Case study, government ICT
The project office behind thirty-two hospitals
Thirty-two sites, 118 live workloads moved without a user noticing, two months early, on budget to the cent, and two audits with no findings.
The job we were given
Ndende Technologies, with PTP Integrated, had won a five-year contract to replace the server, storage, network, backup and disaster recovery estate of the Mpumalanga Department of Health across thirty-two sites, spread over a province the size of Portugal. Ndende’s engineers knew the technology. What the contract also needed was a project office that could hold thirty-two sites, four equipment vendors, a provincial steering committee and the Auditor-General to one plan for a year. Ndende asked DBiLL to lead it, and Thandwefika Tshabalala signed the charter as Project Manager on 30 October 2025.
What the Department was living with
The Department’s own charter says it in one line: end-of-life platforms, fragmented backup and limited disaster recovery, increasing operational risk and downtime. In practice that meant a different platform at nearly every hospital, backups at some sites and on some tools, and no way to restore the data centre at SITA Mbombela if it failed. The site assessment in October 2025 found nine server rooms with broken air conditioning and five without proper physical security. A pharmacy at a district hospital dispenses from paper when its server goes down, and the nearest engineer is often three hours away.
How we ran it
We wrote one plan and made everything answer to it: five gated phases, a weekly status report every Friday, a monthly steering committee chaired by the Department, a joint change control board and one RAID register that the Department’s ICT manager could read any day of the year. Three decisions in the first month shaped the whole project.
- Readiness before delivery. Equipment went to a site only after its power, cooling, rack space and security had been checked and signed. Nine air-conditioning repairs and five room fixes were closed before a single server left the depot, and the sequence ran data centres, then district offices, then tertiary hospitals, then the twenty-two district hospitals.
- No cutovers in December. Hospitals run short-staffed over the holidays, so the plan blacked out the month and the migration was built around it.
- Audit-ready from week one. Every deliverable had a named acceptance criterion, a document, a signatory and a place in a register. We built the paperwork to be read by an auditor rather than assembled for one.
- A switchover nobody noticed. The 118 production virtual machines were planned into three migration groups, each cutover rehearsed, and Veeam’s online replication kept the old server serving while its copy was built on the new cluster. The final switch took about thirty minutes per machine inside an agreed maintenance window. Clinicians, pharmacists and clerks worked through the whole migration without noticing it had happened; the migration report was signed on 30 June 2026.
- Two months early, to the cent. The baseline ran to September 2026; handover finished on 17 July 2026 with nothing dropped from scope. When Rob Ferreira Hospital’s server room proved too small and Ermelo’s had no space for the new rack, the change control board re-sequenced those two sites inside a week and the provincial date held. The final account matched the agreement to the cent, with no penalties and no service credits.
- Two audits, no findings. The Department’s own audit in February 2026 was answered with the deliverables register. On 14 May 2026 the Auditor-General issued Request for Information No. 38 of 2025-26 on health-sector ICT investments, due the next day; the charter, plan, status reports and evidence went back inside the day, and the report carried no findings against the project.
The outcomes
One platform the Department can run
Sixty-five servers and thirty-six switches on one standard design, visible from one console, with backups the ICT team can see and restore from, and a disaster recovery position measured in seconds instead of rebuild days.
Clinical service never interrupted
All 118 workloads moved at the first attempt inside maintenance windows, with network redundancy up throughout and a rollback path that was never used. Nothing was announced, and the helpdesk had an ordinary week.
A record an evaluator can check
UAT passed and signed at every site, a close-out report signed by the ICT Steering Committee, the Project Lead and the Project Manager, and two audits answered from the register with no findings.
A project office that stays
Four Department engineers trained in the handover week of 13 to 17 July 2026 with runbooks and as-built records, and by the second audit Ndende’s own project manager was running the response with the Department while DBiLL assured it.
Why this matters for your next project
When the work is bigger than the team has run before
The risk sits in the plan, the paperwork and the thirty-two sites that each want a different date, long before it sits in the technology. That is the work DBiLL takes on. Ask us for the close-out report and Ndende’s contact.