Performance Max in 2026: the agency playbook that actually works

Performance Max launched in late 2021 as Google's bet that giving up control would produce better results. Five years in, the campaign type now dominates retail spend, runs across every Google surface, and accounts for the largest single source of "why did our ROAS drop this week?" Slack messages on every paid media team we've worked with.

The 2026 version of PMax is meaningfully different from the 2024 version. Google has shipped real controls — search themes, brand exclusions, segmented Search Terms reports, asset group experiments — that most agencies haven't operationalized yet. And the operator habits that worked in 2023 are now actively destroying campaign performance.

Here's what we're seeing across partner agency books in 2026.

The structural shift: fewer campaigns, more asset groups

The most common 2024 PMax structure was wrong: one campaign per product category, sometimes one campaign per ad group equivalent. Operators thought tighter campaign segmentation would give them more control.

It didn't. It fragmented the conversion signal and starved each campaign of the data volume PMax's algorithm needs to optimize. The 2026 best practice is the opposite: fewer campaigns, more asset groups within each campaign.

• 1-2 PMax campaigns per major business goal (e.g., "new customer acquisition" vs. "existing customer retention")
• 3-5 asset groups within each campaign, thematically distinct
• Minimum 50 conversions per 30-day window per campaign before applying Target ROAS bidding

If your campaign isn't hitting that 50-conversion floor, it's not going to optimize. Consolidate before you do anything else.

Search themes are the most underused lever in PMax

Google added Search Themes to Performance Max in mid-2024 and most operators still haven't built them into their standard workflow. They're hints — not hard targeting — that tell PMax's algorithm which queries you want it to prioritize. They work especially well for surfacing queries that aren't yet in your historical conversion data.

The 2026 use case: use Search Themes to push PMax toward high-intent queries that don't yet have conversion volume but match your ICP. New product launches, expanded service categories, geographic expansion — all situations where you know the demand exists but PMax's algorithm hasn't seen the data yet.

The Search Terms report — read it weekly

Google now gives PMax a limited Search Terms report (added 2023, expanded 2024-2025). It's not as complete as the standard Search campaign report, but it's actionable. Most agencies still aren't pulling it weekly.

What to do with it:

1. Negative-keyword obvious mismatches. PMax campaigns regularly serve on queries that have nothing to do with your client. Add them as account-level negatives.
2. Brand exclusions. If your client also runs a brand campaign, PMax will cannibalize it. Use the brand exclusion list (Google added this in 2023) to keep PMax off branded queries.
3. Surface emerging intent. Queries with low impression volume but high conversion rate are signals to create dedicated Search campaigns alongside PMax.

Feed quality matters more than campaign structure

For e-commerce clients, the single highest-leverage PMax improvement is almost never inside the campaign settings. It's in the product feed.

Title optimization, accurate categorization, complete attributes (color, size, material, brand), high-quality images at the right aspect ratios, accurate availability and pricing data — all of this drives PMax performance directly because the algorithm uses feed data to match products to queries and audiences.

If your client's feed has 15% of products with incomplete attributes or stale availability data, that 15% is dragging the entire campaign. Feed audit and ongoing feed optimization should be part of every PMax engagement. If it isn't, you're optimizing the wrong layer.

What operators are doing that's actively hurting performance

Across the dozens of PMax audits we've run for partner agencies, the same six anti-patterns show up:

Anti-pattern	Why it hurts
Daily ROAS reactions	PMax fluctuates 20-30% day-over-day; daily changes reset the learning phase
Single-day budget changes	Triggers algorithm panic; serves wrong audiences while reoptimizing
Disabling "Low" assets in week 2	Asset scoring takes 6+ weeks to mature; early removal kills variants that would have won
Adjusting Target ROAS more than once per 14 days	Forces algorithm restart; you're never out of learning phase
Adding new asset groups before existing ones have 4 weeks of data	Splits volume across too many groups; nothing optimizes
Switching to Target ROAS bidding before 50 conversions/30 days	Insufficient data; algorithm makes poor decisions

If you're doing two or more of these on any campaign, that campaign is in perpetual learning mode and Google's algorithm is making decisions on fragmented data. Stabilize for 30 days before trying to optimize.

The reporting structure that protects your retainer

PMax's opacity creates a specific client-communication problem: when ROAS drops 18% in a week, the client wants to know exactly why, and PMax doesn't fully tell you. Most agencies handle this badly — either by overpromising explanations they can't actually deliver, or by hiding behind "the algorithm" and looking incompetent.

The structure that works in 2026:

• Weekly: One-paragraph Slack update with ROAS trend, impression volume, top 3 search terms, and one operational note.
• Monthly: Full report including Search Themes performance, asset group breakdown, feed health score, and recommendations for the coming month.
• Quarterly: Strategic review including PMax-vs-Search incrementality (if you've run experiments), audience overlap with other campaigns, and budget reallocation recommendations.

The point isn't to overwhelm the client with data. The point is to demonstrate that you're operating PMax on a structured cadence, not just watching the dashboard and reacting.

What to charge for PMax management

PMax-only retainers (e-commerce accounts, no Search/Shopping mix) run $3,000-$8,000/month wholesale at most agency partnerships in 2026, with retail markup typically landing in the $6,000-$15,000 range depending on ad spend volume.

The pricing variable that matters most is account complexity: SKU count, geographic spread, brand variants, and feed automation needs. A 500-SKU DTC brand running $50K/month in spend is roughly twice the operational cost of a 50-SKU brand running the same spend.

"PMax in 2026 is not 'set it and forget it.' It's 'set it, monitor it daily, intervene weekly, restructure monthly.' Agencies still treating it as automation are losing campaigns to agencies treating it as managed automation."

The TL;DR for agency owners

If you're running PMax for clients in 2026:

1. Consolidate campaigns. Fewer, with more asset groups.
2. Build Search Themes into every campaign and update monthly.
3. Pull the Search Terms report weekly; add negatives and brand exclusions.
4. Audit feed quality before optimizing campaign settings.
5. Stop making daily reactions. Set a weekly intervention cadence and stick to it.
6. Structure your client reporting around three tiers (weekly Slack, monthly written, quarterly strategic).

None of this is glamorous. All of it works. The agencies winning PMax retainers in 2026 are the ones treating it as a structured operational discipline — not a magic black box.