Question 1 of 6: Taguchi Philosophy, Cost of Quality, Supplier Certification, and Six Sigma

Question 1: Taguchi Philosophy, Cost of Quality, Supplier Certification, and Six Sigma (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part (a) — Taguchi's two-step approach and the loss function

Taguchi's two-step approach to quality improvement separates the work into a robustness step and a targeting step. Step 1, parameter design, chooses the nominal settings of the controllable design/process factors so that the response is as insensitive as possible to uncontrollable noise (raw-material variation, environmental conditions, wear, customer usage) — this exploits nonlinearity in the transfer function to reduce variance essentially for free, without tightening any tolerance. Step 2, a mean-adjustment step, then uses a factor found to have negligible effect on variability but a strong, roughly linear effect on the mean to move the now-robust response onto its target value. Only if variability is still too large after both steps does Taguchi resort to tolerance design — tightening tolerances or specifying higher-grade components on the sensitive factors — which is the most expensive lever and is used last.

The traditional (goal-post) definition of quality is binary: a unit is "good" anywhere strictly inside its specification limits and "bad" the instant it crosses a limit, so the associated loss function is a step — zero loss everywhere inside spec, a fixed cost the moment a limit is breached. Taguchi's definition instead treats quality as the loss imparted to society from the time a product ships (warranty and rework cost to the producer, plus degraded performance and eventual failure cost to the customer and society), and this loss grows the moment the characteristic drifts away from its target value $T$, whether or not it has yet left the printed tolerance band. The corresponding Taguchi loss function is the parabola $L(y)=k(y-T)^2$, with $k=A_0/\Delta_0^2$ fixed by the cost $A_0$ of a failure exactly at the specification limit $\Delta_0$ away from target. Because $L(y)$ is smooth and increasing on both sides of $T$, it captures the real economic difference between a unit sitting exactly on target and one that is merely inside spec — a distinction the traditional step function cannot see at all — and it gives management a continuing economic incentive to keep reducing variation around $T$ even once 100% of output already conforms, which is precisely the two-step approach's justification: minimizing variance (parameter design) and then centering the mean on $T$ (the mean-adjustment step) together minimize expected loss under $L(y)$.

Part (b) — Cost-of-quality categories, TQM, and the role of management

The classical cost-of-quality framework has four categories. Prevention costs are incurred to keep defects from occurring at all — quality planning, process/product design review, supplier qualification, training. Appraisal costs are incurred to determine the degree of conformance — incoming, in-process and final inspection, test equipment and calibration, audits. Internal failure costs arise from nonconformances caught before the product reaches the customer — scrap, rework, re-inspection, downtime caused by defects. External failure costs arise from nonconformances the customer discovers — warranty claims, returns, complaint handling, liability, and the much harder-to-quantify loss of reputation and future sales.

A successful quality-improvement program should drive internal and external failure costs down sharply (fewer defects are produced and fewer escape to the customer), and this fall is normally many times larger than the modest rise in prevention spending that caused it, because failure costs multiply the further downstream a defect is caught (the well-known "1-10-100" rule: a defect caught at the source costs roughly an order of magnitude more to fix at each subsequent stage). Appraisal costs should also decline over the longer term — once a process is demonstrably capable and stays in statistical control, routine 100% inspection can be scaled back toward audit-level sampling, because the process itself, not inspection, is what is producing conforming output. Only prevention costs are expected to rise, and this is by design: shifting spend "upstream" into prevention is exactly what drives the much larger reductions in the other three categories.

Total Quality Management (TQM) is the organization-wide philosophy that operationalizes this shift. Its key elements are: a relentless customer focus (quality is defined by the customer, not by internal specification alone); continuous improvement (kaizen) as a permanent activity rather than a one-time project; full employee involvement and empowerment, since the people doing the work are best placed to find and fix root causes; process (not just product) orientation, using tools such as SPC, root-cause analysis, and design of experiments; and management-by-fact, i.e. decisions supported by data rather than opinion. Management's role is to lead this shift rather than delegate it — setting the quality vision and long-term strategy, allocating resources to prevention (training, process design, supplier development) instead of chasing symptoms with appraisal, removing the organizational and cultural barriers (blame, short-term production pressure) that block employees from raising and fixing problems, and holding itself, not just the shop floor, accountable for quality outcomes. Without visible management commitment, cost-of-quality spending tends to stay locked in appraisal and internal failure, and the largest, most durable savings (prevention-driven reductions in failure cost) never materialize.

Part (c) — Supplier–producer relations and quality certification

The historic model of supplier–producer relations was adversarial and inspection-based: producers qualified suppliers mainly on price, accepted incoming lots on the strength of receiving inspection (or acceptance sampling), and treated a rejected lot as the supplier's problem to absorb. The modern trend has moved decisively toward long-term, collaborative partnerships: producers now favour a smaller number of qualified suppliers, involve them early in product/process design (concurrent engineering), share cost and quality data openly, and jointly invest in process capability at the supplier's site rather than screening its output after the fact. The underlying economic argument is the same one Taguchi's loss function makes for internal processes: it is far cheaper to prevent a nonconformance at the supplier's process than to detect it on receipt or, worse, downstream in the producer's own product.

Quality certification serves both parties in this new relationship. For the supplier, certification (e.g. to ISO 9001, or a customer-specific certified-supplier status) is a credential that demonstrates a documented, audited quality management system to any customer, reducing the cost of qualifying for new business and, once earned, substituting for that customer's own receiving inspection. For the producer, certifying suppliers reduces or eliminates costly incoming inspection and its associated appraisal cost, gives confidence to source single-supplier for a component (JIT-compatible), and shifts the responsibility for process control to the party that actually runs the process, where it is most effective. Typical phases of the certification process are: (1) a paper/documentation review of the supplier's quality manual and procedures against the standard; (2) an on-site audit verifying the documented system is actually followed in practice (process control records, calibration, corrective-action history); (3) a qualification/trial period, often including a formal process-capability study on critical characteristics, before certified status is granted; and (4) periodic surveillance audits and re-certification to confirm the system remains in control over time, with a re-audit or de-certification triggered by a significant quality escape.

Part (d) — Six Sigma philosophy, key principles, and its five-phase methodology

Six Sigma is a disciplined, data-driven, project-based methodology for reducing process variation and defects, originally developed at Motorola in the mid-1980s and popularized through General Electric's adoption in the 1990s. Its name refers to a capability target of six standard deviations between the process mean and the nearer specification limit — roughly 3.4 defects per million opportunities once the conventionally assumed 1.5$\sigma$ long-term mean shift is applied. Its key principles are: every improvement is run as a chartered project with a defined financial benefit and an executive sponsor; every claim of improvement must be supported by statistical evidence rather than opinion (a strong "voice of the process/customer, not just experience" ethic); trained practitioners (Green Belts, Black Belts, Master Black Belts) lead and support projects under a formal belt hierarchy; and the emphasis throughout is on reducing variation around a target, not merely on shifting a mean or catching defects after the fact — the same underlying philosophy as Taguchi's loss function in part (a).

The five phases of the Six Sigma problem-solving methodology are DMAIC: Define the problem, the customer requirement (CTQ — critical to quality characteristic), and the project scope/goals/financial benefit; Measure the current process performance, validating the measurement system before trusting any data and establishing a credible baseline (e.g. baseline $\sigma$ level or defects-per-million); Analyze the data to identify the vital-few root causes of variation and defects, using tools such as control charts, hypothesis tests, regression, and designed experiments; Improve the process by selecting and implementing changes (often guided by a designed experiment) that address the confirmed root causes, and verify the improvement statistically; and Control the improved process by putting control charts, standard operating procedures, and a response plan in place so the gain is sustained rather than eroding back to the baseline once the project team moves on.