Incrementality Case Study: Longer Video, Larger Impact: Rigorous Geo-Testing Reveals :30s Outperform :15s in Top-of-Funnel Awareness Channels

In many marketing plans, :15 second video ads are the default choice for top-of-funnel media. They are less expensive to produce, cost less to serve, and fit easily into most ad placements. But recent evidence suggests that when it comes to revenue impact, shorter is not always better.

A national ecommerce & consumer-packaged goods brand in the makeup / beauty space partnered with Stella to explore whether longer ads deliver stronger business outcomes in passive, non-clickable environments like Connected TV (CTV), Online Video (OLV), and YouTube TV. The test was designed to isolate creative length as the primary variable while controlling for other factors that typically cloud media results.

‍

The Experimental Design

The experiment was structured as a geo-holdout study across 54 DMAs that were statistically balanced using the brand's historical sales data. The country was divided into three groups:

• Control Cell (A): No media exposure
  
  
• Test Cell 1 (B): :15 second non-skippable video ads
  
  
• Test Cell 2 (C): :30 second non-skippable video ads

‍

All ads ran across the same inventory sources: CTV, OLV, and YouTube TV. These were intentionally chosen because they are lean-back environments where users are unlikely to click, and where view-through impact is harder to detect without incrementality testing.

To ensure the creative itself wasn’t a confounding variable, the brand used the same core concept in both the :15 and :30 second versions. The :30 spot simply allowed for more message depth and brand storytelling, while the :15 was a condensed edit of the same script and footage. Both versions were high-quality, professionally produced, and approved for upper funnel use.

Budget was evenly distributed by region, not by impression count, so the :15s ultimately reached more people. That tradeoff was intentional and reflects the real-world decision marketers face when allocating spend across formats with different costs.

‍

Stella’s synthetic control modeling estimated the expected revenue in each region had no new ads been served. The actual results were then compared against this baseline to calculate the true incremental lift. The model showed a strong fit in both test cells:

• MAPE (Mean Absolute Percentage Error):
  
  
  • :15s Cell: 12.1%
    
    
  • :30s Cell: 10.4%
    
    
• Statistical Significance (p-value):
  
  
  • :15s Cell: p = 0.032
    
    
  • :30s Cell: p = 0.008
    
    
• Confidence Interval (90%):
  
  
  • :15s iROAS: 2.45 ± 0.21
    
    
  • :30s iROAS: 2.82 ± 0.17
    
    

MAPE values between 9 and 16 percent are generally accepted for field experiments of this scale. The narrower confidence interval for the :30s indicates more consistent lift across regions. Both tests reached significance, though statistical significance alone was not treated as proof. These results were evaluated alongside model accuracy, data stability, and post-test observations.

‍

The Results

The :15 second ads produced an incremental return on ad spend (iROAS) of 2.45. The :30 second ads delivered an iROAS of 2.82. That is a 15 percent improvement in revenue per dollar spent, despite the longer ads costing more to serve. CPMs for :30s averaged $32 compared to $25 for :15s.

Because the study observed total revenue across all channels, including ecommerce and marketplace sales, the lift reflects real business impact rather than platform-reported conversions. The observation window extended two weeks beyond the final ad delivery to capture any delayed purchase behavior, particularly important for mid-to-high consideration products.

‍

• :15s Ads
  
  • iROAS: 2.45
    
    
  • MAPE: 12.1%
    
    
  • Confidence Range: 2.24 to 2.66
    
    
• :30s Ads
  
  • iROAS: 2.82
    
    
  • MAPE: 10.4%
    
    
  • Confidence Range: 2.65 to 2.99

Despite a 28 percent higher CPM, with $32 for :30s versus $25 for :15s, the longer ads delivered a 15 percent higher incremental return. This was not an isolated result. Nearly all DMAs in the :30s group showed consistent outperformance. The expanded creative time allowed for stronger narrative depth and higher message retention, which translated directly into measurable revenue lift.

‍

Caveats and Controls

This was not a test of whether longer ads are always better. It was a structured comparison in a very specific media environment. The results apply to passive, top-of-funnel channels where the audience is not actively browsing or converting in real time.

For performance platforms like Meta, TikTok, or YouTube Shorts, we consistently see different dynamics. Shorter ads often perform better in those environments, where attention is limited and the creative must work in the first few seconds. That is not the case with non-skippable CTV or YouTube TV, where viewers are more likely to absorb a longer narrative.

The experiment also controlled for media overlap and frequency bias by assigning regions at the geo level and isolating campaign flights. We reviewed delivery logs across platforms to ensure there were no major differences in inventory quality or placement across cells.

‍

Why This Pattern Is Channel-Specific

It is important to note that this advantage does not extend to performance-focused platforms. In Meta, TikTok, and YouTube Shorts, we consistently see that :30s do not outperform :15s. In some cases, they underperform. These channels are optimized for rapid, high-frequency interaction. Success often comes from brevity and immediacy, not storytelling.

In contrast, channels like CTV and YouTube TV offer a passive, immersive viewing experience. Users are not expecting to click. They are watching. This creates space for longer messages to land and drive results in a way that feed-based media cannot replicate.

What This Means for You

If you are running upper-funnel media in passive channels like CTV or OLV, it may be worth revisiting how you allocate creative lengths. In this test and others we have conducted, :30 second ads tend to drive more incremental revenue than :15s, even after accounting for higher CPMs. The added time appears to help with brand recall and message retention, which translates into downstream sales.

However, creative strategy is not one-size-fits-all. These findings may not apply to your audience, your product category, or your brand voice. The only way to know what will work for your business is to test for yourself.

Want to see how Stella makes structured testing this rigorous available to every brand?
Check out the virtual demo of Stella’s incrementality platform.