We turned on Performance Max for a dental client last spring after months of hearing they were missing out on demand by not running it. Two weeks in, the campaign showed a lower cost per lead than their Search campaigns had ever hit. We were ready to call it a win until we pulled the search term insights and found nearly 40% of that spend was going to their own practice name, traffic that would have shown up for free.
That’s the trap with Performance Max. It optimizes for whatever conversion signal you hand it, and it doesn’t care whether that conversion was incremental or something you already owned. In a healthy account with the right guardrails in place, it’s one of the better additions we’ve made to a media mix in years. In a thin account, or one without brand exclusions set up first, it quietly inflates your numbers while spending your budget on the easiest wins available.
What Performance Max Actually Does
Performance Max is a single campaign type that bids across every Google inventory source at once: Search, Display, YouTube, Discover, Gmail, and Maps. You give it a handful of assets, headlines, images, video if you have it, along with a conversion goal, and Google’s automated bidding decides which channel and audience to show your ad to for each individual auction. There’s no keyword list to build, and until recently there was almost no way to see which search terms actually triggered a click. That’s a real shift from Search campaigns, where you control match types, negative keywords, and bid adjustments down to the query level. For a dental practice, it means one campaign might be showing your ad to someone typing “emergency dentist near me” into Google, someone scrolling a local news site, and someone watching a YouTube video about tooth pain, all pulling from the same budget and the same bid strategy. Google frames this as efficiency. We’d call it a trade: you give up granular control in exchange for reach you couldn’t build by hand.
Where Performance Max Helps
Established Accounts With Clean Conversion Data
Performance Max needs volume to learn, and it learns fastest in accounts that already have a year or more of accurate conversion tracking and a healthy number of monthly conversions to work from. If your call tracking is set up correctly, your form submissions are tagged properly, and you’re not double-counting or missing conversions, the algorithm has something real to optimize against. In that setup, we’ve seen Performance Max find demand a Search-only structure never touches: someone who saw a display ad two weeks before they searched, or someone who watched a dental implant explainer video on YouTube and called the next day. Those touchpoints don’t show up anywhere in a Search campaign, and Performance Max is genuinely good at finding them when the underlying data is trustworthy.
Multi-Location Groups
Practices running five or ten locations get a real time savings from Performance Max, especially with a location feed connected. Instead of building and maintaining separate Search campaigns for every office, asset groups can route a single campaign’s traffic to the nearest location automatically. We still recommend segmenting asset groups by service line rather than lumping everything into one group for the whole practice, but the location-routing piece alone is worth testing for any multi-location group that’s currently managing that complexity by hand.
Filling Gaps, Not Replacing Search
The accounts where Performance Max performs best are the ones where it’s running alongside a well-built Search campaign, not instead of one. When it’s picking up Display, YouTube, and Discover inventory that Search alone never reaches, it’s additive to what you’re already doing. The problems start when a practice turns off Search entirely and asks Performance Max to do that job too.
Where Performance Max Hurts
Brand Term Cannibalization
Performance Max will spend on your own practice name if nothing stops it, because branded searches convert at a high rate and the algorithm gets rewarded for that. Without brand exclusions set up before launch, you can end up paying for clicks you would have gotten organically for months before anyone catches it. That’s exactly what happened with the client we mentioned earlier: nearly 40% of a campaign that looked like a record-low CPA win was branded traffic, and it hid what non-brand demand was actually costing. The campaign looked great in the dashboard. It wasn’t, once we separated the two.
New Or Thin Accounts
An account with a handful of conversions a month, unreliable call tracking, or a brand-new Google Ads history doesn’t give Performance Max enough to learn from. The algorithm fills that gap with broad signals, and for a healthcare vertical, broad signals can mean impressions in places you wouldn’t choose on purpose. We generally want to see at least 30 to 50 conversions in the prior 30 days before we’ll recommend meaningful budget on a Performance Max campaign. Below that, you’re paying to teach the algorithm instead of generating patients.
Losing Visibility Into What’s Converting
Search term insights inside Performance Max are still an approximation, not a real search terms report the way Search campaigns give you. If you can’t see and exclude the queries and placements actually driving spend, you can’t tell the difference between a campaign finding new patients and one burning budget on irrelevant clicks. This is the main reason we keep a real Search campaign running with full negative keyword control even in accounts where Performance Max is active. It’s the only place left where we can see exactly what someone typed before they clicked.
Audience Signal Risk In A Healthcare Vertical
Performance Max leans hard on audience signals and Google’s own automated expansion to decide who sees your ads. For dental and healthcare accounts, that raises the same sensitivity questions we flag for HIPAA-conscious ad targeting generally: you want to be deliberate about what signals you’re feeding in, and you don’t want a platform inferring health-condition interest categories on its own. Left on full automation, Performance Max’s expansion settings can drift further than a practice would choose if it were picking audiences by hand. We turn off audience signal expansion by default on healthcare accounts and only add it back if a client wants to test it deliberately.
Here’s what that split typically looks like once we separate branded and non-branded traffic inside a Performance Max campaign that never had exclusions applied:
| Segment | Typical Share Of PMax Spend | Relative CPA | What It Actually Means |
|---|---|---|---|
| Branded | 30-40% | Lowest | Traffic you'd likely get for free |
| Non-Branded | 60-70% | 3-4x Branded | Your true acquisition cost |
| Blended (Dashboard CPA) | 100% | Misleadingly Low | Hides which dollars did the work |
Ask for the brand versus non-brand split on any Performance Max campaign before you trust the blended CPA. If that split can't be produced, the reporting is incomplete.
How We Decide Whether To Run It
Before we turn on Performance Max for any dental account, we go through the same list every time. This isn’t about being cautious for its own sake, it’s about not letting the algorithm learn from bad data before we’ve had a chance to check it.
- Conversion tracking is accurate and call tracking numbers are verified before Performance Max gets a dollar of budget
- Brand exclusions are in place from day one, not added after we notice cannibalization
- Search campaigns stay live alongside Performance Max so we keep keyword-level control and a real negative keyword list
- Asset groups are built around specific services, like implants, Invisalign, or emergency care, rather than one generic group for the whole practice
- Search term insights and placement data get checked weekly for the first month, then monthly once performance stabilizes
If your Performance Max campaign is the only thing running in your Google Ads account, or nobody’s checked what search terms and placements are actually converting inside it, that’s worth a second look before you add more budget to it. The campaign type isn’t the problem. Running it on autopilot is.
