Twenty-eight years from the furnace chart to the chief data office

Mandate

Create the company's first data function: make product records answer customer questions quickly, set shared operational definitions across six plants, and govern analytics, AI and personal information.

Outcomes

• Cut the time to answer a customer genealogy question from twenty-three days in 2022 to three hours and ten minutes in the March 2026 drill.
• Raised heat-lot completeness at source from 91.4% to 99.7% across six plants.
• Reconciled plant-reported scrap to the general ledger within 0.6%, from a gap of 11%, through 46 shared definitions.
• Narrowed a spindle-prediction programme; unplanned spindle downtime fell 27% and saves about C$2.2m a year.
• Introduced an AI register of 23 uses, each with an owner, a stated limit and a review date.

Decisions in context

Our chief executive created this role after a customer's request about suspect titanium exposed how slowly we could assemble records that, individually, were complete. I brought together the traceability, plant systems and analytics teams I had led as vice-president, added a data governance group and took on the delegated role of person in charge of protecting personal information. The work since has been to build the serial genealogy record above our plants' existing systems, agree shared definitions with plant managers, decide where models may inform maintenance and inspection, and set limits on AI use near product acceptance and export-controlled drawings. I present a traceability drill result and our data risks to the audit and risk committee, and I report on personal information protection once a year. I am not a director of the company, and I hold no role in its cyber security function, which reports to our chief information officer.

Reason for the move

Current appointment; completing the extension of the genealogy record to older parts on active programmes.

Mandate

Connect the plants' execution systems, machines and quality records, and build an analytics team that plant managers would use.

Outcomes

• Connected 212 machining centres, furnaces and measuring machines to plant data collection across five plants.
• Brought Querétaro's systems and records into the group standard within eighteen months of the plant's opening.
• Built a 14-person analytics team whose first projects were chosen by plant managers.
• Kept all six plants producing and shipping with complete records through the 2020 pandemic restrictions, using remote record review for special processes where customers allowed it.

Decisions in context

This role moved me from quality records into plant systems. Tessandier's plants ran three execution systems of different ages, and machine data was collected, where it was collected at all, for the maintenance department's own purposes. I connected machines and furnaces in stages, starting with those whose data customers or accreditation auditors asked for, and I brought the young Querétaro plant into the group's record-keeping before its volumes grew. I also built an analytics team and insisted that plant managers chose its first projects, which produced smaller wins and fewer abandoned dashboards. In 2020 we kept every plant producing under pandemic restrictions; my contribution was making special-process records reviewable remotely where customers accepted it. The role taught me how much of a data programme depends on plant managers wanting it, and how little a head-office mandate achieves without that.

Reason for the move

Appointed the company's first chief data officer, reporting to the chief executive.

Mandate

Respond to an engine customer's audit finding that special-process records could not always be tied to the parts they described, and build one quality records system across the plants.

Outcomes

• Closed the customer audit finding within fourteen months, with no repeat finding in the following two audits.
• Replaced four nonconformance systems with one across all plants.
• Introduced electronic furnace records at Boucherville, retiring paper chart recorders on 17 furnaces.
• Created the first serial-number links between plant execution systems, the basis of the later genealogy record.

Decisions in context

Tessandier recruited me to fix a specific problem, and I spent most of my first year on it. An engine customer's audit had found that some heat-treatment and penetrant-inspection records at Boucherville referred to furnace loads and batches rather than to serial numbers, so linking a part to its process run required judgement. We changed the records at source, retired the paper chart recorders on seventeen furnaces and closed the finding within fourteen months. I then replaced four nonconformance systems with one, which gave the vice-president of quality a single view of defects for the first time. Late in the role I built the first links between the plants' execution systems by serial number. They were crude, but they showed what a genealogy record could be, and they made the case for the role I took next.

Reason for the move

Promoted to lead digital manufacturing and operations data across the plants.

Mandate

Run the plant's manufacturing systems and use machine telemetry and warranty data to anticipate component failures for the service business.

Outcomes

• Rebuilt warranty failure coding with service technicians, reducing free-text-only claims from 64% to 9%.
• Identified a hydraulic pump fault pattern that led to a supplier design change and cut that pump's warranty cost by 41%.
• Implemented a new execution system on the Sudbury assembly lines without missing a customer delivery.
• Stopped a planned failure-prediction product whose models were not yet reliable enough to sell.

Decisions in context

I moved to Sudbury for my husband's work and found the most instructive job of my career. Brantwick's loaders and drill rigs sent telemetry from mines, and the service business wanted to sell failure prediction to customers. My first models were poor because warranty claims were written by dealers paid per claim, in free text, and the models learnt the dealers' habits. I spent two years rebuilding the failure coding with service technicians before we modelled anything else. That work found a real pattern in a hydraulic pump, which the supplier fixed. It also led me to recommend that we not sell a prediction product until the models were reliable, a recommendation the service vice-president accepted reluctantly. I also ran the plant's own systems and replaced its execution system on the assembly lines. I left to return to Quebec and to aerospace, where records carry more weight.

Reason for the move

Recruited by Tessandier Aerospace to lead quality systems and traceability, and returned to Quebec.

Mandate

Run metrology and inspection data, then the quality management system, for a machine shop supplying engine manufacturers.

Outcomes

• Built the shop's first database of first-article inspection reports and nonconformances, replacing paper files for new parts.
• Cut the average time to complete a first-article inspection package from nineteen days to eight.
• Led the shop through two successful quality-system recertification audits.
• Reduced measuring-machine program errors by introducing peer review of new inspection programs.

Decisions in context

Castelmont was a family-owned machine shop of about 380 people making discs, shafts and seals for engine manufacturers. I ran its metrology lab and the data that came out of it, and from 2006 its quality system. Our records were almost entirely paper, and the first-article inspection packages our customers required took weeks to assemble because each one was built from scratch. I built a database that held inspection results and nonconformances by part number and characteristic, which halved the time to assemble a package and gave the quality director a view of recurring defects. I studied applied statistics part-time during these years, which changed how I read measurement data. When a US group bought the company in 2008 and moved its quality systems to its own head office, I decided that administering someone else's design was not the work I wanted.

Reason for the move

The 2008 acquisition moved quality systems to the buyer's head office; I chose to leave rather than administer a system designed elsewhere.

Mandate

Support quality on aerospace forgings, then introduce statistical process control on the forging presses.

Outcomes

• Investigated the 1999 furnace-chart rejection and introduced a second, independent temperature record on every aerospace furnace cycle.
• Introduced statistical process control on four forging presses, designed with press operators.
• Reduced dimensional rejects on aerospace rings by 35% over two years.
• Trained 40 operators and inspectors in control charts and measurement basics.

Decisions in context

Ferronde was my first job and gave me responsibility quickly. Ten months in, a customer rejected a shipment of titanium rings because a furnace chart recorder had run dry during part of a cycle. The rings were probably sound, but we could not prove it, and they were scrapped. I led the investigation with the furnace operators and introduced a second, independent temperature record on every aerospace cycle, which the forge kept for as long as I knew it. From 2001 I ran statistical process control on the forging presses. I designed the charts with the press operators rather than for them, and they used them, which taught me more about data adoption than any course has since. I left in 2003 to work closer to machining and metrology, where the measurement data was richer and the parts more complex.

Reason for the move

Moved to Castelmont Précision to work on machining and metrology data for engine parts.