Precision Parts' SaaS ERP Business Case
Precision Parts Ltd. operates three plants and a 22-year-old on-premises ERP. Unplanned ERP downtime costs about $18,000 per hour; the system has averaged 14 hours of unplanned downtime and 30 hours of planned maintenance each year. IT spends 70% of its $2.4 million annual budget on run activities and 30% on change. The CFO is comparing a SaaS ERP at $620,000 per year for five years, with a one-time implementation cost of $1.1 million and $300,000 of parallel-run costs, against refreshing the on-premises system: $900,000 of hardware in year 1, $380,000 per year in licensing and support, and 4 FTEs at $110,000 fully loaded. The vendor claims '30% lower TCO' and payback in 14 months. The CFO uses an 8% discount rate and will approve only benefits tied to measurable outcomes, not 'agility' or 'innovation'.
1. Critique the vendor's 30% TCO reduction and 14-month payback claims. What categories are likely omitted, and how should the comparison be framed?
Read the worked answer
Rebuild the five-year cash flows rather than accepting the headline percentage.
On-premises: $900k hardware + 5 × $380k licensing/support + 5 × (4 × $110k) labor = 900 + 1,900 + 2,200 = $5.0M undiscounted.
SaaS: $1.1M implementation + $300k parallel run + 5 × $620k subscription = 1,400 + 3,100 = $4.5M, before internal integration, training, data migration, decommissioning, and exit costs.
The gross difference is about $0.5M, or roughly 10%, not 30%. The vendor likely omitted parallel-run, implementation, internal labor, integration, data cleansing, and decommissioning, and may have compared subscription only against on-premises license plus hardware.
Payback: annual run-rate on-premises is $820k ($380k + $440k); SaaS is $620k; the difference is $200k per year. The one-time premium is $1.4M, so simple payback is about seven years, not 14 months. Even crediting downtime reduction, value it explicitly: 14 hours × $18k = $252k per year at most, and only if SaaS eliminates the outages.
Frame the decision as a risk-adjusted five-year NPV with sensitivity to implementation overrun, adoption delay, and subscription escalation. Show the assumptions that drive the answer, and separate cost reduction from cost avoidance.
2. Identify three outcome-based KPIs for this programme and explain how each ties to business value rather than IT activity.
Read the worked answer
Choose measures that would change a plant manager's or the CFO's decision, not measures of project activity.
- Unplanned ERP downtime hours per quarter and associated cost. This ties directly to the $18k/hour exposure. Target, for example, fewer than four hours per year, measured by system monitoring rather than ticket counts. It tests whether the new platform actually improves availability.
- Order-to-cash cycle time. This measures whether the ERP change improves cash flow and customer service. A migration can be 100% complete while order-to-cash gets worse; this KPI catches that.
- Inventory turns or perfect-order rate. These link the ERP to working capital and plant operations, which is where the business value sits.
A fourth option is financial-close days, which is concrete for finance. Avoid activity metrics such as 'users migrated' or 'percentage of workloads in cloud'; they can be fully achieved with zero value. Each KPI needs a baseline, target, measurement owner, and review cadence. Tie vendor fees or internal incentives to these only where attribution is credible; otherwise you pay for noise.
3. How should the company treat the 14 hours of unplanned downtime and 30 hours of planned maintenance in the business case?
Read the worked answer
Treat downtime as an expected cost with explicit assumptions, not as a footnote or a guaranteed saving.
Baseline expected cost: 14 hours × $18k = $252k per year. Planned maintenance of 30 hours also has a cost if it blocks order entry or shipping; if it occurs off-shift, value it at a lower rate and state that assumption. Do not assume planned maintenance is free.
For SaaS, use the SLA credit as a floor, not as the expected benefit. A 99.9% availability SLA permits about 8.8 hours of downtime per year, and credits rarely cover business loss. If the vendor offers 99.95%, the allowance is about 4.4 hours. Model residual risk explicitly: probability of a major outage × business cost, plus the cost of manual fallback procedures and integration monitoring.
Avoid double counting. If you claim downtime savings, remove the on-premises downtime cost from the baseline. If you keep both, the business case inflates benefits. Also include one-time resilience costs such as standby processes, training, and a tested manual workaround. The defensible statement is: 'We expect to reduce expected downtime cost from $252k to $X, with a residual risk of $Y,' not 'the cloud eliminates downtime.'