Keep the reasoning and the consequence together

In March 2024, an oral-dose development programme had completed its planned 50-kilogram engineering runs. The investment timetable assumed that a 300-kilogram route could follow with the same operating instructions. Manufacturing colleagues preferred to reserve the larger equipment immediately: delaying the reservation could move the programme behind another project. I was concerned that the proposed sixfold increase described mass without explaining the changed equipment and material experience.

I asked process and operations specialists to identify the assumptions that the larger run would actually test. The programme concentrated on loading, mixing behaviour and the drying endpoint for this formulation. We did not invent universal operating limits from a smaller run. The team prepared a separate engineering plan for the larger equipment, with quality involvement and clear separation from any later validation or release decision. Reserving capacity remained useful, but it could no longer be presented as evidence that the process was ready.

The twelve-week programme cost CNY 4.6 million and used three planned 300-kilogram engineering runs. The first exposed a drying difference that required investigation. All observations remained in the development account. The revised plan examined a changed endpoint approach in the remaining runs, with the rationale and conditions recorded. The work supported choosing a particular equipment route for further development; it did not establish commercial validation, product approval or patient benefit.

The consequence was a six-week movement in the investment decision date and the loss of the originally preferred equipment slot. That was an actual cost to the schedule, not a delay erased by later technical progress. The manufacturing sponsor accepted the revised route after seeing both the development evidence and the new capacity implications. We had purchased a better-founded next decision rather than certainty about every future batch.

I would bring equipment specialists into the initial programme estimate earlier next time. The important omission was not a missing test added late; it was allowing a scale assumption to become a scheduling fact before the people responsible for the receiving equipment had examined it.

In September 2024, two laboratories preparing an analytical transfer produced different results from a common set of development samples. The project sponsor wanted the receiving laboratory to repeat its work quickly so the transfer could be reported complete. Our analytical colleagues had not yet established whether the difference arose from sample preparation, instrument execution or the material itself. A repeat without a defined question would add numbers without necessarily adding understanding.

I supported a four-week investigation proposed by the analytical and quality leads. Its CNY 1.2 million cost covered specialist time, planned comparative work and the resulting review. Tinglan Que retained the quality decision; my role was to secure capacity and make the technical question explicit. The original results remained visible. Additional examinations were identified as new work under the investigation plan, not replacements for observations that had become inconvenient.

The comparison used twelve retained development samples, with preparation steps observed jointly by both laboratories. The investigation identified a difference in how one preparation instruction was interpreted. That finding supported clarifying the instruction and undertaking the next agreed transfer work. It did not justify declaring the first results invalid simply because an explanation appeared plausible. The authorised review considered the complete record before deciding the next step.

The receiving laboratory's transfer decision moved by four weeks. The programme could explain the preparation difference and had an agreed, executable instruction; its scheduled milestone was still missed. I rejected a proposed presentation that showed only the later aligned observations. The sponsor's wish for a concise success statement was understandable, but omitting the original difference would have removed the reason the work mattered and made the account less useful to a future laboratory.

My lesson concerns language before it concerns equipment. Next time I would ask the receiving analysts to demonstrate their interpretation of the procedure during planning, before comparative samples are used. A document that both teams have signed can still contain different practical meanings. Making those meanings visible earlier would improve the transfer design without assuming that any future discrepancy has the same cause.

In February 2025, a development team in Nanjing was preparing an oral-dose process for further engineering work at Suzhou. The receiving team had the written sequence, equipment list and development reports. Its questions nevertheless revealed that several choices were understood mainly by the people who had made them. A material addition order looked arbitrary in the instruction because the observations behind it were scattered through earlier work.

Operations preferred a short handover followed by learning during the scheduled campaign. I proposed a six-week joint preparation period instead, costing CNY 2.3 million. The development lead argued that writing another report would consume the same specialists needed for new candidates. I agreed that a larger document alone would not solve the problem. We used paired technical sessions in which receiving specialists explained the process back to the developers and identified the decisions they could not yet account for.

The resulting transfer account linked important material and equipment choices to the observations supporting them. It retained rejected routes and unresolved questions rather than presenting the final instruction as the only route ever considered. Three receiving specialists visited the development work, and two developers joined the first Suzhou engineering campaign. Their presence supported interpretation; it did not transfer quality authority or allow informal changes outside the agreed work.

The preparation delayed the campaign start by three weeks because specialist availability was limited. During the campaign, the teams could explain the addition-order rationale and identify when a proposed equipment adjustment needed a new assessment. The outcome was a usable shared account for that transfer, not a measured reduction in all future deviations. We did not claim a numerical productivity gain from a programme designed principally to improve understanding.

I would make the receiving team's questions part of development reviews earlier, rather than create a concentrated explanation effort near transfer. Knowledge that is obvious to an originating team becomes expensive to reconstruct after its members disperse. The next programme should reserve time for the recipient to challenge the account while the relevant observations and people are still readily available.

In August 2025, the technical portfolio faced a choice between an estimated CNY 18 million pilot facility and a CNY 72 million larger line. The larger proposal offered more nominal capacity and fitted the operations team's longer-term ambition. The current development portfolio, however, contained twelve projects with different equipment and learning needs. A single capacity total concealed whether either proposal would serve the work actually approaching its next decision.

I asked the programme owners to map each project's requirements against both options. Nine could use the pilot route for their planned development work. Three needed capabilities outside its intended scope and would require separate provision. I recommended the pilot investment, with those exclusions stated in the capital paper. Finance challenged whether a smaller purchase would merely defer another unavoidable expenditure. My answer was that the remaining projects needed their own assessment; their existence did not establish that the larger line was the right answer for all twelve.

The board approved a CNY 18 million envelope. The six-month installation and engineering-readiness programme finished in February 2026 at CNY 17.6 million. The unspent CNY 0.4 million remained against the approved envelope; it was not a return on the investment. Operating expenditure and later project work were separate from that capital figure. Quality and operations retained the decisions required before any particular use of the facility.

The pilot was available for the nine identified development routes, subject to each programme's own readiness. The three excluded projects still required another path and were retained in the portfolio discussion. I resisted describing the purchase as resolving the whole capacity problem. We had chosen a facility suited to a defined body of learning, while accepting a continuing obligation to decide what the other work needed.

Next time I would include a broader range of plausible portfolio changes in the initial comparison. Our project-by-project analysis was useful, but a pilot that fits today's nine projects may fit a different future mix less well. That sensitivity belongs in the investment discussion before a preferred equipment scale becomes the centre of the argument.

In May 2026, a formulation candidate reached a development review after CNY 6.8 million of expenditure. Its planned stability work had not supported the route needed for the next commitment. The development sponsor proposed a further three-month programme costing CNY 2.1 million. The commercial team had already included the candidate in an internal planning scenario, which made ending it feel like losing both the technical work and a future opportunity.

I asked the team to distinguish what another programme would test from the understandable desire to rescue the existing route. The proposed work repeated much of the earlier approach without a sufficiently grounded explanation of why the result should change. Quality, analytical and formulation colleagues reviewed the evidence within their respective responsibilities. My recommendation was to stop this development candidate and retain its complete record, including the observations that might inform future research.

The executive review accepted the recommendation. The CNY 6.8 million remained the cost of work undertaken, not an amount recoverable by avoiding the next stage. The unapproved CNY 2.1 million proposal was not described as a realised saving. Six specialists became available for reassignment after closing the programme's documentation and materials responsibilities; no payroll reduction followed. The commercial planning scenario was revised to remove the candidate's assumed contribution.

A different formulation concept was considered separately in July. It did not inherit an approval, timetable or probability of success from the stopped programme. Its sponsor had to set out its own rationale and evidence plan. That separation mattered because an organisation can appear to stop a weak project while quietly carrying its unsupported promises into a renamed successor. We preserved the learning without preserving the forecast by default.

I would agree more explicit continuation questions at the start of the next development stage. In this case the technical evidence was available, but the team had not articulated in advance what would make another commitment worthwhile. Doing that earlier would make a stop recommendation less dependent on a difficult late discussion after commercial expectations and personal effort had accumulated.