I spent the first few weeks of a brand-new Pax8 product without designing a single screen.
Everyone says they want to build something from scratch. No legacy code and nobody else’s mess to work around. It sounds like freedom.
Then you get one, and you realise every decision you make is about to become the system, and anything you put off will turn into someone else’s expensive problem later. There’s also nothing to point at when two people disagree, because nothing exists yet.
That last part is the hard bit. On a normal project you can settle an argument by looking at what’s already there. On a greenfield one, every stakeholder has a different product in their head, and they’re all describing it with the same words.
The Pax8 monetisation work started with no brief and five stakeholders who each pictured something different. I didn’t design anything for the first few weeks. I spent them getting everyone to agree on one question: what does success look like for a partner using this in their first week?
From then on, every argument about what to build could be checked against that question.
There’s research behind this. Successful design teams use disagreement to reduce uncertainty, while unsuccessful ones come out of every argument more confused than they went in. The difference is usually whether there’s a shared model concrete enough to disagree with.
Kahneman’s work on anchoring matters here too. When there’s no reference point, people anchor on whatever comes first, like the first wireframe or the first confident person in the meeting. On a greenfield project, whoever frames the problem first has a huge amount of influence over everything after it. I’d rather that be the designer.
The other thing nobody warns you about is how long your early decisions last. You’re building the product and its conventions at the same time, with no history to check against. The patterns you set in the first release will still be there years later, in screens nobody has imagined yet.
So I treat a blank slate as a responsibility to set good constraints early, rather than permission to do anything I like.
References