01Why does "process design" decide quality?

Most rework is born not of a lack of ability but of the absence of process. "Just make it, and check when it's done" — this ad-hoc way concentrates checking at the end, and when a problem is found there, all prior phases must be redone.

The idea that prevents this is lifecycle design. Distribute checks (gates) through the middle of the process, and at each phase judge "may we proceed to the next?" Catch a problem near the phase where it arose. The earlier you catch it, the smaller the rework (this increase is quantified in section 05).

Process design is not something to think out from scratch for each case. Hold one standard lifecycle, and adjust it to the scale and medium of the case. The 10 phases this volume shows are the template for that standard.

02The whole course of a received case — 10 phases

Break the whole course of contract production into 10 phases. For each, make explicit the main work and the main owner.

#PhaseMain work / check
1Receiving the requestConfirm SoW, responsibility boundary, confidentiality; first judgment of advertising applicability (Vol. 1)
2Concept confirmationFix the intended appeal and the primary sources to reference (package insert, approved scope)
3Regulatory mapLay out the applicable norms in four layers (Vol. 3)
4ProductionActual production while referencing the regulatory map
5AI checkFirst-pass screening of violation candidates (Vols. 5 & 6)
6QCQuality check of consistency, typos, citations, visual elements (Vol. 7)
7QAAssurance of regulatory compliance, the deliver/return decision (Vol. 8)
8DeliverySigned delivery, attaching the grounds of judgment (Vol. 10)
9Audit responsePresenting records to client review and regulatory queries
10Record-keepingStoring the decision process and version control (Vol. 10)

This phase list can be merged or omitted according to scale. Adjustments like merging 5 and 6 for a small case are free. But the four — "applicability judgment (1)," "fixing primary sources (2)," "regulatory map (3)," and "QA (7)" — are not omitted for any case. This is the minimum line of process design.

03Where to place the quality gates

Placing a gate between all 10 phases is excessive. From rework cost and detection effect, concentrate resources on four main gates.

Gates A and B are upstream gates, C and D downstream gates. What is most slighted in vendor quality management is the upstream gates — pressured to start making things fast, the confirmation of concept and regulatory map gets skipped. But rework cost is maximized when an upstream gate is skipped (section 05).

04Make the gate's pass criteria explicit

A gate must not be passed on "it seems fine somehow." Define for each gate explicit pass criteria (a checklist).

A vague gate

"It's roughly done and seems fine, so on to the next."

The pass criteria are not put into words. The judgment depends on the person's feel, varying by person and by day. Later you cannot explain "why it was passed."

A gate with explicit criteria

"Gate B pass criteria: (1) the applicable medium's norms are listed in four layers; (2) each efficacy/effect statement is tied to the corresponding line of the package insert; (3) points to confirm are listed and assigned to owners — confirm all three are met."

The conditions of passage are itemized, and whoever judges reaches the same conclusion. The judgment is also recorded.

The principle of building pass criteria is one: build "points that could be flagged later" into the gate criteria in advance. Turn points flagged in the past by internal review or the client into criteria, and the same omission never happens again (this idea is explored in the check regime of Vol. 4).

05The earlier the gate, the cheaper — the escalation of rework cost

Why are upstream gates important? Because rework cost grows exponentially the further downstream the process.

Phase where the problem is foundRange of reworkRelative cost
Concept confirmation (2)Only the direction is correctedSmall
During production (4)Remake of the relevant partMedium
QA (7)Major revision of the material, re-QCLarge
After delivery / audit (9)Recall, revision, trust damageMaximum (includes irreversible)
Principle: The same single problem, caught at the concept stage, is settled with a direction fix; found after delivery, it leads to a recall. Invest in quality gates with weight upstream — that is rational. Push the upstream gates thin under the field's voice of "we want to start producing fast," and you pay many times the cost downstream. Speed and quality are not a trade-off; they are reconciled by making the upstream gates thick.

06Make responsibility allocation explicit — separate the maker and the checker

Another pillar of process design is responsibility allocation. Assign to each phase a "main owner (who makes / drives)" and a "checker (who judges the gate)", and separate the two.

Making responsibility allocation explicit supports not only quality but the response when trouble arises. If "at which phase, who, checked what" is recorded, even when a problem occurs you can identify the phase of cause and move to improving the process rather than a person-dependent hunt for the culprit (the specifics of records are in Vol. 10).

07A self-check for lifecycle design — three questions

When designing the process of a new case, three questions to check.

  1. "Are the four mandatory phases (applicability, primary sources, regulatory map, QA) built into the process?" — If any is missing, do not omit it on grounds of scale. Hold the minimum line
  2. "Does each main gate have explicit pass criteria?" — If not, turn "points that could be flagged later" into criteria. Do not pass on feel
  3. "Is a main owner and a checker assigned separately to each phase?" — If they are combined, separate making and checking

Start production only after a process design to which you can say "yes" on all three. The time spent on design returns many times over in downstream rework saved.

08Summary — "design" the process per case

The points of this volume. (1) Quality is secured not by adding at the end but by gates distributed through the process. (2) A received case breaks into 10 phases, and at least 4 (applicability, primary sources, regulatory map, QA) are not omitted. (3) Make pass criteria explicit at the four main gates, and allocate investment with weight upstream. (4) Assign a main owner and a checker separately to each phase.

In closing

"Rather than spend time on design, I want to start making things fast" — on the floor of contract work, this temptation is always there. But a case that runs off without designing the process will surely stall somewhere downstream and be remade all the way back upstream. The first time spent on design is not a cost but the greatest time-saving investment.

Hold one standard lifecycle and adjust it per case. Place pass criteria at each gate, and separate the one who makes from the one who checks. These are not "carefulness" but the engineering of fast and reliable production. From individual-dependent quality to quality anyone can reproduce — that skeleton is the process design of this volume.

Vol. 3 (in preparation) takes up the core of phase 3 — the regulatory map. It organizes the PMD Act, advertising standards, the MSA guideline, and JPMA into four layers and makes concrete the procedure for mechanically mapping received materials. From the next volume, we fill in the content of the process designed here.