On/Off Cause Verification

Prove the suspected cause by switching it on and off under control: the defect must appear and disappear with it.

Time1 h 30
FormatSmall group
StageRoot Causes

On/Off Cause Verification: what it is and why it works

On/off cause verification proves a suspected cause by switching it on and off under controlled conditions: if it really is the cause, the defect must appear when it is present and disappear when it is removed. The hypothesis is written as a testable prediction, and a trial is planned that changes only the suspected factor while everything else is held constant, within agreed safety limits. The switch is repeated enough times to rule out chance. If the defect does not follow the switch, the hypothesis is rejected and the team returns to its candidate list.

Many corrective actions fail because they were built on a cause that was plausible but never demonstrated. Correlation, expert opinion and a convincing story are not proof; making a problem come and go on demand is about as close to proof as industrial practice gets. The logic is that of the controlled experiment: vary one factor, hold the others constant, replicate. It is the standard of evidence expected in 8D discipline D4. When the live process is too risky or costly, the switch can be made on a pilot rig, in the lab, or by swapping a suspect component into a good assembly. For several interacting factors, a designed experiment extends the same idea. Change analysis and scatter diagrams supply the candidates; a pilot run later confirms the countermeasure at scale.

What you need

A candidate cause from earlier analysis, written as a prediction
A safe way to control the suspected factor
A measurement method that detects the defect reliably
Agreement on safety limits, trial duration and disposition of trial product
Support from production, engineering and quality

What you get

A documented trial plan and the conditions actually run
Results for each on and off condition, repeated
A clear decision: cause confirmed or rejected
An estimate of the size of the effect, useful for designing the fix

When to use it

When a cause is “obvious” to everyone but nobody has demonstrated it.

How to do it, step by step

State the hypothesis as a prediction: “if feed temperature drops below 60 °C, deposits appear within two hours”.
Plan a controlled trial: change only the suspected factor, keep the others constant, agree the safety limits.
Switch the cause on and confirm the defect appears; switch it off and confirm it disappears.
Repeat enough times to rule out chance, and record every condition.
If the defect does not follow the switch, reject the hypothesis honestly and return to the candidate list.

Worked example: Seal leaks on snack-food pouches

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

A snack-food plant sees seal leaks on stand-up pouches rise from 0.2% to 1.5% on one vertical form-fill-seal machine. Change analysis points to a new film lot with a thinner sealant layer, yet the supplier's certificate shows the film within specification.

Hypothesis written as a prediction: with the new film lot at current jaw settings (320 °F, 0.8 s dwell), leaks will exceed 1%; with the previous lot, they will stay below 0.3%.
Plan: same machine, product, crew and settings; alternate the two film lots every 30 minutes over two shifts; leak-test a 200-pouch sample from each run; hold all trial product until results are in.
Results: six runs on the new lot gave 1.2% to 1.8% leaks, and six runs on the old lot gave 0.1% to 0.3%. The defect appeared and disappeared with the film every time.
A second switch tested the proposed fix: the new film with jaw temperature raised to 335 °F, inside the supplier's recommended sealing window, gave 0.2% leaks over three runs.

Result. The cause was confirmed as an interaction between the thinner sealant layer and a jaw temperature near the low edge of the sealing window. The plant adopted the new setting, asked the supplier to tighten the sealant-layer tolerance, and added a seal-strength check at the start of every film lot. Without the trial, it would probably have rejected film that met its specification.

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

Common pitfalls and how to avoid them

Changing several factors at once.Change only the suspected factor; if something else has to move, record it and repeat the trial.
Switching on and off only once.Repeat the switch several times, ideally in alternating order, so that chance and drift cannot explain the result.
An unreliable defect measurement.Confirm that the test method detects the defect consistently before running the trial.
Explaining away a failed prediction.If the defect does not follow the switch, reject the hypothesis and go back to the candidates rather than adjusting the story.

Frequently asked questions

How do you verify a root cause?

Write the suspected cause as a prediction, then show that the problem appears when the cause is present and disappears when it is removed, under controlled conditions and more than once. Where live testing is unsafe or too costly, use a pilot rig, lab tests or component swaps. Supporting evidence, such as timing, location and physical findings, should also be consistent with the cause.

What is the difference between correlation and root cause verification?

Correlation shows that two variables move together in existing data; it does not show that one causes the other, because a third factor may drive both. Root cause verification actively changes the suspected factor while holding the others constant and checks the effect. That deliberate, controlled change is what turns a correlation into evidence of cause.

Is on/off verification always possible?

No. Some causes cannot be recreated safely, such as a fire, a structural failure or a toxic release, and others are too costly to reproduce. In those cases, use lab simulation, failure analysis of preserved evidence, physical models or engineering calculation, and state clearly how much confidence has been reached. Never create an unsafe condition to prove a cause.

Origin

Root-cause verification — 8D discipline D4 practice; logic of controlled experiments (R. A. Fisher, The Design of Experiments, 1935).

Used in these playbooks

Chronic scrap reduction month 1 month

One month against a loss everybody has learned to live with: count at the source, rank the losses, split the data, list the causes and prove the real one before spending money.

Related methods

More in “Root Causes”

List the possible causes, dig down the chains and prove the real one before acting.