Skip to content
Wonder Constant Labs

Methodology

How a question becomes evidence—and then understanding.

Ten stages, grouped under the three parts of the operating principle, repeated as a loop. Each stage owes something specific — a record, a definition, an instrument, a measurement, a test that could fail.

The operating principle

Wonder is the constant.
Observation is the method.
Engineering data into understanding.

We transform observations into structured data, data into models, and models into tools that make complex systems understandable.

  1. 01Wonder generates questions

    Wonder is the constant.

    Every system we study changes — the evidence, the models, the conclusions. The impulse to ask what is actually happening does not. It is the one term that stays fixed across every investigation.

    A field of curves displaced around a single straight axis, with one point held fixed at the centre.
  2. 02Observation captures evidence

    Observation is the method.

    A question becomes tractable the moment you can say what would count as an answer. Disciplined observation is how curiosity is converted into evidence: defined quantities, characterised instruments, recorded uncertainty.

    A coordinate system with sampling lines falling across it, capturing a data point where each line crosses the underlying signal.
  3. 03Engineering produces understanding

    Engineering data into understanding.

    Data is a record of what happened. Engineering builds the models, systems, and tools that turn that record into something you can reason with, test against, and be wrong about in public.

    The same data points with a fitted curve drawn through them and residual ticks showing the disagreement between model and measurement.

The field under observation

Every complex system leaves a field of evidence. Measurement is how we learn to read it.

The structure below drifts, aligns, and resolves the way an unmeasured system does under observation — never a fixed picture, always a reading in progress.

The order matters most at the start, where the temptation is to explain before describing. It is written down not to be rigid, but to be accountable: every stage below names the artefact it owes, and an initiative that skipped one should be visible from what is missing.

Select a stage to see what it produces.

Phase 01Wonder

Wonder is the constant.

Questions come first, and they are cheap to state badly. These stages are about making one worth answering.

  1. 01Observe

    Watch the system as it actually behaves, before imposing a story on it. Record what happens, including the parts that do not fit.

    OwesA description with no explanation attached to it yet.

  2. 02Define

    State the question in terms that could be answered. Name the variables, the boundaries, and what would count as evidence either way.

    OwesA question with a falsifiable shape.

Phase 02Observation

Observation is the method.

Where curiosity becomes evidence: the phenomenon is instrumented, measured under known conditions, and structured into data. Nothing downstream can be better than the instrument that produced it.

  1. 03Instrument

    Build or choose the tools that turn the phenomenon into signal — sensors, logs, simulations, collectors — and characterise what they miss.

    OwesAn instrument, and a written account of its blind spots.

  2. 04Measure

    Collect observations under conditions we understand. Track provenance, resolution, and error, so a number can still be trusted a year later.

    OwesRaw observations with their conditions recorded.

  3. 05Structure

    Turn observations into data: a schema, units, identifiers, and the transformations applied. What was discarded is written down too.

    OwesA structured dataset and its lineage.

Phase 03Engineering

Engineering data into understanding.

Data is a record of what happened. These stages build the thing that explains it — and expose it to being wrong.

  1. 06Model

    Compress the data into something we can reason with — an equation, a distribution, a simulation, a mechanism — and state its assumptions.

    OwesA model, with its assumptions in the open.

  2. 07Test

    Confront the model with data it has not seen. Design the test so a wrong model can actually fail it.

    OwesA result the model could have failed.

  3. 08Analyse

    Separate signal from noise and effect from artefact. Quantify how much of the result is real and how much could be chance.

    OwesAn effect with an uncertainty attached.

  4. 09Refine

    Fold what we learned back into the question. Discard what failed, sharpen what held, and expose the next assumption to test.

    OwesA better question than the one we started with.

  5. 10Publish

    Write it down so others can check it — methods, data, uncertainty, and the places we are still unsure. Understanding that cannot be shared is incomplete.

    OwesA record someone else could argue with.

The loop closes

Publishing is not the end of the cycle. What we learn — and especially where we were wrong — becomes the next thing we observe.

Refinement returns to observation, and the understanding compounds. A model that survives one turn of the loop is not proven; it is still standing, ready to be tested again.

Measured wonder: see the method applied.

Explore the research