Legacy System Replacement Under a Regulatory Availability Requirement
Medway Labs runs a 12-year-old on-premises laboratory information system (LIS). Monthly unplanned downtime averages 9 hours; two outages last quarter delayed 1,400 tests. The vendor will end support in 18 months. An on-premises refresh costs $1.2 million upfront plus $250,000 annual support; a SaaS replacement costs $35,000 per month with a 99.9% uptime SLA and 4-hour RTO. The lab must retain test data for 10 years. The CIO favors SaaS; the CFO notes the cumulative five-year SaaS cost is higher. A regulator requires documented availability and auditability. Staff have limited cloud experience.
1. Compare the options financially and identify costs that are easy to miss.
Read the worked answer
Five-year on-premises cost: $1.2 million + (5 x $250,000) = $2.45 million. Five-year SaaS subscription: $35,000 x 60 = $2.1 million. On the surface SaaS is cheaper, but the comparison is incomplete.
Easy-to-miss costs:
- Migration, interface rebuilds, and parallel-run costs for both options.
- SaaS data egress, archival, and long-term retention for 10 years.
- Integration, identity management, and network upgrades.
- Training and change management for staff with limited cloud experience.
- On-premises facilities, disaster recovery, backup, and staff time; the refresh may require a second site.
- Contract exit, transition assistance, and potential price escalation.
After loading these, the five-year totals may be close. The decision should turn on risk and service quality, not subscription price alone.
2. How should business impact analysis and RTO/RPO drive the choice?
Read the worked answer
The current 9-hour average monthly downtime is an availability failure with measurable operational impact: 1,400 delayed tests in one quarter. A BIA should quantify the cost per hour of LIS downtime, including delayed patient care, staff idle time, and regulatory reporting. If the tolerable downtime is below the current outage duration, the status quo is unacceptable regardless of cost.
The SaaS 4-hour RTO is meaningful only if the vendor's SLA is backed by evidence and the lab's interfaces and downstream systems can recover within the same window. RPO matters separately: how much test data can be lost? For a lab, near-zero RPO is typically required, so replication and backup design must be verified.
Judgment: use the BIA to set required RTO/RPO, then test whether each option meets them. Do not accept a 99.9% SLA as proof; 99.9% still permits about 8.8 hours of downtime per year, and credits are not continuity.
3. Recommend contract terms and controls to put in place before signing.
Read the worked answer
Before signing, require:
- Audit rights and evidence: current SOC 2 Type II or ISO 27001 report, penetration-test summaries, and subprocessor list.
- Data ownership, residency, encryption, and return/destruction obligations, including format and cost of export at exit.
- A transition-assistance clause and escrow or exit plan so the lab is not locked in.
- SLA definitions that match the BIA: measurement method, exclusions, service credits, and termination rights for repeated misses.
- Breach notification timelines and cooperation obligations.
- A validated disaster-recovery test and periodic restore testing.
- Regulatory auditability: retained logs, change records, and access evidence.
Internally, assign a business owner, run a parallel pilot before cutover, and document the residual vendor-dependency risk for the board.