In an era obsessed with ROAS dashboards, attribution models, and performance marketing teams, the idea that paid media could be non-essential sounds like heresy.
But that’s exactly what several of the world’s biggest brands found when they turned off digital ads, business didn’t change.
Around 2017–2020, major advertisers, including P&G, JPMorgan Chase, eBay, and Uber, began conducting internal tests to measure the true impact of their digital ad spend.
What they discovered challenged the very foundations of modern advertising.
Here’s what they did:
JPMorgan Chase slashed the number of sites its programmatic ads ran on from 400,000 to just 5,000 carefully vetted domains.
Uber paused $100 million in digital ads and found two-thirds of it had no incremental effect.
eBay famously ran a large-scale experiment showing that much of its paid search was cannibalizing organic traffic, not creating new sales.
The Uber Shock: $100M in Spend, Zero Incrementality
Uber’s data science team ran a controlled experiment where they stopped spending on certain performance channels. The analysis showed that up to 2/3 of the attributed conversions would have happened anyway, even without ads.
They cut $100M and saw no dip in core growth metrics.
JPMorgan’s Ad Quality Pivot
JPMorgan Chase used to run programmatic ads across 400,000+ websites. After a brand safety incident, they cut that list to just 5,000 curated sites—a drop of 99%.
The result? No change in performance metrics.
The bank kept the tighter list permanently, vastly reducing fraud risk and wasted impressions.
eBay’s Infamous Paid Search Test
In one of the most talked-about academic studies in digital advertising, eBay’s marketing team collaborated with economists to run randomized experiments on brand and non-brand search.
They paused branded search ads across U.S. markets.
Result: No change in sales.
For non-brand terms, the results were mixed. Some drove incremental revenue, others didn’t.
The headline? Billions in paid search spend wasn’t moving the needle.
As digital advertising platforms become more automated, more black-boxed, and more crowded, marketers risk overpaying for conversions that would’ve happened anyway.
What’s really at stake?
Wasted spend on non-incremental impressions.
Overreliance on attribution models that claim false credit.
Brand safety and fraud risks in unvetted programmatic inventory.
The Takeaway: If you can test Incrementality
And they learned:
“Just because a platform reports a conversion, doesn’t mean it caused it.”
