A gated roadmap turns the first line into a learning system

The baseline gate should establish whether the factory can reproduce its own account of a shift. This requirement follows a real measurement problem: NIST's monitoring workshop report describes differing OEE definitions and calculations. A sponsor should therefore ask what each measure includes before accepting a comparison between lines or periods.[S03]

GFT recommends a signed measurement dictionary owned jointly by production, quality and industrial engineering. It should identify the line boundary, style, operation bulletin, reporting interval, time exclusions and responsible recorder. Define good output at an explicit acceptance point. Separate first-pass acceptance from eventual acceptance after repair. Define how reworked garments re-enter the record so they cannot earn output twice. Keep rejects, repair activity and goods awaiting judgement visible even when the headline output measure excludes them.

Declare planned production time before the shift starts. Record revisions with a reason and an approver rather than silently changing the denominator after an interruption. Distinguish the event that stopped production from the department expected to resolve it. An outage, a missing component and an equipment breakdown may all stop sewing, but the proposed response and accountable role differ. Maintain an unresolved category when the cause is unknown; a forced explanation makes the loss tree less credible.

Keep earned standard minutes separate from OEE. The Better Work assessment defines garment standard minute value as the sum of the standards for its constituent operations. GFT recommends retaining the approved standard and its version with every style record, then calculating earned minutes from accepted output under that version. A revised standard should start a visible new comparison series or a disclosed restatement, rather than manufacture an apparent gain.[S13]

Start the capture routine with an hourly line sheet and a shift reconciliation. The proposed sheet records accepted output, first-pass failures, repair movements, downtime by cause, constraint output and WIP by position. Record missing intervals explicitly. Calculate completeness from expected records and usable records, while separately tracking unresolved quantities and causes. A completed form with implausible values is not a clean baseline. The line owner should sample the physical quantities and event times against the sheet before approving the shift.

For the proposed baseline gate, define a representative observation window before collecting the trial data. Include the operating conditions the next commitment must withstand, such as ordinary staffing, replenishment and a style transition. Set acceptable completeness, reconciliation error and unresolved-loss limits in advance, using the decision's sensitivity and measurement capability. There is no universal threshold in this chapter. A missing observation that could reverse the release decision is a reason to hold, even if average completeness looks satisfactory.

The acceptance pack should contain the dictionary, raw sheets, corrections, reconciliation and unresolved questions. Production should demonstrate that another authorised reviewer can reconstruct the published result. If that fails, keep the manual routine, correct the definitions and repeat the baseline. Do not fund broader monitoring simply to accelerate an ambiguous calculation. The first learning outcome is a shared account of what happened, with enough provenance to challenge it.

GFT also recommends recording measurement resolution. A stop observed only at a periodic walk should retain its timing uncertainty rather than acquire a precise duration during transcription. Compare the uncertainty with the improvement being claimed. If plausible recording error could explain the apparent gain, improve the observation method or extend the test. This makes the baseline gate a test of decision usefulness as well as administrative completeness.

The next gate should test a feasible operating method at the point limiting good throughput. The Lean Enterprise Institute describes standardised work through takt, work sequence and required in-process inventory, and treats it as a revisable basis for improvement. Its account of constraints also prioritises the limiting link before major expansion. These support a trial sequence, not a guarantee of improvement in a particular garment line.[S26][S30]

GFT recommends that the industrial engineer and supervisor walk the actual material route with operators and quality staff. Compare observed work with the operation bulletin, machine requirements and skill assignments. Locate persistent queues, starvation and blocking, then check whether the suspected constraint limits accepted output at the line boundary. A queue alone is not sufficient for this proposed gate: it may reflect a temporary replenishment surge, inspection hold or unsuitable release rule. Record the competing explanations before choosing an intervention.

The trial instruction should state the change, its mechanism and its limits. A proposed reallocation must preserve operation precedence, machine suitability, safe reach and demonstrated competence. Name who authorises a reassignment and who can stop it. Retain the prior layout and allocation as a recoverable configuration. Where the proposed balance depends on a specialist operator, test whether the result survives normal cover arrangements before using it as the basis for another line.

Choose the comparison method before looking at the outcome. Use matched style and staffing conditions where practical; otherwise disclose the changes in style complexity, attendance, hours, material quality and interruptions. Retain interval results as well as totals, since a strong finish may conceal an unstable start. GFT recommends recording intervention time and supervisory support separately. A pilot sustained by additional troubleshooting effort may still be promising, but that effort belongs in the scale decision and operating cost.

Set the acceptance test around good throughput sustained within agreed quality and safety conditions. Compare first-pass yield, repair workload, WIP, overtime and operator feedback alongside output. Define what deterioration would trigger a hold before the trial begins. A serious safety concern should stop the affected method immediately; it should not wait for the next commercial review. If output rises while downstream inspection accumulates, the trial has not yet demonstrated a system gain.

The proposed action register in Exhibit 66 assigns accountability without inventing an organisation chart. The sponsor would appoint these roles and their deputies before the trial. Its entries are an illustrative governance design, not evidence that GFT has delivered the actions.

The register leaves approval with the sponsor and technical acceptance with the relevant leads. An owner should report a failed condition directly, with the evidence needed to repeat the test. Releasing the next commitment should require explicit closure or a narrower scope that genuinely removes the dependency. Reassigning the issue to a different department does not close it.

Before the trial, GFT recommends stating what would disprove the proposed mechanism. If reallocating work is expected to relieve the constraint, the record should show both the changed station condition and the resulting accepted line output. If the station improves while output remains unchanged, investigate the next limiting condition. That finding may justify another bounded trial, but it does not justify recognising the original forecast as an achieved benefit.