Performance Max Hides Your Data. MMM Gets It Back.

Performance Max is Google's flagship campaign type, and by 2026, it handles the majority of Google Ads spend for most advertisers. The pitch is appealing: one campaign across Search, Shopping, YouTube, Display, Discover, Gmail, and Maps. Let Google's AI figure out where to allocate.

The problem? You cannot see where your money goes. PMax is the biggest black box in digital advertising, and Google designed it that way.

What PMax Hides From You

PMax does not report spend by network. You cannot see how much went to Search vs. YouTube vs. Display. You get "asset group" performance, which tells you which creative combinations are being used, but not where they are running.

You cannot see search term data for most queries. PMax shows you search categories (broad buckets), not actual search terms. If PMax is bidding on your brand terms, you would not know, and it almost certainly is.

You cannot control placements. If Google decides your budget is best spent on low-quality Display Network sites at $0.40 CPMs instead of YouTube pre-roll at $12 CPMs, it will do that. The CPA might look good because those Display clicks are cheap. But the incremental value of those impressions is often near zero.

The Brand Search Cannibalization Problem

This is the biggest issue with PMax, and it is well-documented: PMax cannibalizes your branded search. Someone types your brand name into Google. They would have clicked your organic result. But PMax bids on that query, shows a Shopping ad, and claims the conversion.

We have seen PMax campaigns where 30-60% of conversions come from branded queries. Those are not incremental sales. Those are sales that would have happened if PMax did not exist. But Google reports them as PMax conversions, making the campaign look incredibly efficient.

Some advertisers have tried brand exclusions in PMax (available since late 2023), but the implementation is imperfect. Even with exclusions, PMax still captures near-brand and misspelling queries.

How MMM Measures PMax Independently

Marketing mix modeling does not care about Google's attribution. It looks at your total PMax spend and your total outcome (revenue, leads, whatever you measure), then estimates the causal relationship.

The model controls for all other channels, seasonality, promotions, and organic trends. So if PMax is mostly cannibalizing brand search, MMM catches that because the incremental lift from PMax spend is small relative to what Google reports.

In practice, when we compare PMax ROAS from Google Ads (typically 8-15x) to MMM-measured ROAS (typically 2-5x), the gap is massive. That gap is mostly brand cannibalization and view-through inflation.

What to Do With PMax in Your Mix

1. Do not kill PMax. Even at the MMM-adjusted ROAS, PMax is often still profitable. It just is not as profitable as Google tells you. That distinction matters for budget allocation but not for the on/off decision.

2. Run branded search separately. Keep a standard branded search campaign outside PMax with brand exclusions in PMax. This gives you cleaner data on what PMax is actually doing for non-brand queries.

3. Use MMM to right-size PMax budget. If Google reports 12x ROAS but MMM shows 3.5x, you are probably over-investing in PMax relative to other channels. Shift budget to where your response curves show room for growth.

4. Monitor the PMax spend split indirectly. Track YouTube impressions, Display impressions, and Shopping clicks alongside PMax spend. If PMax spend goes up but Shopping and YouTube impressions stay flat, most of the increase is going to Search (likely brand).

The irony is that Google built one of the best open-source MMM tools (Meridian) while simultaneously making its own ad product harder to measure. Use the first tool to keep the second one honest. See how it works in our demo report.