Part of the Research Edge Series
You’ve read the business bestsellers. You know the playbook.
“Successful companies have flat hierarchies. They hire for culture fit. They move fast and break things. They obsess over customers.”
So you implement these traits in your organization. You spend months restructuring. You change hiring practices. You adopt new mantras.
And… nothing happens. Or worse, things get worse.
What went wrong?
Here’s what typically happens:
A tech giant offers free meals and nap pods. A smartphone maker obsesses over secrecy and design perfection. A social media platform moves fast and breaks things.
They succeed spectacularly.
Business journalists write glowing profiles. Consultants create frameworks. Books get published. And we all conclude: “That’s the formula. Do what they did, get what they got.”
This is the halo effect in action—when companies succeed, we glorify everything they do. When they fail, we blame everything they did. Same traits, different outcomes, completely different narratives.
But here’s the uncomfortable truth: thousands of failed companies had the exact same traits. We just never heard their stories.
This isn’t just about wasting time copying superficial traits. This is about the fundamental difference between anecdotal pattern-matching and rigorous causal analysis.
Most business advice is built on the former. Sharp business minds and researchers insist on the latter.
Let me show you why.
When you see a successful company with a particular trait, one of three things is usually true:
Example: Large successful companies can afford luxury perks like free meals, nap pods, and generous parental leave.
Small struggling companies copied these perks thinking they’d drive success.
But the perks didn’t create success—success enabled the perks. You copied the outcome, not the cause.
Example: A startup gets massive venture funding. This allows them to:
They succeed. We attribute it to their “risk-taking culture.”
But the real driver was the funding, which enabled both the risk-taking AND the success. The trait was a symptom, not a cause.
Sometimes traits and success just happen to occur together without any causal relationship at all. Like the famous “ice cream sales and drowning rates” correlation—both increase in summer, but neither causes the other.
This is where it gets interesting. When academic researchers and rigorous business analysts tackle this question, they don’t just look at successful companies and list their traits.
They deploy methods specifically designed to isolate causation from correlation. Here are four approaches:
This is the classic test/control approach. Randomly assign companies or teams to either implement the trait or not, keep everything else constant, and measure outcomes.
The problem: You can’t randomly assign corporate strategies to real companies. This works great in labs, not so much in boardrooms.
Track companies over time and see what happens FIRST. Did the trait appear before success, or after?
If you notice that most companies adopted “flat hierarchies” AFTER they grew large and successful (not before), that’s a strong signal you’ve got reverse causality.
This is where mathematics saves us. Instead of comparing all successful vs. unsuccessful companies, you compare companies that are IDENTICAL except for the one trait you’re testing.
Let me show you how this works with a real example.
Let’s say you want to answer: “Do companies with female CEOs perform better financially?”
The naive approach: Compare average profitability of all female-led companies vs. all male-led companies.
The problem: Female CEOs are rare in heavy manufacturing and oil/gas (low profit margins) but more common in tech services and retail (higher margins). You’re not comparing CEO gender—you’re comparing industries.
The regression solution: Mathematically “hold constant” industry, company size, market conditions, founding year, and economic climate. Now compare female vs. male CEOs who are operating in otherwise IDENTICAL circumstances.
This is what “controlling for confounding variables” means. You’re creating an apples-to-apples comparison even when you can’t run a controlled experiment.
Think of it like testing fertilizer on plants. If some plants get more sunlight, you can’t tell if better growth came from fertilizer or sunlight. You need to compare plants with the SAME sunlight but different fertilizer.
Regression does this mathematically for dozens of variables simultaneously.
Here’s another powerful technique: finding statistical “twins.”
The question: Does attending business school increase entrepreneurial success?
The problem: People who attend business school are already different—wealthier families, better networks, more prior work experience, different personality traits.
If MBA holders succeed more, is it because of the MBA, or because they were already on a trajectory toward success?
The solution: Match entrepreneurs in pairs who are virtually IDENTICAL across 15-20 measurable characteristics:
Now compare:
You’ve created statistical twins who differ only on the ONE variable you care about. This gets much closer to proving causation.
The limitation: You can only match on what you can measure. If there’s some invisible factor (like “hunger for success” or “risk tolerance”) that both drives people to get MBAs AND drives entrepreneurial success, you’ll still have bias.
Even with these sophisticated methods, here’s what rigorous researchers will tell you:
Proving causation with certainty is nearly impossible in business contexts.
We’re establishing probable cause, not absolute proof. We’re ruling out alternative explanations and building evidence.
But this probabilistic approach is infinitely better than the alternative: copying traits from success stories with zero evidence they actually caused the success.
The next time you read a business book or article about “7 Habits of Successful CEOs” or “The Culture That Built This Unicorn,” ask yourself three questions:
If they only looked at successful companies with the trait, they’re showing you correlation, not causation. You need to know: what about unsuccessful companies with the same trait?
Are they comparing apples to apples? Or are they comparing tech startups to manufacturing companies and attributing differences to “culture”?
Did the trait come BEFORE success, or did success enable the trait? Timing matters enormously.
If the answer to any of these questions is “no” or “unclear,” you’re reading anecdotal pattern-matching, not rigorous analysis.
Most business advice is built on studying winners and listing their traits. This creates an illusion of insight while providing little actual guidance.
Rigorous research controls for confounders, establishes temporal sequence, and rules out alternative explanations.
The difference between these approaches?
One wastes years and millions of dollars chasing superficial traits. The other gives you actual odds of success based on probable causation.
Which approach is your strategy built on?
If this resonated with you, I strongly recommend “The Halo Effect” by Phil Rosenzweig. It’s a surgical dissection of why most business research misleads us, with specific examples of famous studies that got it wrong.
Also check out:
These books will fundamentally change how you evaluate business advice.