Microsoft's PMax Experiments Beta: The Skinny

Microsoft’s PMax Uplift and PMax Upgrade both promise the same thing: a way to see what Performance Max would do to your Microsoft Ads account before you commit to it. That's a real improvement over launching blind and hoping the numbers move the right way, but it’s not a catch-all.

My caveat: these experiments will tell you what happened inside a controlled test window under conditions the model optimized for, and that’s not the same as what will happen once the campaign goes live and gets influenced by budget, optimizations, and market shifts. Give the campaign the usual guardrails (I’ll explain below), treat the output as a strong directional read, not a guarantee, and the Beta becomes genuinely useful. Treat it as a verdict, and you'll be surprised later.

TL;DR

• Microsoft launched a Beta for two PMax experiment types: Uplift measures incremental lift from adding PMax alongside existing campaigns, Upgrade compares an existing non-PMax campaign against a PMax version.
• Trust the results directionally, not as a promise. Live performance shifts once the test window closes.
• Your existing Search maturity predicts your results before you launch. Mature Search setups tend to show smaller lift, often concentrated in MSAN rather than Search.
• Where the incremental value comes from matters: Search carries higher CPCs and stronger conversion rates, MSAN is cheaper but converts lower, and that mix changes what a "win" actually means.
• The Beta controls for traffic allocation and auction conditions, but it won't set your business guardrails. You still define success metrics, conversion actions, bidding, test budget, and exclusions.

Uplift vs. Upgrade: Two Different Questions, Two Different Tests

These aren’t two names for the same experiment; they answer different questions.

PMax Uplift measures the incremental impact of adding a new PMax campaign on top of your existing campaigns. The question it settles is one of incrementality: will PMax generate genuinely new conversions or revenue, or will it just shift conversions that your current campaigns would have captured anyway?

PMax Upgrade runs a side-by-side comparison between an existing non-PMax campaign and a PMax version of it. The question there is simpler: would switching improve performance over what you're already running? Use Upgrade when you're deciding whether to replace something. Use Uplift when you're deciding whether to add something. Confusing the two is how you end up "proving" PMax works when all you measured was reshuffled credit.

Why Your Existing Search Maturity Predicts Your Results Before You Even Launch

Your current account is the baseline for the experiment, so the sophistication of what you already run largely decides how much lift PMax appears to deliver. The more mature and optimized the account, the smaller the incremental gains tend to be. That's not a knock on PMax. It's math. PMax can only find value that your existing setup is leaving on the table.

If your Search strategy already includes Broad Match, AI Max, and comprehensive keyword coverage, you have less unclaimed Search territory for PMax to discover. In that case, expect smaller incremental lift, with a larger share of it coming from the Microsoft Audience Network rather than Search itself. If your Search strategy is less developed and doesn't use Broad Match or AI Max, PMax may surface Search opportunities you weren't capturing, and the incremental lift can be meaningfully larger.

The practical move is to benchmark your Search sophistication honestly before you read a single result. An advertiser with a mature account and a modest lift number is looking at diminishing returns on top of an already strong foundation, and that reads very differently than the same number on an underbuilt account.

Search vs. MSAN: The Incremental Traffic Mix That Determines the Verdict

A lift number in aggregate hides the thing that actually matters: where that lift came from. Search and MSAN behave differently, and the mix changes what the win is worth to your business.

Search is intent-driven. It carries higher CPCs but stronger conversion rates. MSAN generally delivers lower-cost traffic at lower conversion rates. A PMax campaign that shows incremental lift concentrated in MSAN is a different proposition than one driving incremental Search conversions, even if the top-line lift number looks similar. Cheaper traffic that converts poorly can inflate volume metrics while doing little for your actual return.

Before you accept a verdict, break the incremental value down by source. Ask whether the lift is coming from higher-intent Search demand or from cheaper network reach, and decide whether that mix matches what you're trying to grow. A test can be "positive" and still point you toward traffic your business doesn't want more of.

The Guardrails Microsoft's Beta Won't Set for You

The Beta is built to control for the variables it can see: randomized traffic allocation, consistent auction conditions, a defined test window. That's the part designed to maximize the reliability of the comparison. The part that stays your job is everything the model can't know about your business.

The model optimizes toward the inputs and signals available inside the experiment. It does not understand your profitability requirements, business thresholds, or strategic priorities. Those live outside the experiment framework and require your oversight regardless of how clean the test setup looks.

Here are the guardrails I'd apply to any PMax experiment on Microsoft, the same ones we use for any campaign test:

• Define clear success metrics up front, and use the same conversion actions and bidding strategy across both campaigns so you're comparing like for like.
• Hold a fixed test budget to avoid introducing variability that muddies attribution.
• Allocate enough budget and run the test long enough to exit the learning phase. Microsoft typically carries lower volume than Google, so reaching statistical significance takes longer, and drawing conclusions before you have meaningful data is the most common way these tests go wrong.
• Minimize cannibalization with negatives or exclusions where appropriate, so the results reflect PMax's true incremental impact rather than overlap between campaigns.
• Keep your existing campaigns as stable as possible during the test, so any performance difference can be attributed to PMax rather than to other account changes.

For context, Google offers similar PMax experiments to measure lift and compare against eligible campaign types. The difference is philosophical: Google has broader reach and a more mature AI model but generally offers less transparency and control, while Microsoft has smaller reach and leans harder into visibility and campaign controls. These Uplift and Upgrade experiments fit that Microsoft pattern. They give you more to see and more to set, which is an advantage only if you actually set it.

The Practical Read: Test PMax Without Betting the Account on It

The value of these experiments is in what they let you avoid: committing to PMax on Microsoft without evidence or dismissing it without a fair look. Used well, Uplift and Upgrade give you a defensible read on whether PMax adds incremental value or would outperform what you're running.

The advertisers who get the most out of this are the ones with meaningful Search spend and enough volume to reach significance, who benchmark their current Search maturity first and read the results with the Search-versus-MSAN mix in mind. Low-volume accounts should proceed with more caution, because significance takes longer on Microsoft than on Google. And if your Search strategy is already highly mature, expect the lift to be marginal and possibly concentrated in MSAN, which makes the experiment one input rather than a standalone decision.

If you're weighing whether to test PMax Uplift or PMax Upgrade against your current Microsoft Ads setup, we can help you build the guardrails the Beta won't set for you. Book a strategy call with us.

FAQs

What’s the difference between PMax Uplift and PMax Upgrade in Microsoft Ads?

PMax Uplift measures the incremental conversions or revenue you gain by adding a new PMax campaign on top of your existing campaigns, answering whether PMax drives additive volume or just shifts existing conversions. PMax Upgrade runs a side-by-side comparison between an existing non-PMax campaign and a PMax version to see if switching would improve performance. Use Uplift when deciding whether to add PMax, and Upgrade when deciding whether to replace a campaign.

Can I trust the results of Microsoft's PMax experiments after I launch the campaign?

Trust them directionally rather than as a guarantee. The experiments run over a defined test period under controlled conditions, so they give you a strong read on potential impact, but live performance can shift once optimizations, budget changes, and market conditions start influencing results. Read the output as evidence for a decision, not as a fixed forecast of post-launch numbers.

Why might a mature Microsoft Ads account see smaller lift from PMax experiments?

Your existing account is the baseline for the experiment, so a well-developed Search strategy with Broad Match, AI Max, and full keyword coverage leaves less unclaimed territory for PMax to find. In those accounts, incremental lift tends to be smaller and often concentrated in the Microsoft Audience Network rather than Search. Less mature Search setups typically show larger lift because PMax can surface Search opportunities they weren't already capturing.