Decision intelligence case study
Building Pūtake Labs: evidence-disciplined decision intelligence.
How Changeable co-founded Pūtake Labs, a bicultural decision intelligence practice built on eight integrated analytical Labs, a strict evidence-rating discipline, and a purpose-built engine that turns each analysis into a living, trackable record.
Project overview
Better decisions, more transparent outcomes.
Pūtake Labs is a 50/50 bicultural partnership between Changeable and Danielle Hodson, built around a shared conviction: consequential decisions, in government, iwi organisations and regulated industry, deserve the same rigour as a scientific finding. Pūtake, meaning root, source or foundation, reflects the practice’s aim of getting behind an outcome to what actually produced it.
Rather than start with software, the practice started with a method. Eight integrated Labs, each answering a distinct question about a decision, sit on a shared evidence-rating spine strict enough that every claim in every report carries an explicit confidence level, and a persona engine that surfaces the stakeholder positions a headline support figure usually hides. Only once that method had been stress-tested against real, difficult material was a purpose-built application layer added to make an analysis trackable over time.
Method before software
The analytical method was built, versioned and stress-tested before a single line of application code was written.
Evidence discipline as the spine
Every finding across every Lab carries an explicit High, Moderate or Low rating, with no exceptions.
Bicultural by design
A genuine Māori-Pākehā partnership, with a Kaupapa Methodology Module built into the practice, not bolted on.
Living, not terminal, analysis
A purpose-built engine turns a finished analysis into a structured record that can be tracked and stress-tested over time.
The problem
Most decision analysis is a one-off report, argued with more confidence than it earns.
Consequential decisions in the public and iwi sector are usually assessed once, in a report that reads with more numerical confidence than the underlying evidence supports. A projected outcome five years out is presented with the same authority as a documented fact from last week. Once delivered, the report is a terminal artefact: nobody tracks whether its assumptions held.
Changeable’s own internal audit of the emerging method found this trap directly: one of its own early reference analyses argued authority-drift percentages and scenario probabilities with more precision than its own evidence appendix could support. That finding shaped the entire build that followed.
What the gap looked like
- Consulting reports that conflate a well-reasoned inference with an established fact
- Stakeholder analysis that reports headline support while missing the hidden cost to specific groups
- No structured way to track whether a report’s key assumptions still hold months later
- Governance frameworks that ask whether a system works, not what it is doing to human decision-making
- Māori organisations and Te Tiriti obligations treated as an add-on rather than a genuine practice discipline
The Labs
Eight Labs across a decision’s full journey
Each Lab answers a distinct question. None is applied where it does not fit, and the practice explicitly names which Labs were set aside on any engagement, and why.
Insights Lab
How does work actually happen here now? The present-state baseline.
Retrospective Lab
Given what was known at the time, was each decision in a closed chain sound?
Decision Assurance Lab
What mechanisms must exist to know a decision is working, throughout its operational life?
Forecast Lab
What is likely to happen next, and what should be watched for?
Civic Lab
Is this decision sound, and under what conditions? The default lab for most forward-looking engagements.
Engage Lab
Tests how a decision will land with the people and groups it affects.
Change Lab
How will this transform the organisation and its people, beyond the org chart?
Consult Lab
Independent challenge, available at any point in a decision’s journey.
Decision Visibility
What happens to human decision-making after a system is embedded? Power, accountability, missing voices, human consequence.
How it was built
The method first, stress-tested, then versioned, then built into software
Rather than build an application and hope the analysis behind it held up, Pūtake Labs proved its method against difficult material before writing a line of application code.
A genuine 50/50 bicultural venture, not a branding exercise
The practice began as a formal joint venture between Changeable and Danielle Hodson, with intellectual property from both sides assigned into a new company at incorporation. The name Pūtake, meaning root or foundation, was chosen after alternatives were tested and set aside, including one rejected for an unwanted religious connotation, and the visual identity was deliberately briefed as a genuine coming-together of Māori and Pākehā design traditions rather than a selection of one over the other.
Five Labs and three shared engines, each earning its place
The method was built as a structured analytical system rather than software: each Lab answers a distinct question, explicit guidance defines when to decline a Lab that does not fit, and three shared engines, evidence discipline, a persona engine and output standards, apply consistently across all of them. A formal audit of the early method rated each component’s strength against the same evidence discipline the Labs themselves apply, treating the method’s own credibility as something to be earned rather than assumed.
Prove the discipline holds under real pressure, not just clean examples
Before building any software on top of the method, it was run against genuinely difficult validation material designed to expose weaknesses: multi-year projected outcomes, contested authority-drift patterns, and qualitative organisational shifts with no natural numeric threshold. The audit found that early material had, in places, argued with more numerical confidence than the evidence supported, a finding that directly shaped the rule now built into the method: a projected figure is presented as a qualitative band with its basis stated, never as a false, precise percentage.
A schema settled through deliberate stress-testing, not assumed correct
Turning a finished analysis into something trackable required a structured data format capable of representing genuine uncertainty, not just clean numbers. That format was deliberately stress-tested against the hardest material in the method’s own validation set before any ingestion logic was written, and it produced a small but important discovery: projected trajectories and qualitative organisational shifts are structurally different things and needed two separate artefact types, not one convenient merge.
A tracked record, not a terminal report
With the data contract settled, a purpose-built application was built to ingest a completed analysis and let it be interrogated over time: reviewed as a baseline, tested with what-if changes, and eventually monitored on a schedule. Every trackable finding carries its original position, its current position, and a full movement history, governed by a single enforced rule: confidence in how far a finding has moved can never exceed the strength of the evidence that moved it. That rule is the practice’s discipline against watering down uncertainty over time, expressed as one mechanical, code-checkable guarantee rather than good intentions.
Evidence discipline
The spine of the method, and the reason it survives scrutiny.
Every substantive finding, in every Lab, carries an explicit rating and is checked against the same self-check before it is delivered. This is the same discipline Changeable brings to client AI governance work: confidence that is earned, not asserted.
What this produced
A method that survives scrutiny, and a practice built to scale it
The result is a decision intelligence practice with a defensible method at its core, not just a consulting brand.
A versioned, audited method
The method carries an explicit version history, with every change to the Labs or engines recorded and justified.
Stakeholder analysis that surfaces hidden cost
A dedicated persona engine separates support for a goal from support for a specific design, the distinction most consulting analysis collapses.
Production-grade, consistent delivery
A fully specified house style means every deliverable is visually and structurally consistent, regardless of subject matter.
Analysis that can be tracked, not just filed
A completed engagement can be interrogated with what-if scenarios and monitored over time through the living-analysis engine.
A defined, scalable engagement model
Small, medium and large engagement tiers with co-principal delivery, so the method is not locked inside one person’s head.
Genuinely bicultural practice
Kaupapa Māori methodology is a built-in activation, not a marketing layer, applied wherever an engagement’s consequences call for it.
Questions
Questions about how Pūtake Labs was built?
Common questions about the approach behind Pūtake Labs’ decision intelligence method.
Why build the method before the software?
The credibility of any tracking application depends entirely on the quality of the analysis it is tracking. Proving the method’s evidence discipline against genuinely hard material first meant the application was built on a settled, audited foundation rather than being reverse-engineered from whatever the method happened to produce.
What makes the evidence discipline different from typical consulting rigour?
Every substantive finding carries an explicit High, Moderate or Low rating, findings and hypotheses are never allowed to blur together, and projected figures are represented as honest qualitative bands rather than false, precise percentages. It is enforced by a self-check applied before any finding is delivered, not left to good intentions.
Why does a decision analysis need to be tracked over time at all?
Most consulting reports are treated as finished the day they are delivered, even though the assumptions behind a five-year projection can change within months. A tracked, living record lets a client see whether the assumptions still hold and stress-test the analysis against new information.
How does the movement-confidence rule actually work?
Every trackable finding carries an original position, a current position and a full movement history. A hard rule, enforced in code rather than asserted in prose, means the confidence recorded for any movement can never exceed the strength of the evidence that drove it.
What does the Kaupapa Methodology Module actually do?
It is activated wherever an engagement involves Māori organisations, communities, data, mātauranga Māori, Te Tiriti obligations or significant consequences for Māori, applying particular evidence-discipline care to how those interests are represented. It is a built-in activation triggered by an engagement’s substance, not an optional add-on.
Does every engagement use all eight Labs?
No. Part of the method’s discipline is explicitly declining Labs that do not fit a given engagement, and naming which ones were set aside and why, rather than applying every framework to every problem regardless of fit.
Can Changeable help design a similarly rigorous method for our organisation?
Yes. Pūtake Labs reflects the same evidence-led, governed approach Changeable applies in client engagements, including use case development, AI data modelling and AI governance.
Want to see how Pūtake Labs approaches a decision?
Explore Pūtake Labs directly, or talk to Changeable about applying the same evidence-first, governed approach to your own AI product or decision process.