← Back to blog

Does Performance Max Really Work? An Honest Take After 3 Years

What Performance Max actually is

Briefly, for context — because Google's definition is fairly vague. PMax is a campaign that uses one budget to advertise you across Search, Shopping, Display, YouTube, Gmail, Discover and Maps at the same time. The decision about where to show the ad, in what format and to whom — is made by Google's algorithm, based on signals (your conversions, audience signals, search themes, asset groups).

PMax launched in Google Ads in November 2021; in March 2022 it replaced Smart Shopping and Local campaigns. For 3 years it has been Google's flagship product — heavily promoted by reps, pushed inside the new-campaign wizard and surfaced through AI recommendations in Google Ads. Some of that promotion is justified, some is Google's sales narrative that you should resist.

When PMax works — 4 scenarios

1. Mature e-commerce with a solid feed

This is PMax's "home turf". A store with at least 30–100 SKUs, GTINs, good titles, photos, a return policy. 30+ conversions per month with transaction values (50+ if you use tROAS — see below). Custom labels in the feed for segmentation. Here, in my practice, PMax adds 10–22% higher ROAS than Standard Shopping run alongside it (the range varies by industry) — provided it is fed with data.

2. Brand with strong branded search as a scale layer

If people search for you by name and competitors are bidding on your brand keywords, PMax adds scale across Display, YouTube and Gmail without eating into your branded Search budget (if you properly isolate it in Standard Shopping and Search). It is a good complement for brands that already have demand — not for startups building awareness from scratch.

3. Value-based conversions + offline conversions (B2B with CRM)

This sounds paradoxical — earlier I wrote that PMax does not work for niche B2B. Exception: B2B with a CRM that returns SQL and Closed-Won signals via OCI (Offline Conversion Import). The algorithm gets information about the real deal value and optimizes for cost per Closed-Won, not cost per lead. From my experience in B2B SaaS and business services — PMax only makes sense once OCI is in place. Without it, it optimizes for the cheap lead.

4. Customer Match signals enriched with the CRM

A list of 5,000+ customers in Customer Match, lookalikes, intent signals — that is fuel for the algorithm. The more context, the better PMax's decisions. Accounts that feed PMax only the default asset group plus a budget perform much worse than those that give it the full context.

To wrap up B2B: without offline conversions (OCI) do not bother. With OCI — it is one of the best PMax use cases for businesses with long sales cycles.

When PMax does NOT work — 4 scenarios

1. New account with no conversion history

The system learns from the account's historical data. Without 3+ months of conversion data, PMax starts blind and burns budget trying to find a pattern. Classic case: the owner turns on PMax from day one, sees a terrible ROAS after 14 days, panics, turns it off. The money is gone, the system did not have time to learn. Better: start with Standard Shopping + Search, turn on PMax after 30–60 days when you have data.

2. Low conversion volume (less than 30/mo.)

Smart Bidding requires at least 30 conversions in 30 days for tCPA, 50+ for tROAS — those are Google's official thresholds. Below that, PMax becomes a random decision generator. One conversion every two weeks does not create a pattern. If you have less — either wait, or optimize for an earlier funnel event (e.g. demo request instead of SQL) so the algorithm has more signals.

3. Niche B2B with a long sales cycle, WITHOUT OCI

This is the trap I see most often at B2B clients. PMax without offline conversions (CRM → Google Ads) optimizes for "leads" — i.e. submitted forms. The system finds students filling out whitepapers and consultants testing demos. Real sales never appear in the data. ROAS in the panel looks OK; in the CRM it is a disaster. Fix: deploy OCI before launching PMax, not after.

4. You require full control (brand defense, premium placements)

Some industries need control: where I show, on which queries, in which apps (B2B SaaS, law firms, finance, medical). PMax by definition hands that to the algorithm. You can add signals, but you never know exactly where the budget is going. For these clients, the choice between PMax and Standard Search is not a ROAS question — it is a brand-risk question.

Four biggest problems with PMax

Problem 1: Lack of transparency

Standard Shopping shows you exactly: which queries drove clicks, which products sold, what the CTR was per product. PMax shows aggregates. You know the campaign earned 50,000 PLN, but not whether from YouTube or Search. You know the top queries are "expanded" but you will not see most search queries. That makes diagnosis and optimization hard.

Google adds a bit of transparency every few months (Insights tab, asset reporting, search themes), but you are still much more in the dark than with Standard Shopping.

Problem 2: Brand cannibalization

This is one of the most common problems I find in accounts I take over. PMax with default settings catches branded search ("brandname + product") and reports it as "its own" conversions. Result: PMax shows great ROAS because it is scooping up cheap branded-search traffic. The rest of the campaigns (which actually acquire customers) look weak.

Fix: brand exclusion list in PMax (Account-level Brand Exclusions in Google Ads) — add your brand and competitors so PMax does not catch them. That is the first thing I check after taking over an account with PMax.

Problem 3: It catches cheap queries instead of valuable ones

Without value-based conversions (different values per type), PMax optimizes for the number of conversions. It catches the cheapest queries, the cheapest products, the cheapest customers. Classic case: e-com with mixed margins sees rising sales volume but falling revenue. PMax sold 200 products at 50 PLN instead of 50 products at 500 PLN. Fix: value-based conversions + custom labels in the feed + tROAS instead of Maximize Conversions.

Problem 4: Slow learning and overreaction

Change anything (target ROAS, asset groups, budget) and the system goes back into the learning phase. For 7–14 days the results are unstable. Every nervous decision resets the counter. Conclusion: PMax requires patience and discipline — either you make thoughtful changes once every 2–3 weeks, or you do not change anything. The "today I bump ROAS, tomorrow I cut it" strategy kills results.

PMax vs AI Max — what they are and how they differ

One of the most-asked questions in 2026 — because Google announced AI Max in May 2025 and many people mix the two up. In short: this is not a replacement, these are two different features.

Aspect	Performance Max	AI Max for Search
What it is	A separate campaign type — combines Search + Shopping + Display + YouTube + Gmail + Discover + Maps	A feature you turn on inside existing Search campaigns — expanded keyword matching + dynamic headlines + landing page expansion
Where it shows	All Google networks (cross-network)	Only Search + Search Partners
Keyword control	Search themes (signals, not exclusions)	Standard match types + AI keyword expansion + brand controls
Creative	Asset groups (headlines, descriptions, images, video)	Your existing RSAs + AI-generated headlines / descriptions
Landing page	Asset group → final URL	Final URL expansion on by default — AI may pick a different page on your domain than the one you set. You can disable it.
Who it is for	E-commerce with a feed, brands with mature accounts	Lead gen, B2B, services, brands that want to stay in Search

In short: PMax is "go everywhere, the system decides where", AI Max is "stay in Search, but let AI expand matching and creative". I often see clients thinking AI Max is "the new PMax" — that is not true. They can also run in parallel, if you properly isolate brand and queries.

I will cover AI Max in detail in a separate series (TODO: link to #22 "What Is AI Max", #23 "AI Max — Pros and Cons After a Year", #24 "AI Max Setup Step by Step"). Here it is enough to know that they are two different features and they do not replace each other.

What I see after 3 years — observations by industry

Anonymous ranges from different segments where I have run PMax since March 2022:

• Natural cosmetics (mid-segment) — in my practice PMax added 15–22% ROAS over Standard Shopping (in a hybrid setup). Key: feed with GTIN, custom labels per margin, brand exclusions from day one.
• Diet catering (subscription e-com) — from my experience PMax delivered mediocre results until 2024, and only got better after I introduced value-based conversions per subscription cycle (1 month vs 3 months). Before that, the system was catching cheap "first-order testers".
• Mid-segment fashion — very volatile results, heavily dependent on feed quality and season. PMax without return exclusions (custom_label "high_return") tended to eat margin. After feed cleanup — a solid 8–15% over Standard Shopping.
• B2B SaaS (cybersecurity / data) — PMax without OCI did poorly. After deploying Closed-Won import — the opposite, the system started acquiring leads twice as good at a similar cost. Key: 90+ days of patience to learn.
• B2B research services — low volume + long cycle = PMax basically did not work. Better results from Search exact match + LinkedIn.
• Low-margin accessories store — PMax caught the cheapest SKUs, ROAS looked OK, but POAS (profit) was negative. Fix: tROAS adjusted to margin + custom labels + low-margin exclusions.

The pattern I see: PMax wins when it has feed quality + signals + patience. It always loses when one of those three is missing.

Quick wins for PMax

If you already have PMax and the results are underwhelming — five things that improve performance in 80% of cases:

1. Turn on Account-level Brand Exclusions. Add your brand and 3–5 competitors. You stop branded-search cannibalization from day one.
2. Value-based conversions. Dynamic transaction value in e-com, point scale per CRM stage in B2B. Without it, PMax catches cheap queries.
3. Search themes. Add 10–20 thematic phrases (intent-based, not keyword) to help the system understand your customers.
4. Customer Match audience signals. Upload customer lists (2,000+ records), lookalikes, in-market segments. The algorithm gets additional, valuable context to learn your real customers.
5. tROAS instead of Maximize Conversions. After 50+ conversions per month, switch to tROAS — it teaches the system to optimize for value, not volume. With lower volume, stick with Maximize Conversion Value (with an optional ROAS ceiling).

How to check whether PMax is working for you

Level 1 — does it work technically

• 30+ conversions / mo. (Smart Bidding threshold)
• Conversions with values (the "Conv. value" column is not empty)
• Brand Exclusions set (account level + optional campaign level)
• Search themes filled in (10+ themes)

Level 2 — is there context

• Customer Match with a list of 2,000+ customers
• Audience signals added to asset groups
• Separate asset groups per category / segment (not one for everything)
• tROAS as the bidding strategy (not Maximize Conversions)

Level 3 — are you making money

• PMax ROAS is higher than Standard Shopping run alongside (if both exist)
• POAS (Profit on Ad Spend) is positive — if you track margin per product
• Branded search in Standard Shopping still generates conversions (i.e. PMax did not eat it)
• The Insights tab shows growing search themes — the system is learning

The most common diagnostic outcome on accounts I take over: Level 1 OK, but Level 2 is weak (no Customer Match, no audience signals, one asset group). That is the scenario I see most often — and it is the easiest to fix. From my practice, fixing this typically delivers +15–30% ROAS in 60 days, without raising the budget.

About the sources: PMax mechanics (asset groups, search themes, audience signals, brand exclusions) per Google Ads — Performance Max. Hybrid strategy with Standard Shopping — smec — PMax + Standard Shopping. AI Max — Google Blog — AI Max for Search. Industry patterns and ROAS ranges — from 3 years of running PMax at JP Ads (since March 2022). Numbers from my own practice are flagged "from my experience".