Google Demand Gen vs Performance Max for B2B Pipeline Campaigns

Start with the buyer, not the tool. Gartner's 2025 research puts 80% of the B2B buying journey happening with no direct contact with a vendor at all. Buyers are deep into evaluation long before they raise their hand.

Forrester's 2024 data backs this up from another angle: 92% of B2B buyers already have a vendor in mind before they start a formal evaluation, and 41% walk in with just one name on the list. A campaign built to drive conversions is often showing up after the shortlist is already locked. Most B2B ad spend is aimed at a decision that's already been made.

Add in scale. Gartner puts the average B2B purchase at 27 touchpoints, with 6 to 10 stakeholders involved (more on bigger deals). Each person can enter the funnel at a different point, through a different channel, at a different time. No single campaign type can own that journey. It's spread too thin, across too many people, over too long a window.

Here's where that hits Performance Max directly: feed PMax a conversion goal based on a raw form fill, and it trains itself on the last click of a long, messy path. It gets very good at generating more of that last click, whether or not the click means anything.

Here's where it hits Demand Gen: a campaign type built around awareness and engagement has to eventually justify itself in pipeline terms, or it loses its budget. Most B2B marketing teams don't have the measurement setup to make that case cleanly.

Gartner also finds 61% of B2B buyers actually prefer to avoid talking to a sales rep for as long as possible. Better tracking cannot patch that gap. This is just the shape of how B2B buyers behave now, and campaign strategy has to be built around it, not against it. The teams still picking a campaign type first and a buyer behavior model second have the order backwards.

What Performance Max actually requires to work for B2B lead generation

PMax runs on one core mechanic: give it a conversion goal, and it optimizes bidding and placement across every corner of Google's inventory (Search, Shopping, YouTube, Display, Discover, Gmail, Maps) to hit that goal as often as possible. The quality of that goal decides everything else.

Feed it a weak signal, like an unqualified form fill, and it will chase that signal with total commitment. The Starr Conspiracy has a name for this: "junk-lead gravity." PMax will happily fill a pipeline with students, competitors, people in the wrong country, and job titles that have nothing to do with the actual buyer, because those are the easiest conversions to produce.

There's a second trap: branded search cannibalization. Branded queries convert easily on their own, so PMax will claim credit for them and quietly eat budget that would have converted anyway, with no separate Search campaign around to protect that traffic.

The fix is offline conversion tracking, and it's a mechanical fix, not a conceptual one:

• Capture the GCLID with a hidden field on lead forms.
• When a lead becomes an SQL or opportunity in the CRM, mark that event and export it.
• Import that CRM event back into Google Ads as the primary conversion action.

That last step is what actually teaches PMax which leads turn into pipeline, not just which forms get submitted. Sales-accepted-lead and opportunity-stage events from the CRM need to be the primary conversion actions. Raw form fills should sit as secondary signals, or get excluded from optimization. Skip this step and the rest of the setup doesn't matter. PMax will keep optimizing toward the wrong finish line no matter how good the creative or targeting looks.

Done right, this works. Growleads.io reported in 2025 that a properly structured PMax campaign optimized around sales-qualified leads can cut cost per lead by 34%.

But there's a volume floor, and this is the part most teams skip past. The Starr Conspiracy's Campaign-Fit Selection Matrix puts it around 50 qualified conversion events a month. Below that line, the algorithm doesn't have enough data to learn the right pattern, so it starts grabbing whatever signal is lying around, good or bad. Running PMax under that floor is a fundamentally different strategy. It's a different, worse strategy wearing the same campaign name, and it should be treated that way when someone decides whether to use it at all.

Value-based bidding paired with offline conversion tracking is the current best-practice setup for B2B PMax in 2025.

PMax doesn't natively show performance by placement, search term, or channel, which is a limitation worth knowing upfront. Teams work around this with supplementary analytics integrations and account-level negative keyword lists to get better visibility into what's actually working. None of that is optional if the goal is knowing what's actually working, rather than just hoping.

What Demand Gen is actually built for in a B2B context, and where it breaks down

Demand Gen flips the control model. Pick the audiences and placements, and Google serves visual and video creative across YouTube, Shorts, Discover, and Gmail, optimizing for engagement and consideration, not conversions.

That audience control is the whole point. Target by job title, company size, in-market segment, age band. Build lookalike audiences off existing customer lists. Turn off optimized targeting entirely, and lock the audience as a fixed input instead of a signal the algorithm keeps nudging around. As of March 2025, opting out of specific channels is possible too, meaning YouTube-only Demand Gen campaigns now exist. Existing customers and current pipeline can get excluded outright, to keep prospecting lists clean.

Format matters more than most teams realize. Google expanded Shorts inventory inside Demand Gen through 2024 and 2025, and vertical video is now the default placement for cold B2B demand. Teams still shooting only standard 16:9 video are locking themselves out of the placements getting the most efficient reach, full stop.

One structural change: Video Action Campaigns got folded into Demand Gen in 2025. Any team still running YouTube conversion campaigns under the old name is, whether anyone updated the strategy or not, now running a Demand Gen campaign.

Measurement is the catch, and it's the reason Demand Gen gets underfunded even when it's working. Its conversions are mostly view-through and assisted, and neither one maps cleanly to pipeline without offline conversion imports and multi-touch attribution layered on top. Reporting gives decent visibility into YouTube placements but stays fairly shallow on Gmail and Discover.

On budget: Google recommends at least $100 a day per campaign, or 20 times the expected average conversion value, whichever is higher. For B2B teams working with tight budgets, that floor matters when deciding how many campaigns to run at once.

Demand Gen earns its keep with buyers who haven't started searching yet: the latent or early-active demand crowd. That's the same 80% of the buying journey Gartner found happening before anyone contacts a vendor. Treat Demand Gen as a bottom-of-funnel lead source and it will look like it's failing. It was never built to close deals.

How the Campaign-Fit Selection Matrix routes the decision before spend is committed

The Starr Conspiracy's Campaign-Fit Selection Matrix routes any given offer along two axes before a dollar goes out the door.

Demand state: where is the audience sitting, from latent demand all the way to actively in-market?

Conversion tightness: how close is the conversion event to actual qualified pipeline, from a soft engagement (a video view) to a CRM-imported opportunity?

From there, four routes:

• Route A, Demand Gen primary: latent demand, softer conversion signal, strong visual creative ready to go. The audience isn't searching yet.
• Route B, PMax primary: active demand, high-volume tight conversion signal (roughly 50+ qualified events a month), offline import feed already built and working.
• Route C, Search-led with Demand Gen support: active demand, but low conversion volume and branded equity that needs protecting from cannibalization.
• Route D, Hold: not enough signal, creative, or ICP clarity to justify either campaign type yet. Spending now just produces bad data that someone has to explain later.

A concrete example from the matrix: a demo-request conversion firing fewer than 50 times a month, tied to a narrow ICP. Put that into PMax, and it starves, then starts chasing whatever low-intent volume it can find. Route it to Demand Gen or Search instead.

The whole point of the matrix is forcing this routing decision up front, on paper, before a quarter of spend produces data nobody can explain in a QBR. That means checking, before anything gets spent, monthly conversion volume by funnel stage, whether CRM import is actually built and working, a clearly defined ICP, and an honest inventory of what creative formats actually exist. Skip that checklist and the campaign type stops being a strategic choice. It becomes a guess with a budget attached.

How the full-funnel architecture connects Demand Gen, AI Max for Search, and Performance Max

Google bundled Demand Gen, AI Max for Search, and PMax into what it calls the "Power Pack" for 2025, offering two configurations depending on what an advertiser is trying to achieve. Treat that as one connected flow, not three separate campaigns running in parallel and hoping for the best.

Top of funnel, Demand Gen: reaches cold audiences with video and visual creative across YouTube, Shorts, Discover, and Gmail. Introduces the brand, builds consideration among people who aren't searching yet.

Mid-funnel, AI Max for Search: picks up intent from people already exposed to Demand Gen creative who've started actively researching. AI-driven broad match and query matching catch relevant searches from an audience that's already warmed up.

Bottom of funnel, Performance Max: converts that warm audience across all of Google's inventory, optimizing toward the most valuable conversions, guided by offline pipeline data pulled from the CRM.

The signal moves between layers. Someone who watches a Demand Gen video becomes part of the retargeting pool for both AI Max and PMax. The layers stack on top of each other instead of duplicating the same work.

That stacking creates a real risk, though: Demand Gen and PMax can end up competing for the same inventory and the same people if nobody sets up exclusions. So:

• Exclude Demand Gen audiences from PMax asset groups once they've moved into active consideration.
• Use audience exclusions in Demand Gen to keep current customers and existing pipeline out of prospecting.
• Run a dedicated branded Search campaign on its own, so PMax isn't quietly claiming credit for people who were already searching the company name.

Budget allocation should follow the constraint, not a fixed ratio. A useful framing here: an account with weak brand awareness needs more weight on Demand Gen. An account with solid brand recall but slow pipeline velocity needs more PMax, backed by a better conversion signal.

None of this holds together without CRM data flowing into every layer. Cut that feed, and each campaign type goes back to optimizing in its own silo, and the whole architecture falls apart. It's a plain statement of fact. It's the single point of failure the entire funnel design depends on.

Attribution across long B2B sales cycles and why platform-reported conversions mislead both decisions

Google's default data-driven attribution model is a real improvement over last-click, but it can only split credit among the Google Ads interactions it can actually see. LinkedIn touches, organic search, email nurture, a sales call: all invisible to it.

A journey spanning an average of 27 touchpoints and multiple stakeholders means platform-native attribution can only ever see a fraction of the full path, not by accident, but by design. The model only ever looks at a slice of the full path, so it's not wrong exactly. It just can't see most of the room.

The practical result: a Demand Gen campaign that shaped consideration but never got the final click looks like it did nothing. A PMax campaign that happened to catch the branded last click looks like it did everything. Neither read is accurate, and treating either one as gospel is how budgets get moved in the wrong direction, toward the campaign that happened to be standing near the finish line rather than the one that did the actual work.

There's also the dark funnel to account for: buyers who see an ad, don't click, and search the brand name weeks later. Platform data misses that. A simple "how did you hear about us?" field on the demo form often catches influence the ad platform never will.

The structural fix is connecting the CRM directly to the ad platform, whether that's HubSpot, Salesforce, Attio, or Pipedrive. That link shows which companies are actually engaging with ads and ties that engagement to real pipeline stages, not just conversion events sitting inside Google Ads.

At minimum, a defensible reporting setup tracks four things weekly: spend by platform, leads by platform, opportunities by platform, pipeline by platform. That's the floor, not the finished product.

Timing matters too. A Demand Gen campaign running in Q1 might not show pipeline contribution until Q3. Measurement cadence has to match the actual sales cycle length, not just whatever converted this month.

Data-driven attribution inside Google Ads should be treated as a directional signal, not a source of truth. The number the business is actually accountable to is generally best anchored in CRM data rather than platform-reported conversions.

The pipeline metrics that make the Demand Gen vs. PMax mix defensible to revenue leadership

Pipeline marketing means marketing gets judged on pipeline creation and velocity, not lead volume or MQL counts. Every campaign decision gets measured against downstream revenue, not what happened inside the ad platform.

A few benchmarks worth knowing, to calibrate what "working" actually looks like:

• Visitor-to-lead: 1 to 3%
• MQL-to-SQL: 15 to 21%, and 2025 benchmarks flag this conversion stage as a common bottleneck in B2B SaaS
• Opportunity-to-close: 20 to 30%

Pipeline coverage is the bigger health signal. The Ebsta x Pavilion 2025 GTM Benchmarks put average B2B win rates at 19%, which means many teams now need closer to 5x pipeline coverage just to hit their number. Salesforce Research found organizations with healthy pipeline coverage are 2.3 times more likely to consistently hit revenue targets.

Demand Gen and PMax need to show different things to earn their budget:

Demand Gen should show influenced pipeline by cohort, view-through conversion rate tied to SQL creation, and time-lagged attribution, meaning deals where a Demand Gen touch showed up before intent signals did.

PMax should show SQL rate on the leads it generates, cost per SQL (not cost per lead), offline conversion import volume and match rate, and actual pipeline value tied back to PMax-originated leads.

The MQL-to-SQL bottleneck is usually the first place a campaign mismatch shows up. High MQL volume paired with a weak SQL conversion rate is almost always a sign PMax got trained on the wrong conversion event somewhere upstream. It is rarely a sign the audience is wrong, and teams that respond by rewriting targeting instead of fixing the conversion event tend to make the problem worse.

The budget split between the two should get reviewed quarterly against pipeline coverage, not set once and forgotten. If coverage starts slipping, the fix is almost always adjusting targeting or creative rather than spending more." It's finding what's actually blocking the existing spend from working, usually signal quality, or a mismatch between funnel stage and campaign type.

Where full delegation of campaign execution changes what's possible across this decision

Every piece of this, offline conversion imports, audience exclusions, quarterly matrix reviews, CRM-to-ad-platform syncing, is a standing job, not a one-time setup. Someone has to keep the GCLID capture working, keep the CRM export clean, keep exclusions current as pipeline stages shift, and keep rereading the matrix as demand states change.

Splitting that work across a media buyer who never touches the CRM and a separate specialist who never touches the ad account is the most common way this whole system quietly breaks. Handing the full loop to a single owner, or a single team, is what actually makes the Demand Gen and PMax mix hold up over time. The matrix, the offline conversion pipeline, and the exclusion logic all depend on the same person seeing the full path from ad click to closed deal. Split that ownership, and the signal degrades exactly where it matters most: the handoff between what the algorithm sees and what the business actually gets paid for.