Affect Performance Team

|News|Nov 10, 2026

Microsoft Advertising Expands Uplift Experiments for Performance Max

In August 2026, Microsoft Advertising expanded access to Uplift Experiments, a tool for experimental measurement of Performance Max effectiveness. It shows how many additional conversions and how much incremental conversion value PMax delivers alongside Search, Shopping, and Audience campaigns that are already running.

Until now, advertisers could evaluate conversions, spend, and ROAS for individual PMax campaigns, but they could not always tell whether automation actually increased overall account results or simply redistributed conversions from other campaigns.

Microsoft began rolling out experiments for Performance Max in June 2026, and in August it broadened availability and added guidance on how to use the results.

What Is Microsoft Performance Max

Performance Max is an automated campaign type in Microsoft Advertising. The advertiser provides a goal, budget, creative assets, product data, and other signals, while the algorithms select audiences, manage bids, and allocate budget across available placements.

PMax can run alongside Search and Shopping. That expands reach but makes incrementality harder to analyze. Strong PMax performance does not necessarily mean more total sales, because some conversions might have happened through other campaigns anyway.

Why Standard Reporting Is Not Enough

Consider an online store that generated 1,000 conversions per month before launching PMax. After the launch, Search and Shopping delivered 850 conversions, and PMax reported 250. The account received 1,100 conversions in total.

Conversions before and after the PMax launch

Measure

Conversions

Total before PMax launch (per month)

1,000

Search and Shopping after launch

850

PMax after launch

250

Total after launch

1,100

Although 250 conversions are attributed to PMax, total results grew by only 100. This may indicate that some demand shifted between campaigns. However, a simple month-over-month comparison does not prove a causal effect, because seasonality, pricing, demand changes, and competition all influence results.

Measuring the incremental effect therefore requires an experimental control group.

What Uplift Experiments Are

Uplift Experiments compares results for advertising setups with and without Performance Max. In the control condition, existing campaigns such as Search, Shopping, and Audience continue to run. In the test condition, PMax is added to comparable campaigns.

The tool estimates how much conversions and conversion value change when an automated campaign is added. It matters not only to measure the difference but also to assess how reliable it is and whether it is economically meaningful.

How Uplift Differs from a Standard A/B Test

A standard A/B test might compare two ads, landing pages, or bidding settings. Uplift asks a broader question: does adding PMax improve the overall results of an active advertising account?

For example, a headline comparison answers which copy performs better. An Uplift Experiment helps decide whether to allocate budget to PMax on top of existing Search and Shopping campaigns.

What Changed in Microsoft Advertising in 2026

Until mid-year, Microsoft's experimentation tools focused mainly on search campaigns. In June, Microsoft introduced two new Performance Max testing scenarios: Uplift Experiments and Upgrade Experiments. Existing search experiments were renamed Search Optimization Experiments.

Uplift Experiments measure the additional result from running PMax together with other campaigns. Upgrade Experiments let advertisers compare an existing campaign with a version of it after migration to PMax.

In its August product roundup, Microsoft again highlighted Uplift Experiments and the broader ability to measure the incremental value of automation. Some experimentation features, including Upgrade Experiments, continued to roll out on a limited basis.

Uplift Experiments vs. Upgrade Experiments: What Is the Difference

With Uplift testing, the question is whether PMax delivers additional conversions on top of existing campaigns. The result informs the decision to add or scale PMax.

An Upgrade Experiment supports a different decision: whether to replace an existing campaign type with PMax. For example, an advertiser might consider migrating some Shopping campaigns rather than running both campaign types in parallel.

A positive Uplift result does not automatically mean Search and Shopping should be turned off. It reflects the effect of adding PMax in the specific configuration that was tested.

How to Launch an Uplift Experiment in Microsoft Ads

Experimentation tools are located under Campaigns → Experiments. Whether you can launch a PMax Uplift test depends on feature availability for your ad account.

Before the test, define the business outcome you will measure: number of conversions, conversion value, or a suitable qualified-lead metric. For e-commerce, revenue matters, not just order count. For B2B, CRM conversions are useful if there are enough of them for analysis.

Next, check your Universal Event Tracking (UET) setup or supported offline conversion import. The control and test conditions should be comparable in settings and constraints that are unrelated to the change being studied.

The specific group allocation options and available settings depend on the interface and experiment type.

How Much Data You Need

Microsoft's guidance suggests a benchmark of at least 30 conversions in the last 30 days to prepare for tests with automated campaigns. This does not guarantee a statistically reliable result: detecting a small lift may require significantly more events.

For the final analysis, Microsoft suggests a period of roughly 4 to 12 weeks, adjusted for traffic volume and conversion reporting delays. A short test risks missing the bid learning period, weekly seasonality, and late-arriving sales.

During the experiment, avoid major changes to budget, bidding strategy, and target CPA or ROAS unless they are part of the experiment design.

How to Interpret the Results

Suppose that in a properly structured hypothetical experiment, the control group generated 500 conversions and $50,000 in conversion value, while the test group generated 550 conversions and $57,000.

Hypothetical experiment results

Group

Conversions

Conversion value

Control

500

$50,000

Test

550

$57,000

Difference

10%

14%

That is a 10% difference in conversions and a 14% difference in conversion value. These figures alone are not enough for an economic decision: you also need to compare spend, changes in CPA and ROAS, and the statistical uncertainty of the result.

If the additional value comes at a disproportionate increase in spend, scaling PMax may not be profitable. If the effect is small and statistically inconclusive, it cannot be treated as a proven advantage of automation.

Mistakes That Distort Experiment Results

The first problem is changing the budget or the target CPA or ROAS in only one experiment group. The effect of PMax then gets mixed with the effect of a new bidding configuration.

The second problem is a poorly constructed control group. The control and test conditions must be comparable so the result can be interpreted.

The third problem is inconsistent conversion definitions. Without additional normalization, you cannot compare a test measured on form submissions with a control measured on qualified leads.

The fourth problem is insufficient duration. Late conversions, seasonality, and low data volume can create a false impression of performance.

What Is Known About PMax Results

Microsoft has previously reported aggregated data on additional conversions from Performance Max, including a benchmark of about 8%. However, this is not a forecast for any specific Uplift Experiment and not a platform promise for every advertiser.

It is important to separate the performance of PMax itself from the value of the experimentation tool. Uplift Experiments does not change the advertising algorithm. It helps measure whether turning on PMax actually increases account results.

As of October 2026, there is little large-scale independent research on the effectiveness of Uplift Experiments in Microsoft Advertising specifically. Conclusions should be based on your own experiments, not only on aggregated platform results.

How to Use Results in Budget Allocation

If an experiment shows a sustained incremental effect at an economically acceptable cost, that supports scaling PMax. If the lift is minimal and a significant share of conversions is simply redistributed between campaigns, scaling may not be justified.

When the result is statistically uncertain, it is better to extend the test or revisit the methodology than to declare the automated campaign a success or a failure.

The main purpose of Uplift Experiments is to move from evaluating the ROAS of an individual PMax campaign to measuring its incremental value for the advertising account as a whole.

Trust Us Transform Your Performance Marketing.

Before you put budget behind PMax, let us design the experiment, set up conversion tracking, and read the results against your CPA and ROAS targets in a Paid Search audit.

Audit My Microsoft Ads Account