Each wider remit made the meaning of a local record more consequential.

Mandate

Make the basis and limits of enterprise industrial data intelligible across physical records, analytical uses and accountable decision owners.

Outcomes

• Preserved timing qualifications after a 2021 review found observed clock offsets on thirty-seven of two hundred machines.
• Separated current definitions, aliases and superseded definitions in the 12400-entry 2023 parts programme.
• Kept measured, modelled and missing submissions distinct across the 270-supplier 2025 footprint review.

Decisions in context

The CDO appointment required me to explain a combined data account to people who had not created its individual records. A factory's familiar shorthand could not remain an unstated assumption once the material supported an enterprise decision. I organised the function through seven direct reports and kept its expense delegation distinct from the resources controlled by plant, engineering and product leaders. The timing review, parts programme and supplier account tested that distinction in different ways. My team could make a qualification visible without claiming the specialist conclusion that followed from it. I have presented these questions to the board as an executive since March 2019, rather than as a statutory director. A search account should assess the contribution through what became clearer and what remained unresolved. The company's revenue and factory footprint provide context for the work; they do not establish that the data function owned every operating outcome across the manufacturer.

Reason for the move

Promoted from vice-president of industrial data; current appointment.

Mandate

Connect cross-site data development with a precise account of the population and decision an analytical claim was intended to support.

Outcomes

• Questioned a 2017 evaluation whose random row split retained observations from the same forty-two assets on both sides.
• Withheld the proposed claim about unfamiliar-asset performance pending a design appropriate to the intended use.
• Retained S$2.3 million in further analytical work without attributing an unobserved maintenance benefit.

Decisions in context

The vice-president role placed more of my attention on how industrial analysis was presented beyond its originating team. A model result could be technically reproducible while its public description implied a population that the evaluation had not examined. The maintenance study made that problem concrete. I asked colleagues to distinguish rows excluded from training from assets or future periods that were genuinely unfamiliar to the model. The request did not make the previous work worthless, but it did remove the basis for the stronger claim. I accepted the cost and delay of further analysis rather than attach a later success figure that the record did not contain. This appointment broadened my responsibility for the relationship between data development and its proposed use. It did not transfer maintenance decisions from qualified plant professionals to the data team. That boundary remained important as the function grew and its work attracted more executive attention.

Reason for the move

Promoted to CDO as enterprise data responsibilities became a CEO-reporting function.

Mandate

Develop cross-plant analysis that retained the physical meaning and limitations of locally created production records.

Outcomes

• Owned the 2008 report that conflated eighty-six unique codes with eighty-six physical production lots.
• Reconciled twenty-six codes used in two plants and sixty used in one to a population of 112 lots.
• Withdrew the complete-traceability claim and supported S$640,000 in review expense.

Decisions in context

Leading plant analytics made me responsible for an account assembled from several local perspectives. The lot-count error showed that I had underestimated what could disappear in that assembly. I endorsed the uniqueness of a code column without asking whether its scope matched the physical population described. When Mei Arvellan challenged the result, I needed to do more than accept a corrected total. The earlier traceability claim had to be withdrawn, and the review needed room to examine the supporting records. I kept the analytical correction separate from any conclusion about product quality or safety, which belonged to the responsible professionals. The expense and loss of confidence in the original report were consequences of my decision, not inconveniences to remove from the career story. This role prepared me for wider data leadership by exposing the difference between a technically coherent table and a defensible explanation of the work it represented.

Reason for the move

Promoted to lead industrial data beyond the plant-analytics remit.

Mandate

Maintain common definitions and relationships without discarding the context needed to interpret earlier or local records.

Outcomes

• Developed a distinction between current selection records and definitions retained for historical interpretation.
• Worked with users to make shared data understandable without assuming every local convention was redundant.
• Built supervisory experience across the maintenance and explanation of industrial records.

Decisions in context

The supervisory appointment introduced the continuing work behind a common definition. A catalogue could not remain intelligible simply because someone had agreed its labels once. Colleagues needed to know why a relationship existed, what an earlier definition meant and when a change required another owner to examine its use. I led a small function with an expense delegation appropriate to that responsibility, while production and engineering decisions stayed elsewhere. The most useful learning concerned conversations with the people who used the records. A consistent label could still be inconvenient or ambiguous in practice, and a local term could carry information that a central summary had overlooked. I began to see maintenance of meaning as part of the function's work rather than a preliminary clean-up before analysis. Moving into plant analytics allowed me to examine the consequences when those definitions were used to support a broader account of manufacturing activity.

Reason for the move

Promoted into plant analytics to work more directly on the conclusions drawn from combined manufacturing data.

Mandate

Examine production records under supervision and connect reported figures with the conventions and events behind them.

Outcomes

• Learned to ask how a local record had been created before relying on its column heading.
• Developed experience explaining differences between a recorded value and an inferred manufacturing conclusion.
• Progressed from individual analysis into responsibility for shared industrial definitions.

Decisions in context

I began in August 1990 after completing my industrial-statistics degree that June. The twelve-person team gave me access to colleagues who could explain the practical origin of a production figure. My responsibility was to contribute analysis that others could examine, rather than to direct the factory or claim ownership of its output. I learned that a record's apparent simplicity often depended on knowledge shared by its original users. Asking about that knowledge could change the meaning of a comparison before any calculation changed. The experience supplied a foundation for later work on common definitions, but I would not present it as proof that I had already mastered every problem of industrial identity. My subsequent lot-count error shows otherwise. This first appointment is useful in the chronology because it describes the setting in which the questions began, along with the modest authority and scale of the role actually held.

Reason for the move

Promoted to master-data supervision after developing analysis and cross-functional record experience.