A product decision with its conditions still visible

In April 2024, two pump designs were competing for the next industrial product programme. The commercial team favoured the design with the higher peak efficiency on its curve. Our application review described a different problem: the intended duty moved between three operating conditions during a twelve-hour reference day. A peak measured elsewhere on the curve did not tell us which design would use less electrical energy while providing those same required conditions.

I funded a ten-week comparison costing C$1.4 million. Test specialists established matched flow and head requirements at each point and measured electrical input using the agreed method. The reference schedule assigned six hours to the first point, four to the second and two to the third. Design A drew 12.4,17.6 and 24.1 kilowatts respectively. Design B drew 11.2,16.9 and 25.0 kilowatts. B was worse at the third point, a fact the product recommendation needed to retain.

Multiplying each input by its scheduled hours gave 193.0 kilowatt-hours for A and 184.8 for B: an 8.2 kilowatt-hour difference over the defined reference day. These were calculated duty totals using measured points, not an annual customer saving. A changed schedule could alter the comparison, and the experiment did not represent every installation. We chose B for further product development because the stated application weighted its stronger conditions more heavily.

The commercial sponsor objected that a lower peak figure would make the headline specification harder to sell. I accepted that communication cost and asked the application team to explain the relevant duty instead of replacing one selective claim with another. The C$1.4 million study cost remained separate from a modelled energy difference. We did not multiply the result across the installed population to manufacture a financial return.

Next time I would require the duty assumptions before approving the initial comparison brief. The test team could execute either brief accurately; leadership had to ensure it was answering the selection question the customer actually faced. A technically sound measurement at the wrong operating point can still support a commercially poor choice.

In October 2024, a pilot shaft programme used a 50.000 millimetre nominal diameter with a±0.020 millimetre tolerance. One reported measurement was 50.014 millimetres. It lay below the 50.020 upper limit, but the expanded uncertainty associated with the existing measurement approach was±0.008 millimetres. The corresponding interval extended from 50.006 to 50.022. The displayed central value alone could not establish acceptance under the quality team's agreed rule requiring that interval to remain within the limits.

Production wanted to use the existing gauge because its calibration was current and a more demanding measurement arrangement would slow the pilot. Metrology colleagues explained why a calibration reference did not establish suitability for this particular decision. I supported an eight-week C$740,000 programme to improve the measurement arrangement, including preparation, fixturing, environmental control and specialist evaluation of the uncertainty account. Martin Orswick retained the acceptance decision.

The revised approach produced a documented expanded uncertainty of±0.003 millimetres for the stated measurement conditions. The same 50.014 illustrative central value would then correspond to 50.011–50.017, inside the limits under that rule. We did not use this arithmetic to declare every pilot part acceptable. Each actual result required the appropriate measurement and review, and the improvement did not retrospectively change the uncertainty associated with earlier observations.

The new arrangement added approximately four minutes of preparation per examined part in the pilot work. I accepted that cost while manufacturing and metrology considered a supportable longer-term method. We did not repeatedly measure until a favourable number appeared or quietly widen the drawing tolerance to avoid the issue. Any requirement change would need its own technical justification and authorised approval.

My improvement next time would be to review measurement capability while the tolerance is being proposed. The team should know whether the required decision can be supported before a drawing and production schedule assume that it can. More decimal places on a report cannot supply the missing understanding, and a traceability statement should never be asked to do that work by itself.

In January 2025, a seal-cartridge redesign offered a common assembly for two pump families. The engineering proposal reduced duplicated design work and made future purchasing more straightforward. Service colleagues identified a complication: earlier installations used a housing interface that could not accept the new cartridge directly. A catalogue showing one replacement part would look simpler while leaving maintainers to discover the difference on site.

I chose a staged commonality programme with two explicit service routes. New production could use the common assembly after its own qualification work. Earlier equipment would retain the supported original replacement or use a separately defined conversion kit where the installation assessment permitted it. Sales preferred to describe the new cartridge as a universal replacement; I declined that wording because the interface condition was material to the purchase and the maintenance task.

The nine-month programme cost C$3.6 million and completed in September 2025. It included development examination, service documentation and preparation for both routes. The conversion kit added C$420 to the parts cost for an eligible earlier installation, excluding labour and other site work. The original service cartridge remained in the support plan. We did not count its continuing obligation as a failure to standardise; it was part of the decision to keep the installed population supportable.

The useful outcome was a common design for new production with a service account that preserved the earlier interface. It did not establish that every installed pump had been converted, nor did the programme claim a fleet-wide maintenance saving. The service team could identify the appropriate route from the equipment record and the assessed circumstances before making a promise to the customer.

I would involve field-service specialists at the concept review next time, before a common-component drawing becomes the preferred commercial story. Their knowledge of installed conditions changes what commonality can responsibly mean. A product family is simpler only when the people maintaining it can understand the exceptions that remain, rather than being expected to solve them after the replacement arrives.

In June 2025, a product team proposed a modular wear assembly for a pump used in abrasive service. The existing design required a larger subassembly to be removed when the wear element needed replacement. The modular route offered a more accessible task but added interfaces, 2.4 kilograms of material and C$180 to the initial unit manufacturing cost. Marketing wanted to introduce it as a lower-impact product before those trade-offs had been examined together.

I supported a fourteen-week development and service trial costing C$960,000. The programme examined the proposed interface, the replacement task and the instructions under defined workshop conditions. Six service technicians each performed both tasks after familiarisation, with the order alternated across the group. The median observed replacement time was 94 minutes for the existing route and 61 minutes for the modular route. Those twelve observations described a small workshop exercise, not every field repair.

I approved further introduction work with the service result stated at that scope. We retained the extra material and initial cost in the product account and did not convert a faster task into a whole-life environmental claim. Such a comparison would need assumptions about operating life, replacement frequency, materials, transport and the fate of removed components. The workshop exercise had not answered those questions, and no customer downtime claim followed from the timing result.

The commercial team accepted a narrower description focused on replaceability. The decision preserved a useful design feature while giving up a more ambitious launch claim. Engineering also retained the examination and maintenance responsibilities created by the additional interface; modularity did not make them disappear. The product's value would depend on the service circumstances in which that feature was actually used.

Next time I would agree the proposed public claims at the start of development. That would help separate the evidence needed for a repair task from the much wider evidence needed for an environmental comparison. A good service feature does not need an unsupported second story to justify being investigated, and the team should not discover that distinction only when the launch language is already written.

In February 2026, a supplier announced the planned end of a control-board component used in a mature pump-drive assembly. A final purchase could support continuing service, but it would also create inventory, ageing and forecasting obligations. The product group wanted to buy enough for the full remaining support horizon. Finance questioned a large stock commitment based on a service-demand estimate with considerable uncertainty. An immediate redesign would avoid some inventory but could not responsibly be completed before the supplier's order deadline.

I recommended a bounded bridge: C$1.8 million of component inventory against an eighteen-month planning period, alongside a C$2.7 million redesign programme scheduled for twelve months. The inventory period was an assumption for planning, not an assurance of demand coverage. We agreed review points for actual consumption and the redesign's progress, with service, procurement and engineering sharing the account. Storage conditions and the suitability of later use remained matters for the relevant specialists.

The final order was placed in May 2026. By September the redesign had completed its architecture review and prototype preparation; examination and release work remained ahead. I reported that stage explicitly instead of describing the replacement as complete because a new design existed. The bridge inventory supported the current service route while the next route was developed. It did not remove the possibility of excess stock or a need to reconsider the schedule.

The decision gave up the apparent simplicity of either buying for every possible future need or declaring an immediate replacement. It created two commitments that had to be managed together, with different costs and completion conditions. No saving was claimed from comparing the purchase with an unapproved larger inventory estimate. The board discussion focused on the exposure retained and the evidence that would trigger another decision.

I would make component-life review a regular part of mature-product planning earlier. The supplier notice made the timing urgent, but the installed support obligation existed long before it arrived. A deliberate review would give engineering and service more room to choose a transition before an order deadline narrows the available options.