Skip to content
GREEN FUTURETECH.
Explore

GFT / How we work

Evidence earns
the next phase.

Diagnose the constraint. Design the operating system. Deploy against gates. Prove value in operation.

Operate the phase console
Scrub the phase

Gate state The material, market, operating and infrastructure questions have evidence, an owner and a stated uncertainty.A team can operate the design with the available skills, power, connectivity and fallback procedures.The record is dependable, the routine is used and the application clears its operating test.The comparison is like for like, the denominator is intact and the result does not transfer harm elsewhere.

Enter the phase orbitProcedural visualisation. Schematic, not a facility.

Diagnose, Design, Deploy and Prove are the public editorial sequence GFT uses to explain its programme method. Verification runs through all four, and no phase promises a fixed delivery duration.

A connected working sequence

Each phase ends with evidence for the next commitment.

Move the phase cursor
Diagnose

Establish what is true before drawing the solution.

The work begins with the material, market, operating and infrastructure conditions that could change the answer. Observations, estimates and policy targets remain separate so the baseline can be challenged without breaking the analysis.

Decision
Which constraint actually binds the decision?
Evidence gate
The material, market, operating and infrastructure questions have evidence, an owner and a stated uncertainty.
  • Dated baseline
  • Assumptions register
  • Decision options
  • Conditions precedent
Design

Make the operation and its evidence contract explicit.

The operating scope, measurement definitions, owners, source records and review routines are specified together. Architecture follows the decision and the actual constraints, rather than arriving as a predetermined stack.

Decision
What is the smallest useful operating system?
Evidence gate
A team can operate the design with the available skills, power, connectivity and fallback procedures.
  • Operating model
  • Measurement definitions
  • Source records
  • Owners
  • Routines
  • Proportionate architecture
Deploy

Install discipline before complexity.

A bounded implementation establishes clean capture and an active management routine before more complex applications are introduced. Local validation and a usable fallback remain part of the working state.

Decision
Can the routine produce a reliable operating history?
Evidence gate
The record is dependable, the routine is used and the application clears its operating test.
  • Clean capture
  • Working review routines
  • Trained owners
  • Bounded applications with local validation
Prove

Test movement, attribution and unintended effects.

The result is compared with the agreed baseline and operating conditions. Quality, delivery, labour and resource guardrails stay beside the intended benefit, and the evidence can justify adaptation or a stop as well as scale.

Decision
Does the evidence justify scale, adaptation or a stop?
Evidence gate
The comparison is like for like, the denominator is intact and the result does not transfer harm elsewhere.
  • Benefits register
  • Guardrail review
  • Corrective-action record
  • Next decision
Phase
01 / 0402 / 0403 / 0404 / 04
Gates cleared before
0 of 41 of 42 of 43 of 4
Working outputs
4 listed6 listed4 listed4 listed

Diagnose

The baseline must be strong enough to change the answer.

Diagnosis establishes what is known, what is estimated and what could make the venture or intervention fail. It makes uncertainty usable instead of hiding it inside a single recommendation.

Choose an evidence strand
Strand 1 of 6

Material supply

Seed cotton, fibre and finished output are dated and kept apart. A missing input becomes a research task, not a silent default.

Question
What is the dated material baseline?
Condition
Seed cotton, fibre and finished output are never mixed.
Strand 2 of 6

Demand, trade access and competition

Observations, estimates and policy targets stay separate, so the baseline can be challenged without breaking the analysis.

Question
What do demand and trade access support?
Condition
A target is never read as an observation.
Strand 3 of 6

Workforce and supervision

Trained supervision is a condition to resolve, not an assumption. A positive feasibility answer can still depend on it.

Question
Who will run the routine, and is that supervision trained?
Condition
A named, trained owner exists before the design is relied on.
Strand 4 of 6

Unit economics and risk

Assumptions sit beside the decisions they could change. Uncertainty stays usable instead of hiding inside one recommendation.

Question
Which assumption could change the answer?
Condition
The assumption is recorded beside its decision.
Strand 5 of 6

Power, connectivity and maintenance

Infrastructure conditions are recorded because a feasibility answer can depend on them, and a fallback keeps the routine alive.

Question
Do power, connectivity and maintenance support the routine?
Condition
A usable fallback is defined.
Strand 6 of 6

Site and logistics

Site and logistics inputs carry a source, a date and an owner, and travel with the recommendation into design.

Question
Do the site and logistics inputs hold at the stated date?
Condition
Each input has a source, a date and an owner.

Each strand has a source, a date and an owner. The output is a set of decision options with conditions precedent: a positive feasibility answer can still depend on site supply, trained supervision, a working quality record or a reliable operating routine, so the condition travels with the recommendation into design. A stop option is retained when the evidence does not support commitment.

1Material2Market3People4Economics5Infrastructure6LogisticsDATEDBASELINE
Conceptual diagnostic map. The strands are investigation areas, not measured scores.

Design

The operating model and the evidence contract are one design.

A process flow alone does not tell a team how to control the operation. Design specifies the work, the record and the response together, then chooses an architecture that can carry them.

Open a clause of the contract
Clause 1 of 5

Operating scope

Define the product, process steps, accepted-output boundary and explicit exclusions. Upstream or partnered operations remain distinct from the base operating commitment.

Test
A named team can say what is in scope and what is not.
Clause 2 of 5

Measurement

Give each measure a formula, unit, time window, source record and data-quality rule. Include resource intensity beside output, quality, delivery and cost.

Test
The same figure can be recomputed from the same record.
Clause 3 of 5

Ownership

Name who records, who reviews and who can correct the condition. Design the shift handover so an exception does not lose its owner between teams.

Test
No exception loses its owner at a handover.
Clause 4 of 5

Review routine

Connect line-level events to daily control, weekly problem solving and monthly decisions. Specify the escalation path before the first exception arrives.

Test
The escalation path exists before the first exception.
Clause 5 of 5

Architecture

Choose capture, validation, storage and analysis around power, connectivity, maintenance and available history. Keep a fallback that preserves the operating routine.

Test
The routine still runs when the architecture fails.

Design output: an operable contract, not only a recommendation. The test is whether a named team can run it under the stated conditions.

1Scope2Measurement3Ownership4Review routine5ArchitectureOPERABLECONTRACT

Deploy

Install discipline before complexity.

Deployment creates a reliable operating history. The sequence below is a logic of readiness, not a universal timetable. A later state begins only when the earlier one is used and understood.

01 / 04 · Clean capture

Make the first record dependable.

Train the people closest to the work, agree definitions and begin a controlled source record. A manual performance system can be the right starting point. It exposes ambiguous events and missing fields before those problems are automated into a larger architecture.

Gate: The team can record consistently, explain gaps and reconcile the source with the work.
02 / 04 · Active management

Use the record in the operating routine.

Daily and weekly reviews turn the source into action. Supervisors investigate losses, assign corrective work and verify that the action closed. The management routine must be real before a more complex application can claim that its output will be used.

Gate: Reviews happen, owners act and corrective work leaves a verifiable history.
03 / 04 · Validated application

Prove the tool under local conditions.

Introduce a bounded, proven use case with an acceptance test on the actual material, product and operating state. Keep the fallback running until the new method is dependable. History-dependent applications wait for enough governed history and must beat a simple baseline.

Gate: Local acceptance clears, the owner understands the result and the fallback remains usable.
04 / 04 · Evidence-gated scale

Let the record justify the next commitment.

Compare the new state with the agreed baseline and inspect quality, delivery, labour and resource guardrails. Scale only when the result and the operating routine hold. Otherwise adapt the design or stop the use case without disguising the outcome.

Gate: The benefit is attributable, the guardrails hold and the next commitment is justified.

Account in view01 / 05 Accepted output02 / 05 Delivery03 / 05 Cost04 / 05 People05 / 05 ResourcesBaseline question onlyComparable boundary applied

AcceptedoutputDeliveryCostPeopleResourcesINTENDED EFFECTCOMPARABLE BOUNDARY APPLIEDBASELINE QUESTION ONLYGUARDRAIL123
  1. 1Keep the record whole. The register holds baseline, intervention, result and operating conditions together.Inside one comparable boundary, failures are recorded as well as gains.
  2. 2Keep doubt in the wording. Where attribution is uncertain, the wording stays uncertain and the next test separates the candidate causes.
  3. 3Test that it repeats. A short gain from exceptional supervision differs from a routine the normal shift can repeat.
Conceptual comparison. The lines explain a review logic and do not represent measured results.

Prove

Proof tests the mechanism, not only the movement.

A better number is the beginning of an investigation. The team still needs to know why it moved, whether the comparison is fair and whether an apparent gain transferred a loss into another account.

Comparison boundary
Select an account in the benefits register
Accepted output
Baseline question
Was the same acceptance rule and product boundary used?
Evidence for the next decision
The gain persists under comparable operating conditions.
Guardrail
Quality and rework remain visible.
Delivery
Baseline question
Did service change because of the intervention or another condition?
Evidence for the next decision
The mechanism explains the service movement.
Guardrail
Inventory and expediting are not hiding a new loss.
Cost
Baseline question
Which inputs, volumes and one-off costs changed?
Evidence for the next decision
The unit result reconciles to the operating assumptions.
Guardrail
Savings are not created by deferring necessary work.
People
Baseline question
Did the routine change supervision, skills or workload?
Evidence for the next decision
The normal shift can sustain the new routine.
Guardrail
Volume is not rewarded by transferring harm to the team.
Resources
Baseline question
Are water, energy and material compared per accepted unit?
Evidence for the next decision
The improvement holds inside the same system boundary.
Guardrail
Product mix and estimation rules remain explicit.

Management rhythm

The review cadence keeps the method honest.

Each cadence answers a different question. Events are controlled near the work, recurring causes are examined across shifts, and larger commitments are reviewed with a wider evidence frame.

Daily control

What changed at the line, what needs an owner now and which exception must be contained?

The source record stays close to the event and the shift handover keeps the action alive. Control is exercised where the work happens, before an exception becomes a pattern.

Escalation
An owner is named for each open exception.

Weekly problem solving

Which losses repeat, which corrective actions remain open and what does the evidence say about the cause?

The review tests whether earlier actions actually closed the gap. Recurring causes are examined across shifts rather than one event at a time.

Escalation
An action that did not close the gap returns to the list.

Monthly decisions

Does the operating evidence justify scale, redesign or a stop?

Product, investment, workforce and resource questions are reviewed together rather than as separate reporting silos, with a wider evidence frame than the weekly review.

Escalation
A commitment is scaled, redesigned or stopped on the record.
Illustrative review items
  • What changed at the line
  • Which exception needs an owner now
  • Exception contained
  • Handover keeps the action alive
  • Source record stays close to the event
  • Owner named for each open item
  • Which losses repeat
  • Which corrective actions remain open
  • Did the earlier action close the gap
  • What the evidence says about the cause
  • Recurring causes examined across shifts
  • Open action tested against the record
  • Scale, redesign or stop
  • Does the evidence justify the next commitment
  • Product and investment reviewed together
  • Workforce and resource questions in one review
  • No separate reporting silos
  • A wider evidence frame

Gate decision

A gate can say adapt or stop.

Evidence gates are useful only when they can change the commitment. If every phase leads automatically to scale, the gate is theatre and the review has lost its purpose.

Decision shown: Adapt.

Illustrative scale from 0 to 100, for explanation only. The thresholds explain the logic and are not GFT criteria. A breached guardrail caps the evidence that counts.

Withdraw commitment

Stop

The use case fails, the evidence cannot support the spend or the result transfers harm elsewhere. Stopping preserves the learning without turning sunk effort into a reason to continue.

Change the design

Adapt

The question remains valuable but the mechanism, definition, training or architecture needs work. The revision and the next comparison are made explicit.

Commit further

Scale

The comparison supports the intended effect, guardrails hold and the normal team can operate the new state. The next commitment names its owner and its own evidence test.

A working denominator

OEE shows how one result becomes three different questions.

Overall equipment effectiveness separates whether the process was available, whether it ran at the intended rate and whether the output was good. The point is not a universal target. It is the diagnostic separation.

Operate the full OEE model

Illustrative shift, not a measured line

Time85%
Rate90%
Good output95%
OEE72.7%
  • Good output72.7
  • Quality loss3.8
  • Rate loss8.5
  • Time loss15.0

Percentage points of planned time. The default values are illustrative, not a measured factory or an industry benchmark. The loss that last moved is lit.

Continue through the method

Put the phase cursor on your own decision.

The principles explain the gates. The approach page explains the measurement crossing. The company page places the method inside GFT's known three-part programme.

Where is your decision in the sequence?

Related research register

Measure Before You Model

Related research register

Edge to Dashboard

Related research register

Utilisation Decides

Related research register

Utilisation Decides

The pairing of phases with papers is editorial. The papers develop the performance, utilisation and factory-architecture arguments.

Find a thread.

Type to search.