The framework
What stands between evidence and the money.
The same method runs on any operation, in any sector. You do not need to see the machinery, only what a decision is allowed to claim before effort and capital move.
Evidence-governed decision-making for physical assets is the discipline of deciding what deserves action, what to fix, fund, or defer across the operations you run, before effort and capital move, and stating what would make that call wrong.
Same method, any sector. Asset-agnostic by design.
Evidence foundations
A read doesn't lean on one source. It weighs the asset's data, its contracts and economics, the climate it runs in and the rules it answers to, against references that carry authority, and says where the evidence is thin.
- Standards & codes
- Physical record
- Operational data
- Financial
- Contracts
- Climate
- Local regulation
- Operator records
Many inputs in, few admissible claims out, each one level-stamped and ranked, or held back.
- 1Evidence in
- 2Filtered to admissible claims
- 3Ranked by what changes the call
Making heterogeneous sources agree, deterministically and traceably, where every claim reconciles with the rest before it can stand, is most of the work, and most of why it is hard to copy.
The method
Nine questions
A governed read runs the decision through nine questions. Each one can stop effort, engineering and capital from moving toward the wrong variable.
- 1What evidence is admissible?
- 2What is physically plausible?
- 3What depends on what?
- 4What is actually happening?
- 5What would prove or kill this?
- 6What capital is exposed if this is wrong?Finance
- 7What rules change the decision?Regulation
- 8What changes when new evidence arrives?Belief update
- 9What can move now, and what must wait?Decision
Energy is one input. So are physics, operations, finance, and regulation. The method doesn't depend on the sector.
From “we think” to “we know.”
A read may only claim as much as its evidence supports. It climbs four levels as evidence arrives, and always says where it stands before capital is committed.
Why it's hard to copy.
Not one model: several disciplines that must line up before anything is asserted: physics, evidence, finance and regulation, each able to veto a claim. The moat is the architecture, not the screen.
- Physics-based priorsHow assets like this actually behave.
- Operational contextHow this one is actually run.
- Verification pathwaysWhat evidence would settle it.
- Probabilistic financeExposure as ranges, not point estimates.
- Computable regulationWhich rules actually trigger here.
- Belief revisionThe read updates as evidence arrives.
- Traceable claimsEvery statement linked to its source.
- AdmissibilityWhether a claim has earned the right to stand.
ZLab. Governs the whole decision. Nothing is asserted until every layer lines up, and any one can veto.
An AI can produce a recommendation. It cannot decide whether the recommendation is admissible.
FAQ
Questions about the method
What does a governed read check?
Every claim runs five gates before it can stand: is it physics-consistent, backed by observed evidence, within this level's ceiling, checked against the rules that trigger, and able to survive an attempt to falsify it. Miss one and the claim is held back, not shown as fact, so what deserves action is separated from what only looks like it.
What is admissible evidence?
A claim must pass physics, observed evidence, the level ceiling, the rules, and a falsification test before it can stand. Anything that misses a gate is held back, never allowed to carry an operational decision on its own, and never allowed to commit effort, engineering, maintenance, validation, and eventually capital.
Why does the question matter more than the calculation?
A wrong question can survive thousands of correct calculations. The work governs what is actually being decided, the governing variable, before the numbers run, so prioritization under uncertainty rests on what deserves action and not on whatever was easiest to compute.
Does the framework replace our engineers, auditors, benchmarks or meters?
No. The framework does not replace your engineers, your auditors, your benchmarks or your meters. It governs how their evidence is combined, and decides what that evidence is allowed to support before an operational decision is made.
How is this different from an audit, a dashboard, or an AI copilot?
Each does one thing well and stops where the decision begins: an audit observes the asset closely, a dashboard shows the data, a copilot generates fluent language. None governs the decision. An AI can produce a recommendation; it cannot decide whether the recommendation is admissible.
Does the method depend on the sector?
No. Same method, the operations you run. The same architecture runs on a plant, a building, a data center or a cold-chain, in any sector.
What evidence does a read weigh?
Not one source. A read weighs the asset's physical record and operational data, including sensors, telemetry and digital twins, its contracts and unit economics, the climate it runs in, and the rules it answers to, against references that carry authority such as ASHRAE, IPMVP, DOE, EPA and NREL, and it says plainly where the evidence is thin.
Why is it hard to copy, and why is it not just a checklist?
Pulling heterogeneous sources into one read that is deterministic and traceable, where every claim reconciles with the rest before it can stand, is slow, meticulous work. It is most of what the method does, and most of why it is hard to copy. The references are not badges, they are the bar a claim has to clear before it can carry a decision.