The Wrong Kind of Right
There's a famous Steve Jobs clip where he talks about focus. But the part people miss is what he says about being wrong. He describes how the best ideas often came from people he initially disagreed with—and how being "right" individually matters far less than getting to the right answer collectively.
This reframed how I think about conflict. A disagreement isn't a battle to win—it's a signal that we have different mental models. And different mental models mean we have an opportunity to learn something.
From Debate to Experiment
When two people have strong, opposing product intuitions, the worst thing you can do is argue until someone gives up. The second worst thing is to compromise into something neither person believes in.
Instead, I ask: "How can we test this?" Turn the disagreement into a hypothesis. Run with one direction, measure the outcome, and see if reality matches what we expected.
- State each position as a hypothesis with a predicted outcome
- Agree on what "success" looks like before running the test
- Pick the cheapest way to get signal (not the most thorough)
- Commit to following the data, even if it proves you wrong
Cheap Ways to Experiment
Teresa Torres has written extensively about continuous discovery and lightweight experimentation. Here are the methods I use most often:
Fake Door Tests
Add a button or link for a feature that doesn't exist yet. Measure how many people click it. If nobody clicks, you've saved weeks of development. If lots of people click, you've validated demand before writing a line of code.
Wizard of Oz
Make it look automated, but do it manually behind the scenes. This lets you test the experience without building the infrastructure. I've run entire "AI features" that were actually just me responding quickly in a queue.
Concierge Tests
Deliver the service manually to a small group. Don't scale it—learn from it. Watch how people actually use what you're offering. The patterns you see will inform what to build.
5-Second Tests
Show someone a design for 5 seconds, then ask what they remember. If they can't identify the main action or value prop, your design isn't clear. Takes 10 minutes, costs nothing, saves weeks of building the wrong thing.
Prototype Testing
This is where my "code prototyping" approach shines. Build a working version in a day, put it in front of real users, and watch what happens. Not a Figma mockup—something they can actually interact with.
The Meta-Lesson
The goal isn't to avoid conflict. The goal is to make conflict productive. When someone disagrees with me, I've learned to get curious instead of defensive. What do they see that I don't? What assumption am I making that might be wrong?
And when we can't resolve it through discussion, we run an experiment. The data doesn't care about seniority or who argued more convincingly. It just tells us what's true.
Further Reading
- Teresa Torres - Continuous Discovery Habits
- Marty Cagan - Inspired & Empowered
- Eric Ries - The Lean Startup
- Rob Fitzpatrick - The Mom Test