Performance Max Strategy That Actually Scales (2026)

PMax has gone from "experimental" to "default ecommerce engine" in two years. But 7 out of 10 PMax campaigns I audit are set up wrong from day one. Here is the strategy that actually works in 2026 - built from 50+ live PMax accounts.

What Performance Max actually is (and is not)

PMax is not a replacement for Search. It is a cross-channel automation layer that uses one campaign to serve ads across Search, YouTube, Display, Discover, Gmail, and Maps. Google's algorithm decides where, when, and to whom - based on asset groups, audience signals, and conversion data you feed it.

Strengths: massive reach, automatic creative testing, scales fast on solid data. Weaknesses: low transparency, hard to debug, easily cannibalises Search if mis-configured.

The 6 inputs Google needs from you

1. Asset groups: Headlines, descriptions, images, videos, logos.
2. Audience signals: Customer match, in-market, custom segments, your converters.
3. Search themes: Up to 25 themes per asset group (acts as a hint to the algorithm).
4. Conversion goals + values: Tells PMax what to optimise for.
5. Bid strategy + budget: Max Conv Value or tROAS for ecommerce; Max Conv or tCPA for lead gen.
6. Exclusions: Brand keywords, placements, locations.

Skip any of these and you are giving the algorithm an incomplete brief.

Asset group architecture - the right way

The single biggest mistake: dumping all products into one asset group. Use this rule instead:

Examples by business type

• Ecommerce (apparel): Asset group per category - Men, Women, Kids, Accessories.
• SaaS: Asset group per use case - "for startups," "for agencies," "for enterprise."
• Local services: Asset group per service - AC repair, AC installation, AMC.
• Lead gen: Asset group per buyer persona - owners, decision-makers, influencers.

Each asset group needs minimum: 5 headlines, 5 long headlines, 5 descriptions, 5 images (square + landscape), 1 logo, and 1 video (10-30 sec). Skipping the video means missing YouTube inventory entirely.

Audience signals - the secret weapon

Audience signals do not restrict PMax targeting. They guide the algorithm to find people similar to who you tell it to look for. Think of them as accelerators for the learning phase.

What to load into every asset group

1. Customer Match list: Upload your past 6-12 months of customers (1,000+ minimum).
2. Top converters segment: Auto-built from Google Ads conversion data.
3. In-market segments: Pick 3-5 relevant ones from Google's library.
4. Custom segments from URLs: Add 5-10 competitor URLs Google will model from.
5. Custom segments from search terms: Add the 10-20 keywords your best converters use.

Skipping audience signals doubles your learning phase from ~14 days to 30+ days.

Search themes - your only direct keyword input

Search themes were a 2024 addition. They let you feed up to 25 keyword themes per asset group as a "strong hint" to PMax. Use them like this:

• Add your top-performing Search keywords as themes.
• Add category-defining terms ("buy running shoes online," "AC repair near me").
• Avoid super-broad themes ("shopping," "services") - they waste reach.
• Update themes monthly based on Search query reports.

The brand exclusion fix (critical)

Without brand exclusions, PMax will happily eat your brand Search traffic at higher CPCs than your dedicated brand campaign. As of 2024, brand exclusions are configurable directly in the Google Ads UI:

1. Open your PMax campaign → Settings → Brand exclusions.
2. Add your brand name + all common misspellings.
3. Add competitor brand names if you do not want them included.

This single setting can save 15-30% of PMax spend on most accounts.

The 4 PMax scaling phases

Phase 1 (Days 1-14): Learning

Do not touch anything. Volatility is normal. Watch for: zero conversions for 7+ days (problem) or steady upward trend (good). Budget should be at 100% of plan.

Phase 2 (Days 15-30): Optimisation

Refresh underperforming assets (Google grades them Low/Good/Best). Add 2-3 new images, refresh weak headlines. Add audience signals you skipped initially.

Phase 3 (Days 30-60): Tightening

Switch from Max Conv Value to tROAS (or Max Conv to tCPA). Set initial tROAS at 80% of current actual ROAS - not your goal. Tighten 5-10% every 2 weeks.

Phase 4 (Day 60+): Scaling

Increase budget 15-20% per week if performance holds. Add new asset groups for adjacent product categories. Test second PMax campaign for distinct audiences.

The 7 most common PMax mistakes

1. One giant asset group with everything mixed in.
2. No video assets (kills YouTube inventory).
3. No audience signals (doubles learning phase).
4. No brand exclusions (cannibalises brand traffic).
5. Setting aggressive tROAS/tCPA in week 1 (kills learning).
6. No conversion values flowing (forces tCPA when tROAS would win).
7. Letting PMax run final URL expansion unrestricted (sends traffic to weak pages).

When PMax does NOT belong in your account

• You have <30 conversions/month total.
• You sell only 1-2 SKUs/services with stable demand (Search alone wins).
• Your conversion values are flat (tROAS has nothing to optimise).
• You operate in a niche where Display/YouTube reach has zero relevance.

The bottom line

PMax done right is the most powerful scaling tool in Google Ads in 2026. PMax done wrong is the fastest way to burn ₹2L/month. The difference is structure, not luck.

If you want me to look at your PMax setup and find the 3-4 fixes that would compound - book a free 30-minute audit.

Related reading

• Manual vs Smart Bidding decision matrix
• 13 common Google Ads mistakes
• Pillar: Google Ads Services in India