A control plan is not a checklist. It is not a list of dimensions copied from a drawing into a table format. It is not a document produced to satisfy a customer audit requirement and then filed in a folder that no one references during production. A control plan that functions as its name suggests — that actually controls the quality of the product it covers — is a live process management document that identifies the characteristics most likely to fail, specifies the measurement method that will reliably detect that failure, defines the response that will prevent a non-conforming casting from being shipped, and is actively used by operators and inspectors during production rather than existing only as an audit artefact.
The gap between these two versions of a control plan — the paper document and the working quality tool — is the gap between a foundry that produces consistent quality and one that produces variable quality that occasionally passes inspection. Building the working version rather than the paper version requires understanding what a control plan is supposed to achieve and what decisions need to be made to achieve it.
Selecting the Characteristics to Control — The Most Important Decision
The first and most consequential decision in building a casting control plan is which characteristics to control — and more specifically, which characteristics require active in-process monitoring rather than final inspection verification. Not all dimensions on a casting drawing carry equal quality risk, and a control plan that attempts to monitor every dimension with equal intensity is a plan that will not be followed in production because the measurement burden it creates exceeds what the inspection function can practically deliver in a production environment.
The characteristics that warrant active in-process monitoring are those that meet two criteria simultaneously: they are functionally significant — their deviation from specification would cause assembly problems, functional failure, or customer rejection — and they are process-sensitive — they show meaningful variation from casting to casting that cannot be assumed to be controlled by die design and process setup alone. A bore that locates a bearing — functionally significant — that varies by 0.3 millimetres across the die's production life as the die wears — process-sensitive — belongs in the control plan as an in-process monitored characteristic with a defined measurement frequency and a response plan for out-of-tolerance results.
A face-to-face dimension that is controlled by the die's parting surface geometry — which changes very slowly with die wear — and that has a tolerance of plus or minus two millimetres — wider than the die's dimensional variation — does not require active in-process monitoring. It is adequately controlled by periodic verification that the die's parting surfaces are within specification, and by final inspection verification on a sampling basis rather than 100 percent inspection of every casting. Including this characteristic in the in-process monitoring plan alongside the bearing bore would dilute the monitoring resources applied to the genuinely critical characteristic and make the control plan harder to follow in production without improving quality outcomes.
The customer's drawing identifies which characteristics are special — marked with customer-specific symbols indicating elevated functional significance. These special characteristics are the mandatory entries in the control plan's in-process monitoring section; every special characteristic requires a defined measurement method, measurement frequency, and response plan. Beyond the customer-identified specials, the foundry's own process knowledge — which characteristics have historically shown the most variation, which dimensions are most sensitive to the specific process parameters that drift in production — should inform the selection of additional characteristics for active monitoring.
Measurement Method Selection — Matching the Tool to the Requirement
The measurement method specified in the control plan must be capable of detecting the variation that matters for each controlled characteristic — and this requires matching the measurement tool's resolution, accuracy, and repeatability to the tolerance of the characteristic being measured. A rule of thumb widely used in quality engineering is that the measurement system should have a resolution of at least one tenth of the characteristic's tolerance — if the tolerance is plus or minus 0.25 millimetres, the measurement tool needs resolution of at least 0.025 millimetres to be reliably useful.
For most casting dimensions with tolerances in the range of plus or minus 0.3 to plus or minus 1.0 millimetres, digital verniers or digital micrometers with 0.01 millimetre resolution are adequate measurement tools if they are correctly calibrated and used by trained inspectors applying consistent technique. For tighter tolerances — bearing bores to H7 tolerance, mating faces with flatness requirements below 0.1 millimetres — dedicated gauges, bore gauges, or CMM measurement are required to achieve the measurement resolution and repeatability that the tighter tolerance demands.
The measurement method must also be practical for the production environment where it will be used. A control plan that specifies CMM measurement for a characteristic that must be checked on every casting — at a production rate of twenty castings per hour — will not be followed because CMM throughput cannot match production throughput. The practical solution for a critical characteristic that requires tight dimensional control and frequent measurement is a dedicated hard gauge — a go/no-go gauge or a dedicated snap gauge — that provides pass/fail verification at the required frequency without the throughput limitation of CMM measurement. The CMM is then used for periodic capability verification and for investigating borderline results rather than for 100 percent inspection of every casting.
Measurement system analysis — the statistical evaluation of whether the measurement system is adequate for the characteristic it is measuring — is specified in IATF 16949 and advanced quality planning requirements for automotive casting supply. A gauge repeatability and reproducibility study — where multiple operators measure multiple castings multiple times with the same gauge — quantifies the proportion of the characteristic's tolerance band that is consumed by measurement system variation. A measurement system whose variation consumes more than 30 percent of the tolerance band is marginal; more than 10 percent is the target for critical characteristics. A control plan that specifies a measurement method without verifying that the measurement system is adequate for the tolerance it is controlling is not a controlled process — it is a process that produces measurement data without confirming that the data is reliable.
Sampling Plans — How Often to Measure and What to Do With the Results
The measurement frequency in a control plan — how many castings are measured per hour, per batch, or per production run — reflects a judgement about how frequently the controlled characteristic is likely to shift outside its specified range during production, and how many non-conforming castings are acceptable between measurements. A characteristic that shifts slowly and predictably — controlled by a die dimension that wears gradually over thousands of cycles — can be monitored at lower frequency than a characteristic that shifts with every alloy charge or every die opening and closing cycle.
The statistical framework for sampling plan design — acceptance sampling tables, statistical process control charts, and process capability indices — provides a principled basis for sampling frequency decisions. A process with a Cpk of 1.67 on a critical characteristic — meaning its natural variation uses only 60 percent of the tolerance band with adequate margin on both sides — can be monitored at lower frequency than a process with a Cpk of 1.0 — where the variation is using the full tolerance band and any drift will immediately produce non-conforming product. Higher process capability justifies less frequent monitoring; lower capability requires more frequent monitoring to catch deterioration before non-conforming castings accumulate.
The response plan — what happens when a measurement result falls outside the specified control limit — is the element of the control plan that most directly determines whether the plan prevents non-conforming castings from reaching the customer. A control plan whose response plan says "inform supervisor" without specifying what the supervisor does, or that has no response plan at all, will not consistently prevent non-conforming castings from being shipped. A control plan whose response plan specifies: stop production, quarantine castings produced since the last acceptable measurement, identify and correct the root cause before resuming production, and inspect all quarantined castings individually before release — this plan will prevent the non-conforming castings from shipping, at the cost of production disruption that is far less expensive than the customer complaint that would result from shipping them.
Making the Control Plan a Working Document
The control plan that exists as a file on a computer or as a printed document in a quality folder does not control quality. The control plan that is displayed at the relevant workstation, understood by the operator and inspector who use it, has its measurement requirements incorporated into the daily work routine, and has its response procedures known and followed when out-of-tolerance results occur — this plan controls quality. The difference between these two is not the content of the document but the management commitment to making it a live operational tool rather than an audit requirement.
Achieving this requires three things: the control plan must be written in language and format that the operator and inspector who use it can understand — not in quality management jargon that means something to the quality manager but nothing to the person standing at the inspection bench; the measurement equipment specified must be available and calibrated at the workstation where the inspection occurs — a control plan that requires a measurement that can only be performed in the quality lab will not be followed in production; and the response plan must be simple enough to follow without ambiguity — a single-page flowchart that anyone trained on the process can follow without consulting the quality manager is more useful than a multi-page procedure that requires interpretation to apply. These are the practical conditions under which a control plan makes the transition from quality documentation to quality control — and they are the conditions that separate foundries whose quality systems actually improve quality from those whose quality systems primarily produce documentation.