Operationalizing Data Analytics Methodology
Earned trust in AI.

Choose one outcome. Name its owner. Agree on the evidence that justifies action.Choose an outcome. Name its owner. Agree on the evidence.

Start with one outcome →
Adaptive operations

Turn priorities into results you can verify.

ODAM connects the outcome you want, the evidence you have, and the authority to act. Use the Data Compass to reason through the decision, carry out bounded work, and check what changed.

Name an owner. Provide authority and resources. Set limits. Agree on the evidence and a review date. Give the people affected a way to challenge the aim, the measure or the rule.

ODAM as a loopFive steps in a closed loop: define the outcome, name the owner, use the Compass, carry it out, check and adjust. Evidence from the work returns to the check, and the check adjusts the outcome. Authority runs from the owner to the work. The people affected, at the center, can challenge the outcome and the check.The people affectedDefine the outcomeName the ownerUse the CompassCarry it outCheck and adjustevidenceadjustauthoritychallengeODAM as a loopFive steps in a closed loop: define the outcome, name the owner, use the Compass, carry it out, check and adjust. Evidence from the work returns to the check, and the check adjusts the outcome. Authority runs from the owner to the work. The people affected, at the center, can challenge the outcome and the check.The people affectedDefine the outcomeName theownerUse the CompassCarry it outCheck andadjustevidenceadjustauthoritychallenge
  1. 01
    Define the outcome
    What should improve, or must stay true, for the people you serve.
  2. 02
    Name the owner
    Authorize the work, with resources and limits.
  3. 03
    Use the Compass
    Decide from what the evidence supports.
  4. 04
    Carry it out
    Bounded work, recorded as it happens.
  5. 05
    Check and adjust
    Continue, change, pause or stop.
Start with one outcome Work through a decision
Three connected graphs

Connect commitments, evidence and decisions.

ODAM organizes what your team has committed to, the evidence it receives, and the decisions it makes by comparing the two. The three graphs keep those records connected, and keep clear what each one can establish.

Declared
What have we committed to, required or authorized?
Example (illustrative): a vendor's bank details change only after out-of-band verification; the payables lead owns payments.
Join
No. Comparison alone grants no authority to act. The owner decides, within the limits they were given.
How does this evidence bear on this commitment, and what decision follows?
Example (illustrative): the required verification is not evidenced, so the owner holds that vendor's payment and asks for a callback.
Observed
What does the evidence show, including what is missing?
Example (illustrative): an AI agent updated the bank details from an emailed request; there is no record of the required verification.

Comparison informs the decision; it does not grant authority.

Under all three · the Ledger
Every decision stays reviewable.
  • What was knownThe evidence the decision rested on, with its source and time.
  • Who decidedThe owner, the authority they used, and when.
  • What happenedThe action taken and its observed effect.
  • Every correctionAdded alongside, never written over, so you can see how you got here.
See the three graphs in the Explorer Shared reference material, not your organization's live evidence.
The Explorer's library is shared reference material. Its Observed and Join areas show structure, not a live organization's evidence.
A row of check gates with red railings holds back a full canal; green fields and mountains beyond.
Gravity-fed

A gate is a decision.

A gate doesn't decide anything by itself: someone with standing decides who may open it, and when. An acequia moves water downhill, on time, under shared rules. Data becomes useful when it helps someone decide what to do. Sometimes the right decision is to keep the gate shut.

Data Compass

Work through the decision.

The graphs keep the records connected. The Data Compass is how people reason about them: what is happening, what could explain it, what remains unknown, and what action the evidence supports.

How the Compass works
Three numbers

Three numbers

An old round headgate set into a concrete wall, under green leaves.
01

Days at Risk

How long does each known exposure stay unmitigated, counted from the earliest record that shows we knew?

02

Time to Good Decision

How long from the first anomalous signal to a recorded decision that holds through its validation window?

03

Ambient Breach

Are we keeping up? Does ambiguity arrive faster than we can verifiably repair it?

Why now

Machine-speed work needs more than a checklist.

Attackers can vary their tactics, divide a harmful objective into ordinary-looking tasks, and coordinate the effects across agents and tools. Each action can pass its own check while the combined result causes harm. Attempts can arrive faster than people can review them.

The possibilities are open-ended; rules, time and capacity are finite. Rules can still constrain cases nobody anticipated. The hard part is enforcing those limits across changing context, concurrent work and cumulative effects.

How the parts work together.

Connect the outcome, the evidence and the authority to act. Select any card to see what it takes in and what it hands on.

Threat pressure

Open-ended attacks against finite rules

Each action can pass its own check while the combined result causes harm.

  • Machine speed
  • Novel combinations
  • Split objectives
  • Swarm coordination
Open-ended attacks against finite rules

Attack paths keep changing, harmful objectives arrive in harmless-looking fragments, and consequences accumulate across agents. Finite rules and capacity have to address context, composition and cumulative effects.

Inputs
  • Open-ended attack paths
  • Coordinated agents and tools
Outputs
  • Attempted actions at agents and the gateway
  • Anomalous signals in Observed
A small steel slide gate with a red hand wheel, set in a concrete ditch among brush.
Authority opens the gate.Evidence informs the decision.
Purpose

People, purpose and accountable authority

Set the outcome, the owner, the resources and the limits. Hear challenges from the people affected.

People, purpose and accountable authority

One outcome for the people served. Someone accountable grants and scopes authority. The people affected can challenge the aim, measure, rule or evidence.

Inputs
  • Measurements at review
  • Challenges from the people affected
Outputs
  • The outcome, owner, resources and limits
  • Scoped grants and delegation
  • Challenges into Join and the Ledger
Appropriateness · should

Tamed Autonomy

Should this action happen?

Tamed Autonomy

A normative judgment made before anyone asks whether an action may proceed. Norms bind once an accountable party adopts them. Contributed by Robin Martherus.

Inputs
  • A proposed purpose and use
Outputs
  • Adopted purpose and norms, into Declared
Delivery and adaptation

ODAM

The delivery loop, from an outcome to a reviewed result.

  • 01 Define the outcome
  • 02 Data Compass
  • 03 Agents and automated work
  • 04 Review the result
ODAM

The delivery loop. People's tools, services and AI agents request and carry out bounded work; the team reviews effects and unknowns and decides to continue, change, pause or stop.

Inputs
  • The outcome and its limits
  • Requirements and evidence
Outputs
  • Decisions with an owner
  • Requests to act
  • Effects, observed
Data Universe · three connected graphs

Declared → Join ← Observed

Commitments on one side, evidence on the other, compared in Join.

  • Declared: what we committed to
  • Observed: what the evidence shows
  • Join: how they compare
Declared → Join ← Observed

Declared holds commitments, requirements, roles, scope, authority and the decision profile. Observed holds readings and events with sources, times and limits, including what's missing. Join matches scope and context, compares, and records findings and decisions. Comparison grants no authority by itself.

Inputs
  • Requirements and adopted norms
  • Readings and observed effects
  • Presence evidence
  • Shared vocabulary
Outputs
  • Findings and decisions
  • Limits and current context for each action
  • Evidence for reasoning
Decision reasoning

Data Compass

What is happening? What could explain it? What action is supported?

Data Compass

The reasoning practice inside the loop: connect observations, explanations and unknowns to the action the evidence supports.

Inputs
  • Findings from Join
  • Requirements and evidence
Outputs
  • A decision, its owner and time, into the Ledger
  • The supported action
History · the record of every decision

Ledger

What was known, who decided, what happened, and every correction since. It's what makes a decision reviewable. Together, Ledger and Join provide the records behind the three measures.

  • Append-only
  • Owner on every decision
  • Evidence linked
  • Corrections added, never overwritten
  1. SignalAn agent changed a vendor's bank details. No record of the required verification.
  2. DecisionHold that vendor's payment. Owner: payables lead.
  3. VerificationA call to the number on file: the vendor asked for no change.
  4. CorrectionBank details restored. The hold stays until review.
Illustrative record.
Ledger

What was known, who decided, what happened, and every correction. Records keep their source, scope and history; corrections are added, never written over.

Inputs
  • Decisions from the Compass
  • Decision and execution records
  • Challenges and corrections
Outputs
  • History for review
  • Records for the three measures
Measure

Days at Risk

How long does known exposure remain unmitigated?

Days at Risk

A plain count from the earliest record that shows you knew, never shortened by a slow acknowledgment.

Inputs
  • Ledger: awareness to mitigation
Outputs
  • Where to fix next, at review
Measure

Time to Good Decision

From the first signal to a decision that holds.

Time to Good Decision

The decision must hold through its stated window. Measured per incident class at the 90th percentile, with reversed and open cases shown beside it.

Inputs
  • Ledger: first signal to decision
Outputs
  • Whether response is fast enough, at review
Wooden check boards set in a concrete diversion box among green brush.
Every opening is on the record.Who decided, on what evidence, and what changed.
Measure

Ambient Breach

Can verification and repair keep up, with reserve?

Ambient Breach

Compares the arrival of uncertainty with verified repair. A flat backlog doesn't prove the reserve is safe.

Inputs
  • Join: arrivals against verified repair
Outputs
  • Staff, automate or escalate, at review
Execution

Action Gateway

Binds the action. Rechecks at release.

Action Gateway

Binds the decision to the exact action, sends the actor and action to the Infrastructure Layer for evaluation, enforces the result at release, and records the execution. After a required human review, a fresh evaluation runs before release.

Inputs
  • The actor and exact action
  • The decision, limits and supervision from the Infrastructure Layer
Outputs
  • The actor and exact action, to the Infrastructure Layer
  • Release of the authorized action, within limits
  • Decision and execution records, into the Ledger
  • Holds for required human review, to the owner
Infrastructure

Infrastructure Layer

Verifies state. Evaluates each action.

Infrastructure Layer

Evaluates each action and answers the gateway through its own interface.

Inputs
  • Actor and exact action
  • Scoped grants
  • Declared limits
  • Current context
Outputs
  • Decision, limits, supervision
Authorization

Kinetic Trust Protocol

Defines the bounds on authority. Conditions tighten grants; they never widen them.

Kinetic Trust Protocol

Authority is evaluated at the moment of the act, from the grant, the exact action and the current context, not only from a credential issued earlier. A correctly evaluated veto is not bypassed; corrections lead to a fresh evaluation.

Inputs
  • Scoped grants and delegation
  • The exact action request
  • Limits and current context
Outputs
  • Allow, shape, require supervision, or deny
  • Act and record, or hold for review
Reference view

Data Universe Explorer

Browse the library and its links.

Data Universe Explorer

Trace outcomes, work, roles and signals across Macro, Meso and Micro.

Inputs
  • The shared reference library
Outputs
  • Reference structure for Declared
Coordination and registry

Open Trust Commons

Shared terms, declared claims and evidence records.

Open Trust Commons

Projects describe what they claim and what supports it, in one vocabulary, so they can be compared. An entry is not an endorsement.

Inputs
  • Project scope, claims and evidence
Outputs
  • Shared vocabulary for Declared
Presence evidence

Noma

Mutually chosen nearby encounters.

Noma

Presence informs the evidence. Presence is evidence, not permission.

Inputs
  • Mutually chosen encounters
Outputs
  • Presence evidence in Observed
See the whole map on one page Details: every part, its inputs and outputs
Any size

One platform team and a global enterprise start with the same seven questions.

Water pours from the mouth of a pipe into a ditch, bright in the sun.
One outcome at a time.
Scope, authority and resources stay local. The questions don't change.
One platform team · illustrative
1outcome

Agent-initiated payments reach the approved beneficiary account, within authorized limits. One owner, one aim, seven questions answered with what they can show.

A global enterprise · illustrative
400outcomes

Four hundred outcomes, the same seven questions. Each keeps its own evidence and profile, so a strong program can't hide a weak one.

Seven questions
  1. 01Aim
  2. 02Measure
  3. 03Evidence needed
  4. 04Coverage
  5. 05Detection
  6. 06Bounded work
  7. 07Loop
A small steel check plate set in a narrow concrete ditch beside a house.
Small gates count.
An old steel gate frame on a concrete headwall, stones stacked beside it.
Old works too.
Your turn · about five minutes
Set your aim and see where you stand.

Pick, type and slide through the seven questions. No score, and nothing leaves your device.

Start with one outcome →
Five grades, plus unknown · never added into one number
See the maturity profile
  1. Absent
  2. Stated
  3. In place
  4. Checked
  5. Held
  6. Unknown
Rooted in water
Data Universe Explorer

A map of risks, outcomes, measures and signals, and how they connect. Not for sale: ask for it.

Where it came from
ODAM v1, 2023

The original methodology, kept as published. ODAM grew into the Kinetic Trust Protocol .

A painting of night irrigation under moonlit red cliffs near Abiquiú, water spreading in silver sheets.

The water keeps moving.