You finally did the structure everyone recommends. Brand sits in its own Search campaign. Performance Max is supposed to find new customers. Then account ROAS falls hard, MER gets worse, and sales feel softer even though you put more hours into the work.
That drop is usually the honest number showing up. The old PMax tile was carrying a lot of branded demand. Pull brand out, and you see what non-brand acquisition actually costs.
What the old ROAS was really made of
A blended PMax that includes brand often looks calm and profitable with almost no babysitting. Brand queries convert cheaply. People already know you. The campaign leans on that warm traffic, then reports a healthy ROAS as if the whole mix were prospecting.
Once brand is excluded, PMax has to fill the budget with colder inventory. Search themes get thinner. Display, YouTube, and Discover can quietly take a bigger share. Average CPC climbs because the easy branded auctions are gone. The dashboard that used to sit near a comfortable multiple now looks closer to 1.5x or 2x, and it feels like the account broke overnight.
Nothing new broke. The report stopped padding itself with brand.
Two changes often get blamed as one
Most of the wrecks I see stack two moves in the same week.
First, brand gets yanked into a dedicated campaign so the structure looks clean.
Second, someone raises the PMax budget hard so non-brand can replace the volume brand used to contribute. That jump can kick relearning, push the campaign into auctions it used to skip, and buy the same click count at a higher CPC. Efficiency falls even before you judge whether non-brand is profitable on its own.
If you only look at campaign ROAS after both changes, you cannot tell which lever caused the damage. Treat the split and the budget bump as separate tests.
Judge the business, not the pretty PMax tile
Campaign ROAS after a brand split answers a different question than whether the business is growing.
Check whether total sales and MER held. If branded Search is capturing the name traffic cleanly, and overall revenue is flat while PMax looks ugly, you mostly cleaned the scoreboard. If MER worsens and backend sales drop, non-brand spend is not paying for itself at the current mix.
Shops with thin gross margins cannot treat a 2x non-brand ROAS as fine just because a thread said to scale. Shops with strong margins and repeat purchase can live with a lower acquisition ROAS when lifetime value clears. The number only means something next to your margin.
Stabilize before you burn 15 hours a week "optimizing"
When performance tanks, the urge is to touch everything at once... new assets, new negatives, new Search Themes, daily bid tweaks. That much motion during relearning often makes the model worse.
Do this instead:
1. Pull PMax budget back to the last stable range
If you jumped from a quiet daily number to a much higher one the same week you excluded brand, step back toward the prior range for a few weeks. Lower spend can make the campaign choosier about auctions. Prove the structure before you ask it to scale.
2. Read the placement mix with brand gone
Open the channel and placement breakdown. If a big slice of spend shifted into Display, YouTube, or Discover after the exclusion, you are not looking at Search without brand. You are looking at upper funnel fill. Decide whether that mix is intentional. If it is not, tighten assets, add negatives, and steer with Search Themes before you add more budget.
3. Cap branded Search CPCs if your own name got expensive
Dedicated brand campaigns sometimes bid like they are fighting PMax forever. You do not need 90%+ impression share at any price. Manual CPC with a ceiling under the inflated average often recovers ROAS on brand without giving the whole SERP away. A few points of impression share lost can be cheaper than paying rent on every branded click.
4. Decide if non-brand at the new ROAS clears your math
Stop asking PMax to look like the blended era. Ask whether the non-brand CPA or ROAS clears margin after returns and sales cost. If it does not, cut waste and rebuild intent in Search before you force PMax to invent volume. If it does, keep the split and scale in small weekly inches, not double-day jumps.
What a healthy split looks like
A healthy setup after brand exclusion usually looks boring.
Brand Search owns the name queries with a modest budget and controlled CPCs. PMax or Search prospecting reports a lower ROAS than the old blended tile, and you accept that because MER and total sales are the scoreboard. Budget increases stay small enough that CPC inflation does not eat the gain. You stop measuring whether the dashboard feels as pretty as it did when brand was doing half the work in secret.
If you want a tighter playbook for the brand versus PMax boundary, exclusions, and how to keep the account from lying to you about acquisition, use the Brand Cannibalization Guide. It is built for this exact fork in the road, not for chasing a fake blended ROAS.
Know what new customers cost once brand stops padding the report. That is the whole job after the split.