recommendation that feels just a little too enthusiastic. The reflex is understandable: where do I click to make this better?
But that's usually the wrong question. Performance Max doesn't become uncontrollable because Google hides the buttons. It becomes uncontrollable when you don't treat asset groups, conversion value, lead quality, and insights as a system. Then optimization turns into a series of disconnected actions — just 'doing little things'.
Asset groups are your first control layer
Google describes asset groups as the place where you bring together creative assets, signals, and products or services. In plain language: this is the package Google uses to understand who you want to reach and with what message. If everything is mixed together in there, don't be surprised when the results feel equally mixed.
In lead generation, I often see asset groups created by gut feeling. One group for 'general', one for 'SMB', one for 'premium', and then three more variants because someone wanted to test a new headline. That sounds active. But when every asset group has a different audience, a different promise, and a different conversion value, you learn very little. You're testing everything at once and nothing properly.
- Give each asset group one clear role: audience, service, product category, or purchase intent.
- Use headlines that address the same problem as your landing page.
- Remove assets that only exist because Google once suggested them.
- Check that images, video, and copy all deliver the same promise.
- Don't put completely different services in the same asset group if you want to evaluate them separately.
Conversion value is not a luxury field
For e-commerce, conversion value feels obvious. An order worth 800 euros is different from one worth 40 euros. For lead-based businesses, it often gets vague. Every inquiry gets the same value, or no value at all. Then Performance Max is left to guess which leads are good. And then we're surprised when the system attracts cheap form submissions.
Google's own Performance Max documentation keeps hammering on conversion tracking, conversion value, and signals like new customers. Not because Google loves tidy administration. It's because smart bidding without good value input will optimize for the easiest measurable action. A form submission is then just a form submission, even if sales knows within three minutes that it will never become a customer.
| Situation | What Google sees | What you actually want |
|---|---|---|
| All leads have value 1 | Every inquiry looks equal | More value assigned to leads with real buying intent |
| Offline sales data is missing | PMax stops at the form | Feedback on appointments, quotes, and actual customers |
| New customers not valued separately | Existing demand may get too much weight | Room for new-customer growth |
| GA4 and CRM don't communicate | Campaign learns slowly or incorrectly | One measurement line from click to quality |
Insights without follow-up are reporting theater
Search term insights and audience insights are useful, but only if someone actually does something with them. A search category can reveal that people search differently than you expected. An audience insight can show that a segment responds surprisingly well. But if that insight never feeds back into your landing page, asset group, or sales narrative, it stays a nice screenshot for the monthly report.
This is where the 'full funnel approach' becomes concrete. An insight from Ads needs to be able to land in content, CRO, automation, and follow-up. Say PMax is picking up a lot of searches for 'marketing automation costs'. The question isn't just whether you should bid on that. The question is also whether your page addresses the pricing hesitation, whether your automation follows up on that doubt, and whether sales knows what pain lies underneath it.
- Note the three most important PMax insights every month.
- Link each insight to one decision: adjust an asset, improve a page, change a bid, or add a sales question.
- After four weeks, measure whether that decision changed anything in lead quality or conversion rate.
- Archive insights that don't lead to action, so you stop reporting just for the sake of reporting.
- Have sales feed back which PMax leads are a good fit and which are not.
The monthly PMax feedback loop
A healthy PMax approach needs rhythm. Not daily panic, not six months of inaction. Monthly works well as a baseline for many SMB accounts: enough data to spot patterns, short enough to limit budget leakage. Especially when your budget isn't unlimited.
My practical order is simple. First, measurement foundation: are conversions, values, and imports correct? Then structure: are asset groups logical and not duplicated? Then quality: which leads, products, or customers came out of it? Only then optimization: bidding strategy, budget, assets, and landing pages. Anyone who reverses that order is usually polishing the dashboard while the foundation is crooked.
Performance Max doesn't ask for more button-pushing. It asks for Direction, Focus, and only then Scaling. Know what type of customer you're looking for, give Google clear value input, and close the loop with real business quality. PMax will still not be a glass house, but it will be a lot less foggy. And for many accounts, that's already a significant step forward.
