Shape & Signal

How We Work

Building is not the first step.

Shape & Signal forms hypotheses, tests them, and makes explicit decisions based on evidence. Not every signal becomes software — the method is built to reach that conclusion honestly, as often as it reaches the opposite one.

  1. 01
    Signal
    Is the problem real?
  2. 02
    Shape
    What's the smallest useful solution?
  3. 03
    Ship
    Into real use.
  4. 04
    Learn
    What did reality say?
  5. 05
    Reshape
    What should change?

Projects can move backward — Reshape often leads back into Shape, not forward into a new Signal.

01

Signal

Is the problem real?

We start by understanding the environment: workflows, friction, the people affected, the incentives at play, the tools already in use, and what's already been tried. Signal has to be sufficiently real and sufficiently understood before anything gets shaped — a feature list is not a problem statement.

Recon

We observe how work is actually being done, talk to the people doing it, understand existing tools and workarounds, identify recurring friction, and distinguish real problems from assumed ones.

02

Shape

What's the smallest useful solution?

Given what Signal found, we determine what should actually be built — not the most complete version we can imagine, but the smallest one capable of proving or delivering the thesis. Shape is where scope gets deliberately constrained, not expanded.

03

Ship

Into real use.

We put something real into use. Shipping isn't the finish line — it's the point where a product becomes exposed to reality and starts generating the information the next stage depends on.

04

Learn

What did reality say?

We observe usage, behavior, outcomes, and failure points. Evidence only matters if it's capable of changing a decision — collecting it for its own sake isn't Learn, it's just data.

05

Reshape

What should change?

Improve, narrow, expand, reposition, rebuild, or stop — based on what Learn actually found. Reshape usually leads back into Shape for another pass; stopping altogether is a legitimate outcome, not a failure of the process.

Have a problem worth running through this?