Problem Solving & Quality · Act

Pilot Run with Go/No-Go

Try the countermeasure on one line, one shift or one tank, with success criteria and a rollback plan written down before the start.

Time2 h
FormatSmall group
StageAct

Pilot Run with Go/No-Go: what it is and why it works

A pilot run tests a countermeasure on a limited, representative scope, such as one line, one shift, one tank or one production campaign, before committing to a full roll-out. What distinguishes a structured pilot from simply trying something is what is written before the start: the indicator and threshold that define success, the stop criteria that trigger an early end, the rollback procedure, and who takes the go, adjust or no-go decision. The pilot then runs long enough to see normal variation, and every unusual event is logged.

Pilots are worth their cost when a change is expensive, slow or risky to undo, or when the benefit is uncertain because it depends on real conditions such as raw material variability, operator habits or seasonal effects. Writing criteria in advance protects against the natural tendency to declare success after the fact. Running a comparable line or unit without the change at the same time gives a control group, which separates the effect of the fix from background noise. The pilot is the "Do" of a PDCA cycle at small scale, it inherits stop criteria and contingencies from a potential problem analysis, and its data become the baseline for the effectiveness check after roll-out.

What you need

The countermeasure to test and the verified cause it addresses
A representative but limited pilot scope and, if possible, a comparable control scope
Success and stop criteria with indicators and thresholds
A rollback plan and a named decision maker
Baseline data for the indicator, collected with the same measurement method

What you get

Pilot data compared with the criteria and with the control scope
A log of unexpected events and side effects
A documented go, adjust or no-go decision
A roll-out plan or the next option to test

When to use it

When a plant-wide change would be expensive to undo if it fails.

How to do it, step by step

Choose a pilot scope that is representative but limited: one line, one shift, one filter, one campaign.
Write the success criteria and the stop criteria before starting, with the indicator and its threshold.
Prepare the rollback: how to return to the previous state, who decides, how fast.
Run the pilot for long enough to see normal variation, and log everything unusual.
Decide go, adjust or no-go against the written criteria — then plan the roll-out or the next option.

Worked example: Trialing a new collector in a flotation circuit

Illustrative scenario — figures are realistic but not from a real company.

A zinc concentrator runs two parallel flotation lines fed from the same grinding circuit. Zinc recovery averaged 86 %, and the metallurgy team wanted to test a new collector that promised two to three points more recovery. A full switch would require changes to reagent storage and could not be quickly reversed.

Line A was chosen as the pilot and line B as the control. The pilot was set for four weeks, long enough to cover several ore blends from the pit.
Success criterion: recovery on line A at least 1.5 points above line B over the period, with concentrate grade not more than 0.5 points lower. Stop criteria: any froth instability requiring more than two operator interventions per shift, or a grade drop of more than one point over three days.
Rollback was prepared: a temporary dosing skid could be switched back to the old collector within one shift, and the metallurgy superintendent had the decision authority.
During week two, a high-clay ore blend caused frothing problems on both lines, which the control line showed was not related to the new reagent. Samples were taken twice per shift from both lines.

Result. Over four weeks, line A recovered 1.9 points more zinc than line B with no significant difference in grade. The decision was go, with a planned roll-out to line B and a permanent reagent system. Without the control line, the clay week would probably have been blamed on the new collector and the pilot stopped early.

Open it in the FieldKit Problem Solving app → Get the Problem Solving template pack — join the waitlist

Common pitfalls and how to avoid them

Defining success after seeing the data.Write success and stop criteria, with thresholds, before the pilot starts, and have them approved by the decision maker.
Choosing a pilot scope that is unrepresentative, such as the best crew or the newest machine.Pick a scope with normal conditions, or explain and compensate for differences in the analysis.
Running the pilot for a few days and extrapolating.Run long enough to capture normal variation: shifts, raw material lots, weekly cycles.
Starting without a tested rollback path.Prepare and, if possible, rehearse the return to the previous state, and name who can order it.

Frequently asked questions

How long should a pilot run last?

Long enough to see the normal variation that could mask or mimic the effect: typically several shifts, raw material lots or production cycles. For a daily indicator, two to four weeks is common; for a batch process, a set number of batches. A quick statistical estimate of the sample size needed to detect the expected improvement is a good way to set the minimum duration.

What are go/no-go criteria?

Go/no-go criteria are measurable conditions, agreed before a pilot starts, that decide whether to roll out a change. They usually include a success threshold on the target indicator, limits on side effects such as quality, safety or cost, and stop criteria that end the pilot early. Writing them in advance prevents results being interpreted to fit expectations.

Why use a control group in an industrial pilot?

Because many things change during a pilot besides the countermeasure: raw materials, weather, product mix, staffing. A comparable line, shift or unit running without the change at the same time experiences the same background changes. Comparing pilot and control isolates the effect of the fix. Without a control, a lucky or unlucky period can easily be mistaken for the effect of the change.

Origin

Pilot testing with predefined success and stop criteria — common engineering and change-management practice; no single author.

Used in these playbooks

One-page improvement cycle 3 weeks

Three weeks for a local problem owned by a team leader: see it on site, tell it on one A3, choose the countermeasure openly, pilot it and make it the new standard.

Related methods

More in “Act”

Plan, pilot and run the fix — with owners, dates and a structured roadmap.