What Performance Max Actually Needs to Work in B2B

TL;DR

1. Performance Max is a cross-channel campaign running across all six Google surfaces simultaneously. Its structural advantage is reach. On high-intent search queries, standard Search consistently outperforms it on conversion rate - that isn't a flaw, it's how two structurally different tools work. Evaluating PMax against Search benchmarks produces a confusing picture.

2. PMax learns from whatever conversion signal it's been given. When that signal is connected to real business outcomes - qualified leads, pipeline milestones, closed revenue - the algorithm finds patterns associated with those outcomes. When it isn't, it finds the cheapest available conversion. Neither outcome is a malfunction.

3. Three conditions determine whether PMax can learn in a B2B account: conversion signal quality, monthly conversion volume, and how branded search is handled. These aren't advanced optimisation choices - they're the baseline for the algorithm to function as designed.

4. For B2B accounts with sales cycles longer than 90 days, there's a structural timing issue: the identifiers Google uses to connect ad clicks to conversions expire after 90 days. Without intermediate pipeline signals, the algorithm is optimising without receiving the outcomes it's supposed to learn from.

5. The useful question isn't "is PMax working?" It's "have I given PMax what it needs to work?" One question looks at the campaign. The other looks at the account. Only the second one leads somewhere actionable.

Why B2B Teams End Up Unsure Whether to Trust Their PMax Numbers

When a Performance Max campaign shows strong numbers but the pipeline doesn't follow, the instinct is to look at the campaign. The creative. The bids. The audience signals. That's a reasonable instinct - and often, it's looking at the wrong layer.

The situation tends to look like one of two things. Either PMax is reporting conversions that sales doesn't recognise - leads the team can't place in the CRM, contacts no one remembers speaking with. Or PMax ROAS looks healthy on paper, but revenue hasn't moved in a way that corresponds to it. In both cases, the search for a fix starts inside the campaign. Creative gets refreshed. Bid targets shift. Audiences are added or removed. The numbers might move. The uncertainty doesn't.

Picture a Tuesday pipeline review. The dashboard shows 34 conversions for the month. The sales team is going through the CRM, marking which ones they remember. "Where are these people actually coming from?" Not an accusation - just a real question. The campaign is running. Something is being measured. What it means for the business is still unclear.

What keeps this from resolving is that campaign-level changes can't fix an account-level condition. Whether PMax is learning from the right signal, receiving enough signal to find patterns, or even connecting ad clicks to the right business outcomes - none of these live inside the campaign settings. They live in how the account is configured before the campaign runs.

This article is about that other layer. Not whether PMax is working, but what it needs to work - and how to tell whether those conditions are in place.

Performance Max Is a Reach Campaign - That Changes How You Evaluate It

Performance Max is not a more automated version of Search. It is a cross-channel campaign that reaches audiences across all six Google surfaces simultaneously - Search, Display, YouTube, Gmail, Discover, and Maps - from a single budget and a single set of creative assets. No other Google campaign type does this.

Click to expand

Budget allocation across those channels is controlled by the algorithm. You set the objective and provide the signals. Where the spend actually goes is determined by where the algorithm expects to find the most efficient path to that objective. This is structurally different from managing separate campaigns where allocation is explicit and visible.

The consequence is that evaluating PMax against Search benchmarks - conversion rate, cost per lead, impression share on high-intent queries - will almost always produce a confusing picture. Adalysis research across more than 3,300 non-retail PMax campaigns found that when both PMax and a standard Search campaign were eligible for the same search query, standard Search produced a higher conversion rate on 84% of those overlapping queries. That isn't a finding against PMax. It's a description of what PMax is for. Its advantage is the inventory Search cannot reach - YouTube viewers, Gmail users, Discover feed browsers - audiences who are not actively searching for what you sell. Holding PMax accountable to Search-style efficiency on known high-intent queries is asking the wrong question of it.

Google has also substantially improved PMax's reporting transparency over the past year. The Channel Performance Report, released in November 2024, now shows spend and conversion breakdowns by channel. The Search Terms Report, added in March 2025, shows the top search categories driving activity with a column distinguishing PMax from Search impression share. Campaign-level negative keywords, rolled out in December 2024 with the limit raised to 10,000 in March 2025, gave advertisers meaningful control over Search and Shopping inventory that didn't previously exist. The campaign is more visible and more controllable than it was two years ago - which makes the signal quality question more important to get right, not less.

One genuine structural advantage worth noting: PMax is the only Google Ads campaign type with a New Customer Acquisition goal - the only campaign type that lets advertisers explicitly bid for first-time customers with a value premium. That capability doesn't exist in standard Search or Display.

Performance Max Learns From What You Give It. In B2B, the Default Signal Is Often Incomplete.

Performance Max optimises toward the conversion signal it's been given. The algorithm doesn't evaluate lead quality - it can't. It evaluates signal frequency and patterns. What you give it to learn from determines what it learns.

In many B2B accounts, the primary conversion event being tracked is a form submission. A form fill tells the algorithm something happened - not what happened next. Without explicit value differentiation, a form fill from a prospect who closed six months later looks identical to one from a competitor researching your pricing. Given a form-fill signal, the algorithm finds the most efficient path to form fills. That tends to run through branded queries, retargeting audiences, and lower-intent display inventory - all of which convert at a lower cost than cold search. The conversions are real. The question is whether they represent pipeline.

This pattern isn't unique to PMax - the form-fill quality problem exists across all Google Ads campaign types. What's specific to PMax is that its cross-channel reach gives it more surfaces to find cheap conversions on. Display and YouTube reach audiences with structurally lower purchase intent than search, so the drift from pipeline to volume is faster and less visible than it would be in a Search-only setup.

Click to expand

When CRM data is connected - qualified leads, pipeline milestones, opportunities - the algorithm receives a different question. It's no longer looking for the cheapest form fill. It's looking for patterns associated with contacts that progressed. The inventory mix shifts. Raw submission volume may decrease. Quality increases. Google's own lead generation best practices documentation explicitly recommends optimising for "qualified or converted leads that are as close to the final sale as possible" - not raw form submission counts. The algorithm is designed to respond to that signal when it's given one.

How Long Sales Cycles Affect What Performance Max Can Learn

Google's click identifiers - the codes that connect an ad click to a future conversion - expire after 90 days. For B2B accounts where deals close in months, not weeks, this creates a timing gap between when conversions happen and when the algorithm can learn from them.

Every ad click Google serves gets assigned a unique identifier called a GCLID. When an offline conversion is later imported - a qualified lead, a closed deal - it's matched back to the original click via that identifier. The GCLID is retained for 90 days. Any conversion uploaded after that window cannot be attributed to the originating click.

Suppose your average deal takes five months from first click to close. You've set up offline conversion import correctly. The CRM shows won business. But at month four, the algorithm still hasn't received any closed-deal signal - because the GCLID from that initial click expired at the 90-day mark. PMax continues optimising as though those outcomes don't exist, because from its perspective, they haven't arrived yet. The issue isn't a broken integration. It's a fundamental mismatch between the identifier's lifespan and the deal cycle's length.

Click to expand

The solution isn't a campaign-level fix. It's to feed the algorithm signals it can actually receive. Intermediate pipeline events - demo completed, proposal sent, SQL status reached - occur earlier in the cycle, within the 90-day window, and give the algorithm real information about what's progressing. Assigning proxy values to these milestones (a booked demo carries more downstream value than a contact form, so pricing that difference into the signals tells the algorithm what to prioritise) enables value-based bidding even when the final sale takes months. Google's own documentation describes this approach specifically for B2B accounts with long deal cycles - it's not a workaround, it's the intended design for this configuration.

Two Other Factors That Shape What Performance Max Can Do in B2B

Two other conditions shape what the algorithm can work with, independent of signal design.

The first is conversion volume. Smart Bidding - the system PMax runs on - generally needs a meaningful base of conversion signals to learn reliably. Google's Smart Bidding documentation references around 30 conversions per month as the level at which the system begins to function effectively; for PMax's own learning cycle specifically, Google's documentation references up to around 50 conversion events to complete a full learning period. Below these thresholds, the algorithm is working from too few data points to find stable patterns. This isn't unique to PMax - it's the minimum requirement for any machine learning system to function as designed. B2B companies with niche audiences and long sales cycles frequently fall below this level, which is a reason to delay introducing PMax until Search campaigns are generating sufficient volume, not to abandon the channel type.

The second is how branded search interacts with the campaign. PMax competes in branded auctions even when dedicated brand campaigns exist. Whether this is a problem depends entirely on the competitive context, and the evidence reflects that. Haus.io's experiments on brand exclusion in PMax found the result genuinely splits: in roughly half their cases, excluding branded terms from PMax improved incremental performance because budget shifted to higher-quality non-brand inventory; in the other half, including brand terms was defensible because competitors were actively bidding on those terms and PMax's presence provided real protection. There's no universal prescription here - the right answer depends on understanding the specific competitive landscape. Brand Lists, available natively in PMax campaign settings, apply brand exclusions to Search and Shopping inventory only. They do not extend to YouTube, Display, Gmail, or Discover placements.

Performance Max and Search Have Different Jobs - Running Them That Way Changes the Outcome

Standard Search and Performance Max have different structural strengths. Search is efficient on known high-intent queries - the terms where someone is already looking for what you sell. PMax reaches audiences Search doesn't access. When their roles are clearly defined, they complement each other. When they're not, they compete - and that competition produces data that's hard to interpret.

Click to expand

Optmyzr's analysis of 503 accounts found keyword overlap between Search and PMax campaigns in 91.45% of accounts studied, with standard Search producing a higher conversion rate on the majority of those overlapping queries. This matters not because it disqualifies PMax, but because it clarifies where PMax's contribution actually lies. The overlap is structural - when both campaign types are eligible, Search tends to win on efficiency for the queries where intent is clear. PMax's contribution is the reach into channels and audiences where Search is simply not present.

The practical question to ask before evaluating PMax's contribution is: what is this campaign responsible for that standard Search cannot do? The answer should describe specific inventory - display audiences in a target industry vertical, YouTube viewers who fit the ICP profile, Discover-browsing prospects who aren't yet in search mode. Without a clear answer to that question, there's no basis for evaluating whether PMax is succeeding at its actual job. If PMax and Search are competing on the same queries without a defined role for each, the data produced by both becomes difficult to interpret independently. The structural fix is defining the division before running both, not revisiting it after performance looks confusing.

The Account-Level Questions That Campaign Metrics Can't Answer

Campaign metrics tell you how efficiently Performance Max is pursuing its objective. They don't tell you whether the objective is connected to what the business needs. Those questions live one layer above the campaign - in the account setup, the conversion design, and the signal architecture.

The useful diagnostic questions aren't about campaign settings. What conversion event is PMax optimising toward, and does it have a documented relationship to pipeline - not just a count of form submissions? Is conversion volume sufficient for the algorithm to produce reliable patterns, or is the campaign effectively in permanent learning mode? If the sales cycle exceeds 90 days, are intermediate milestones being fed back to the algorithm within the window where it can actually receive them? Is branded search included in PMax deliberately - because competitive pressure on brand terms makes it defensible - or by default, where it inflates reported performance without adding new demand? And what role, specifically, is PMax responsible for that Search cannot cover - and is its performance being measured against that role?

These aren't campaign optimisation questions. They're account and business questions. They can't be answered from inside the platform, and campaign-level adjustments won't surface them.

The person whose answers to those questions are clear and affirmative has a genuine basis for trusting what PMax reports. The person who isn't certain about one or more of them hasn't found a campaign problem. They've found where to look.

If the Questions Above Don't Have Clear Answers

That's not a campaign problem - it's a setup question. And it's exactly the kind of question an account audit is designed to surface.