Are Real-World Data Companies Just $150 Burger Joints?
What Flatiron Health teaches techbio founders about margin, scale, and survivability.
Intro
I’m Danny—former pharma-tech founder who bootstrapped to exit, now host of Life Sciences Today. I help techbio and digital health CEOs grow revenue.
🚀 For sharp thinking on growth, subscribe and read my Friday essays.
This week’s essay is about restaurants and real-world data.
Flatiron Health — much more than real-world data.
I recently hosted Alex Deyle from Flatiron Health on my podcast. He leads clinical research at Flatiron—a company that improves the design and execution of oncology trials by turning messy real-world data into clean, usable insights.
Flatiron didn’t succeed because it built a real-world data SaaS company.
It succeeded because it built a high-margin, defensible system around a unique two-sided business model.
They started by doing something non-scalable.
They assembled an army of clinically trained professionals—including nurses and cancer data specialists—who manually reviewed patient records to extract meaningful insights. This process helped transform messy, unstructured oncology data into structured, usable information at scale. Over time, they combined this human expertise with Flatiron’s software tools to handle millions of data points more efficiently.
Then they launched their own community Oncology EMR.
They didn’t stop at data—they built a network.
Flatiron’s unique two-sided model:
Oncology sites use the community Oncology EMR for free.
Pharma sponsors share risk and reward.
The Flatiron business model works because of three key lessons every techbio founder needs to learn:
Margins matter.
Scalability isn’t obvious.
Go-to-market is everything..
See the Oncology Trials with Flatiron Health - Life Sciences Today Podcast Episode 15
1. Low Margins Kill Good Ideas
I once worked at a restaurant between tech jobs. Thought I’d wait tables. Spent two weeks peeling potatoes.
Restaurants are brutal. Margins for full-service restaurants hover around 3–5%. Fast food runs a bit higher at 6–9%. Most don’t survive five years.
SaaS isn’t much better. We romanticize 90% gross margins, but most startups bleed cash on support, feature churn, and sales. In reality, they run below 10%. More like a burger joint than a gold mine.
IQVIA? Pretax margin: ~10%
Oracle? Pretax margin: 29%
Why the difference? Because Oracle sells software, not services. They get paid upfront. They charge for maintenance. They don’t burn to scale.
In a real-world data company—where data collection, cleaning, labelling and validation can take 5–10 years—margins are everything. They buy you time to survive.
2. Scalability Is a Trap
You and your friend want to open a burger place. He used to work at Oracle—he wants to sell $150 burgers with options. You went to business school—you want to start with $15 burgers and upsell.
You put both on the menu. Customers assume it’s a scam.
You can’t scale by confusing your customer. You have to pick your positioning.
Flatiron didn’t pretend to be everything. They didn’t build a self-service dashboard for “healthcare data consumers.” They focused on oncology. Then they got obsessive about the one thing that mattered: data fidelity in clinical workflows.
Want to scale a real-world data business?
Pick your lane.
Price it for margin.
Build trust before volume.
3. Services vs. Products
A lot of startups begin with professional services.
It pays the bills and gets you inside pharma.
But that’s the trap.
Service projects don’t scale. They create entropy. You become a services firm pretending to be a product company.
And here’s the kicker:
No professional services company in history has transformed itself into a dominant product business.
Why?
Remember The 22 Immutable Laws of Marketing?
The most violated is the Law of Line Extension.
One day you’re focused and profitable. The next, you’re spread thin and bleeding.
Ivory soap. Ivory shampoo?
Coors beer. Coors water?
Chanel No. 5. Chanel for men?
Take IBM.
In the 70s and 80s, IBM dominated mainframes. In 1991, they pulled in $65B in revenue. Adjusted for inflation, that’s $148 billion in today’s dollars.
In 2015, IBM launched Watson Health—hyped as the future of AI in medicine.
By 2022, after repeated failures, they sold it and wrote down $14 billion.
Today? IBM’s 2024 revenue is $62B—less than half their 1991 revenue in real terms.
A sharp fall for a company once considered the backbone of enterprise computing.
When you try to be everything to everyone, you end up meaning nothing to anyone.
Personally, I’d rather be strong somewhere than weak everywhere.
More is less. The more projects you run, the more partnerships you chase, the less money you make. Each service project is different. Each one pulls your team in a new direction. Each project eats management attention.
Less is more. Narrow your focus to earn a position in the buyer’s mind. Focus leads to compounding: improved product, better operations, sharper,tighter team.
If your product is: "We do real-world data development projects in colorectal cancer", then make that your product.
Say no to everything else.
Every custom data project you take that’s off-thesis weakens your product and team focus.
Flatiron’s Genius? Doing the Unscalable—Then Making It a Platform
Let’s break down what Flatiron actually did right:
Started with non-scalable work: an army doing manual data labeling
Listened to the customer: Pharma said “not enough patients,” so Flatiron built its own EMR
Created a flywheel: More sites → more data → more value for pharma → better trials → better outcomed
Aligned incentives: Free for sites. Pay-for-performance for sponsors.
This isn’t just ops. It’s a data network effect.
Data network effects (rarer than VC investment in cold fusion)
A data network effect is when a product’s value grows as a result of more usage via the accumulation of data. This is the most valuable type of defensibility you can build—and it’s rarer than most people realize.
One of the best consumer examples? Waze.
Here are the six characteristics you want:
Automatic data capture from customer usage (e.g., Flatiron’s community EMR)
Product value increases automatically as more data is added
High minimum threshold before the product delivers value—raising the barrier to entry
Incremental data remains valuable—avoiding early plateau
Data is central to product value, not peripheral
Customers perceive the data as valuable—it drives decisions and retention
Flatiron checks most, if not all, of these boxes.It seems that Flatiron ticks off most if not all of the data network effect checkboxes.
And that’s probably why Roche acquired them and as content to let them run as an independent business unit.
See Nfx Network effects Manual for deeper discussions of network effects.
Extro
Let’s go back to the $150 burger.
Consider Oracle. Their pre-tax margin is 29%—3× that of IQVIA.
Oracle revenue is 2× Salesforce. But Larry Ellison is worth 25× Marc Benioff.
That tells you a lot about the limits of the SaaS model.
The old Oracle model—“License our software” + “Pay for maintenance”—is still incredible.
Getting paid upfront, then getting paid to maintain it?
It’s better for techbio data startups than SaaS.
As I learned in the early days of Flask Data:
In that model, your customers pay for your development.
You need less investor capital
You own more of your company
You use your customer base to fund new growth
About Me
I’m a former pharma-tech founder who bootstrapped to exit.
Now I help techbio CEOs grow revenue—by solving the tech, team, and GTM problems that stall progress.
📈Every Friday: sharp, no-fluff insights on how real techbio companies grow revenue—plus candid interviews with the builders.
Join the private network for techbio entrepreneurs.
About Flatiron Health
Flatiron learns from real-world data and improves people’s experience with cancer with better-designed clinical trials, faster drug development and smarter care for patients.
🔗 Visit Flatiron Health