Decisions that slowed us down before they paid

Verifying manufacturers' code data before hospitals scanned it

In the first half of 2024 our nine logistics centres received about 1.6 million cartons a month from roughly 2,400 manufacturers, and hospitals in Sichuan and Chongqing had begun scanning each pack's traceability code as medicines were dispensed under insurance. Most cartons arrived with an outer code that the manufacturer's system associated with the packs inside. When that association was wrong, because a carton had been repacked or a packing line had mislabelled, the error travelled with our delivery note and surfaced at a hospital pharmacy, where a pharmacist found that a pack did not match its record. In March 2024 code-related rejections at hospital receipt reached 1.9 per cent of our delivery lines, and three tertiary hospitals warned that they would stop accepting our electronic deliveries if the rate did not fall.

Our vice president for logistics proposed scanning only the outer carton code and trusting the manufacturers' associations, which kept the docks moving. Our sales leaders did not want to put manufacturers on scorecards. I recommended verification at receipt: sample cartons from every batch opened and each pack scanned, full checking for manufacturers whose error rate exceeded 0.3 per cent, and no electronic receipt of goods without code data. The president approved it over the logistics objection.

The cost was considerable. We spent RMB 46 million on scanning tunnels and handheld devices, hired about 260 people at roughly RMB 31 million a year, and accepted a 14 per cent fall in receipt throughput for six months, which delayed some deliveries to primary-care clinics. Two small manufacturers moved products to other distributors rather than correct their packing lines.

By mid-2025 errors in received cartons had fallen from 0.84 per cent to 0.05 per cent, and code-related rejections at hospital receipt from 1.9 per cent of lines to 0.2 per cent. Every manufacturer now receives a monthly scorecard, and 118 of them have corrected their aggregation processes. No hospital stopped accepting our electronic deliveries.

I would do two things differently. I would have begun verification with our forty largest manufacturers a year earlier, before hospitals were scanning, when errors were cheaper to find. And I would have opened a support desk for small domestic manufacturers at the start instead of after six months of complaints, because many of them wanted to fix their lines and did not know how.

Narrowing the IT estate when procurement reform took the margin

By 2022 volume-based procurement had cut the prices of several hundred medicines and many high-value consumables, and our gross margin on hospital distribution had fallen from 6.4 per cent in 2019 to 5.6 per cent. Net profit fell 14 per cent that year. For the 2023 budget our chief financial officer asked every function to cut 15 per cent, which for IT meant reducing RMB 248 million to about RMB 211 million.

An even cut was the easy answer. I believed it was also the expensive one, because it would thin maintenance and security on the systems we could not operate without while preserving all 438 applications, many of them reports and portals built for a single hospital, region or manufacturer after acquisitions. I proposed instead to retire 162 applications over eighteen months, consolidate three warehouse systems into one, freeze a planned RMB 38 million data-platform programme and raise security spending from RMB 14 million to RMB 22 million.

Our executive vice president for commercial operations disagreed most strongly. The data platform was meant to support paid reporting services for manufacturers, and he saw its freeze as forgoing revenue. Regional general managers objected to losing reports their customers used. Our CFO preferred an even cut because it was easier to defend as fair. The president accepted my proposal on condition that I report quarterly to the executive committee on what had been switched off.

We gave up a good deal. The data platform was delayed by two years. Three regional subsidiaries lost bespoke reports, and one hospital group complained to our chairman. We ended 44 contractor roles and redeployed 19 colleagues.

IT spending was RMB 209 million in 2023. By 2025 the run cost of the estate had fallen from RMB 172 million to RMB 131 million, and security spending had reached RMB 29 million. The single warehouse system proved more valuable than I had argued: when traceability verification began in 2024, we changed one system rather than three.

I would publish the list of applications to be retired, signed by each business owner, before the first switch-off rather than during the programme. I underestimated how much hospitals relied on 23 of the bespoke reports, and we rebuilt nine of them on the standard platform at a cost we should have foreseen.

Keeping pharmacy patient data in China when principals asked for it abroad

In 2022 we distributed medicines for 41 multinational manufacturers, and most of their contracts required us to send daily flow data, showing sales by hospital, batch and quantity, to regional data platforms outside China. Several also asked for data from patient-support programmes run through our specialist pharmacies: enrolment, refill dates and, in some cases, diagnoses. The Personal Information Protection Law had taken effect in November 2021, and rules on security assessments for outbound data followed in 2022. Contracts covering about RMB 9.2 billion of our annual revenue depended on these principals.

Our senior vice president for multinational accounts wanted to keep sending the data and complete whatever filings were needed, because principals were threatening to move products to distributors who would comply. I concluded that hospital-level flow data contained no personal information and could be provided, but that patient-level data from our pharmacies should not leave China for commercial reporting at all; a principal's sales analysts did not need a patient's refill history to plan supply. I proposed an onshore portal giving each principal its own hospital-level data, with patient-programme results reported only in aggregate. The president and our general counsel supported the proposal; the regional legal teams of two principals did not.

The portal cost RMB 12 million and took seven months to build. One principal declined it and in 2023 moved two oncology products in Guizhou and Yunnan to another distributor, costing us RMB 540 million of annual revenue. Our account teams spent most of a year renegotiating data clauses.

The other forty principals now use the portal. No personal information from our pharmacies is transferred abroad, and patient-programme data are held in China under separate consents that state their purpose. The work also produced our data catalogue, which now classifies 1,140 data items by sensitivity, owner and permitted recipient.

I would approach it differently in one respect. I spoke first to the principals' regional commercial teams, who saw only a lost report. Their China data-protection officers, whom I met months later, largely agreed with us and could have shortened the negotiation considerably.

Leaving consumables data inside each hospital

In 2019 Shuheng decided to compete for contracts to manage medical consumables inside hospitals, and I was hired to build the technology. Tertiary hospitals in Sichuan were inviting tenders for these services, and our strongest competitor was offering a cloud platform at no charge, holding each hospital's consumables usage, linked to patient billing, in the competitor's own data centre.

Our vice president for hospital services wanted to match the offer with a multi-tenant cloud of our own, at about RMB 0.8 million per hospital. I proposed deploying inside each hospital's network instead, connected to its clinical and resource-planning systems through the hospital's own integration platform and within its protection grading, at about RMB 2.1 million per hospital. Patient-linked records would remain with the hospital; we would receive item-level consumption without patient identifiers. Having been a hospital's director of information, I did not believe a hospital should have to trust a supplier with records about its patients in order to have its cupboards stocked. Our CFO questioned the cost, and the vice president thought I was handing the competitor the market.

In 2020 we lost four of eleven tenders, three of them on price. Each deployment took about fourteen weeks against six for the cloud service, and by March 2021 we were running in 21 hospitals, fewer than the business plan had assumed.

After the Personal Information Protection Law took effect, hospital information departments began asking every supplier where patient-linked data was held. Between 2022 and 2024 nine hospitals that had chosen hosted services re-tendered their contracts, and we won seven. We now run the service in 58 hospitals, ward stock falls by about a third in the first year of a deployment, and no patient-identifiable consumables record is held outside a hospital.

I would standardise sooner. Our first nine deployments used individual interfaces, and bringing them onto a standard package in 2022 cost RMB 6.2 million. I would also have published our data design to hospital information directors at the outset, instead of explaining it tender by tender.