Brandon Benson
CTO at Leo • July 17, 2025
Managing Your Ad Spend Like a Hedge Fund: The Future of Marketing
Managing Your Ad Spend Like a Hedge Fund: The Future of Marketing
Introduction and Real‑World Motivation
"Half the money I spend on advertising is wasted, the trouble is I don't know which half." – John Wanamaker [1]
Modern marketers still feel this pain, but the digital era gives us a new tool to solve it: incremental Return on Ad Spend (iROAS). iROAS asks a simple counter‑question: How much extra revenue comes only from the ads, compared with what would have happened if you spent nothing?
Real cases show why this matters:
- Uber (2019): Turned off roughly two-thirds of its performance budget (~$100 million) and saw no drop in new riders [2](The Hustle)
- Airbnb (2020): Cut most online advertising yet kept about 95% of normal traffic [3](LinkedIn)
- eBay: Ran large experiments and found search ads created "no measurable benefit" to sales [4](The Guardian)
These stories show that "good" looking ROAS can hide spend that is doing almost nothing.
Forward‑thinking brands now treat ad dollars the way hedge‑fund managers treat capital. They move budgets away from low‑lift campaigns and double down on the winners. One mid‑sized retailer recently cut $40,000 a month from low iROAS retargeting and re‑invested the same money into high‑lift prospecting. The result: a 15% jump in incremental revenue—about $200,000 extra every month, without spending more [5].
From ROAS to iROAS: What's the Difference?
| Metric | What it Measures | Why It Can Mislead |
|---|---|---|
| ROAS = Attributed revenue Ă· ad spend | Counts all sales that tracking software links to ads | Often credits conversions that would have happened anyway |
| iROAS = (Revenue with ads – revenue without ads) ÷ ad spend | Counts only the extra sales caused by ads | Strips out coincidence and double counting |
Think of ROAS as gross return. iROAS is the alpha that Wall Street chases—the part produced by real lift.
Four Practical Ways to Measure Incremental Lift
1. Randomized Holdout Tests
Split your audience randomly. Show ads to one group, withhold from another, then compare results.
- Pros: Cleanest proof of causation
- Cons: Opportunity cost of not advertising to some potential customers
2. Ghost or PSA Tests
The control group sees a public‑service message in the same ad slot instead of your real ad.
- Pros: Maintains reach while measuring true lift
- Cons: Requires placebo ad inventory and careful execution
3. Marketing Mix Modeling (MMM)
Uses years of spend and sales data to tease out each channel's statistical effect.
- Pros: Holistic view across all channels, great for TV and big budgets
- Cons: Data-intensive, slow to update, limited granularity
4. Causal Inference on Observational Data
Data scientists build "twin" user groups with matching traits, then compare outcomes with vs. without ads.
- Pros: Faster than experiments, works when tests aren't practical
- Cons: Relies on model quality and assumptions
Best Practice: Smart teams combine these tools—run experiments for ground truth, then let models fill the gaps between tests.
Ten Signals That Predict High iROAS
| Signal | Impact on iROAS | Why This Matters |
|---|---|---|
| High organic conversion rate | ⬇️ Lower lift | When people already convert without ads, there's little room for improvement |
| Large share of new customers | ⬆️ Higher lift | New customers represent true growth, not just shifted timing |
| Non‑branded search or cold prospecting | ⬆️ Higher lift | Users aren't already on the path to buy from you |
| Moderate ad frequency | ⬆️ Higher lift | Too many impressions hit the same users with diminishing returns |
| Fresh audience reach | ⬆️ Higher lift | Unsaturated audiences still have persuadable people |
| Heavy competitor bidding on your brand | ⬆️ Higher defensive lift | Ads prevent competitors from stealing your customers |
| Off‑peak timing | ⬆️ Higher lift | Ads can create demand when none naturally exists |
| Engaging creative (high CTR) | ⬆️ Higher lift | Strong engagement signals ability to persuade new users |
| Unique offer or value proposition | ⬆️ Higher lift | Converts fence‑sitters rather than just shifting purchase timing |
| Uplift or propensity scoring | ⬆️ Higher lift | Targets users unlikely to buy without the ad nudge |
Use these signals the way an analyst watches market indicators before placing trades.
Why AI Changes the Game
Human teams can analyze results monthly or weekly. AI engines like Leo can monitor every campaign minute by minute, predict iROAS before spend goes out, and move budgets instantly.
Leo's machine learning models study lift tests, creative performance, and audience signals in real time. The system reallocates spend the way high‑frequency traders shift portfolios—aiming for maximum incremental return while humans sleep.
This isn't science fiction. It's the next step in digital advertising, and it will leave slow, manually-run agencies behind.
Conclusion
Managing ad spend like a hedge fund means treating every dollar as an investment that must prove its worth. Traditional ROAS tells part of the story, but iROAS tells the truth.
Marketers who adopt lift testing, causal modeling, and AI allocation will grow faster and waste less. Brands that cling to surface metrics will fall behind as platforms like Leo rewrite the rule book.
The future belongs to whoever can measure, learn, and act the fastest—and that future is arriving now.
References
[1] John Wanamaker, retail pioneer, quotation circa 1900.
[2] The Hustle, "Uber was swindled out of 100 m dollars in ad spend," 2021.
[3] LinkedIn post by Kieran Flanagan, "Airbnb turned off paid marketing in 2020," 2021.
[4] The Guardian, "eBay study warns search ads have 'no measurable benefit'," 2014.
[5] Brenden Delarua, LinkedIn case study on incrementality testing, 2024.