In the evolving landscape of digital advertising, Google’s Demand Gen campaigns have emerged as a powerhouse for brands looking to capture attention across high-engagement surfaces like YouTube, Google Discover, and Gmail. Unlike traditional Search campaigns that capture existing intent, Demand Gen is designed to create it. However, this shift from “pull” to “push” marketing introduces a significant challenge for advertisers: the “attribution illusion.”
When an advertiser looks at their Google Ads dashboard and sees conversions attributed to a Demand Gen campaign, a nagging question often remains: Would these users have converted anyway? Because Demand Gen operates on visually-heavy platforms where users might see an ad but not immediately click, standard attribution models often struggle to distinguish between a user who was influenced by the creative and a user who was already on a path to purchase through organic search or direct traffic. To solve this, Google introduced asset uplift experiments in late 2025, providing a scientific framework to measure the true incremental impact of creative assets.
The Problem with Standard Attribution: Correlation vs. Causation
The “attribution illusion” occurs when marketers mistake correlation for causation. In a typical user journey, a consumer might see a high-quality video ad on YouTube via a Demand Gen campaign. They don’t click the ad immediately because they are mid-video. Later that evening, they remember the brand, search for it on Google, and complete a purchase. Under many attribution models, the Demand Gen campaign may receive partial or even full credit for that conversion.
While this looks good on a report, it doesn’t prove that the ad was the deciding factor. It is possible the user was already planning to buy. Without a control group, you are essentially guessing. Relying on these skewed metrics can lead to inefficient budget allocation, where funds are funneled into creative assets that look like they are performing but are actually just “snatching” credit from users who were already converted. This is where incrementality testing—and specifically asset uplift tests—becomes essential.
Establishing a Baseline with Incrementality
To truly understand the value of your creative, you must use the scientific method. This involves establishing a baseline by withholding your test assets from a specific segment of your target audience. By comparing a “treatment group” (those who see the ad) against a “control group” (those who do not), you can isolate the specific lift generated by the creative. The delta between these two groups represents your true incremental conversion rate.
What You Need Before Testing Creative Uplift
Running an experiment without the proper foundation is a recipe for “noisy” data. Before you dive into the Google Ads Experiments interface, you must ensure your account and campaign meet specific criteria to reach statistical significance. Statistical significance is the threshold at which you can be confident that your results weren’t just a product of random chance.
Meeting Conversion Volume Requirements
Data density is the fuel for any successful A/B test. Google officially recommends that your experiment generates at least 50 conversions across both the treatment and control arms during the testing period. For high-ticket items or B2B SaaS companies with long sales cycles, reaching 50 “final” conversions (like a closed deal) can be difficult within a month. In these cases, it is often more effective to optimize the test around high-intent micro-conversions. For example, instead of tracking “Completed Purchase,” you might track “Add to Cart” or “Schedule a Demo” to ensure you have enough data points to validate the test.
Budget Minimums and Consistency
An experiment is only valid if it is consistent. If your Demand Gen campaign is frequently “Limited by Budget,” the algorithm will intermittently stop showing ads to your treatment group. This creates gaps in the data and skews the results of the control group. To avoid this, ensure your budget is high enough to sustain continuous delivery for at least four weeks. If the campaign hits its daily cap early in the afternoon, the results may not reflect the behavior of users who browse in the evening, leading to an incomplete picture of performance.
Isolating the Creative Variable
One of the most common mistakes in ad testing is changing too many things at once. If you test a new video asset but also change the target audience and the bidding strategy, you won’t know which change caused the performance shift. For a clean asset uplift test, keep every other element—audiences, bidding, and standard static assets—exactly the same. This isolation ensures that any “lift” measured is directly attributable to the specific creative asset being tested.
How to Run an Asset Uplift Test in Google Ads
Setting up an asset uplift experiment is a structured process that requires careful planning. Since the launch of these features in November 2025, the process has become more streamlined within the Google Ads UI. Follow these steps to ensure your test is configured correctly.
Step 1: Define a Clear Hypothesis
A test without a hypothesis is just wandering through data. You need a specific question you are trying to answer. A weak hypothesis would be: “I want to see if this video is good.” A strong, actionable hypothesis would be: “Adding a 15-second customer testimonial video to our Demand Gen asset group will result in a 12% incremental lift in leads compared to our current lifestyle imagery.”
Step 2: Navigate the Experiments Interface
To begin, log in to your Google Ads account and look at the left-hand navigation menu. Select “Campaigns” and then “Experiments.” Click the plus (+) button to create a new experiment. From the list of options, choose “Asset tests provided by you” and specifically select the Demand Gen campaign type. This ensures the system uses the correct logic for cross-surface delivery on YouTube and Discover.
Step 3: Configure a 50/50 Cookie-Based Split
Google will ask how you want to split your audience. For the most reliable results, use a 50/50 cookie-based split. This method assigns a unique cookie to each user, ensuring that once a user is placed in the control or treatment group, they stay there for the duration of the test. This prevents “cross-contamination,” where a user sees the test ad and then later enters the control group, which would ruin the data’s integrity.
Step 4: Lock Your Variables and Start the Test
Once you launch the experiment, you must practice what marketers call “test discipline.” It is tempting to jump in and tweak bids or adjust targeting if you see a slow start, but you must resist. Any major change made mid-test introduces “noise” that can invalidate the statistical confidence of the results. Set it, and let it run.
Step 5: Define the Duration
Patience is a virtue in incrementality testing. A standard test should run for at least four weeks. The first week is generally considered a “learning period” where the Google Ads algorithm adjusts to the new audience split and learns how to best serve the creative. The actionable performance data typically accumulates between weeks two and four. If you are in a B2B environment with a conversion window longer than 30 days, consider extending the test to six or eight weeks to capture the full journey of the users you reached during the test period.
Interpreting Your Experiment Results
Once the experiment concludes, you will find a report in your Experiments dashboard. This report provides more than just raw numbers; it offers confidence intervals. Understanding these intervals is the key to making smart business decisions. Usually, there are three primary outcomes from an asset uplift test.
Outcome 1: Positive Lift with Statistical Significance
If your treatment group shows a positive lift with at least 95% confidence, your hypothesis was correct. The creative asset you tested is a “winner.” At this point, you should calculate your incremental Cost Per Acquisition (iCPA). You do this by taking the total ad spend of the treatment group and dividing it by the number of incremental conversions (the conversions over and above what the control group achieved). If the iCPA is within your brand’s profitability targets, you should scale this asset across other relevant campaigns.
Outcome 2: Negative Lift
It is a hard truth to swallow, but sometimes a new creative asset actually performs worse than your baseline. Perhaps the video is too long, the “hook” is unengaging, or the creative doesn’t resonate with that specific audience. If the data shows a negative lift, you should pause the asset immediately. While it feels like a “failure,” this is actually a massive win for your budget—you have used data to prove that an asset was hurting your performance, allowing you to stop wasting money on it.
Outcome 3: Inconclusive Results
If the difference between the control and treatment groups is negligible, the result is inconclusive. This usually happens for two reasons: either the sample size (conversion volume) was too small, or the creative difference wasn’t bold enough. If you find yourself with inconclusive results, the solution is rarely to tweak the existing test. Instead, try testing a radically different creative concept. In Demand Gen, small changes like a different button color rarely move the needle; big changes in storytelling, format (e.g., UGC vs. Studio), or the value proposition are what drive measurable lift.
The Strategic Value of Creative Testing
In the modern era of Google Ads, many traditional levers—like manual bidding and keyword matching—have been automated by AI. This has shifted the primary differentiator for brands from “technical management” to “creative strategy.” Creative is now one of the few remaining levers that a marketer can pull to significantly outperform the competition.
However, creative production is expensive and time-consuming. Whether you are hiring a production house or utilizing an in-house team for User-Generated Content (UGC), you need to prove that the investment is paying off. Asset uplift experiments allow you to move away from subjective opinions (“I think this video looks cool”) and toward data-backed decisions (“This video drives a 15% lower iCPA”).
By leveraging Google’s asset uplift tests, you can transform your Demand Gen campaigns from a “black box” of mysterious conversions into a transparent, scientifically-validated engine for brand growth. Start by establishing your baseline, run your experiments with discipline, and let the data dictate your creative roadmap. In a world of attribution illusions, incrementality is the only truth.