Economic Analysis of a Payroll Consolidation: NPV, Risk, and Inherently Governmental Work
A defense finance center processes 2.4 million travel vouchers annually across three sites. The current operation costs $46.0 million a year, averages 12.4 days per voucher, and has an 8.1 percent error rate; rework accounts for $3.1 million of the annual cost. The commander has directed an economic analysis of three alternatives over a five-year horizon:
- Alternative A: status quo, $46.0 million per year.
- Alternative B: consolidate to one site with workflow automation. One-time investment $21.0 million; annual operating cost $32.0 million once fully implemented. Savings begin in year 2 at 50 percent and reach the full amount in year 3. Disruption risk during the two-year transition.
- Alternative C: transfer processing to a commercial shared-service provider for $33.5 million per year plus $2.5 million in transition costs, with savings realized from year 1. The provider requires that its own staff perform all voucher review and approval.
The command applies OMB A-94 guidance and a 3.5 percent real discount rate. The certifying officer function is currently performed by government employees at each site.
1. Compute simple payback and a five-year discounted comparison for Alternatives B and C. Show your work.
Read the worked answer
Discount factors at 3.5 percent: year 1 = 0.966, year 2 = 0.934, year 3 = 0.902, year 4 = 0.872, year 5 = 0.843; sum = 4.517.
Alternative B. Annual saving when fully implemented = $46.0M − $32.0M = $14.0M. Savings stream: year 1 $0, year 2 $7.0M, years 3-5 $14.0M each.
- Discounted savings = ($7.0M × 0.934) + ($14.0M × (0.902 + 0.872 + 0.843)) = $6.54M + ($14.0M × 2.617) = $6.54M + $36.64M = $43.18M.
- Net present value = $43.18M − $21.0M = $22.18M.
- Simple payback = $21.0M ÷ $14.0M = 1.5 years of full-rate savings, or about 2.5 years from project start once the ramp is counted.
Alternative C. Annual saving = $46.0M − $33.5M = $12.5M, in all five years.
- Discounted savings = $12.5M × 4.517 = $56.46M.
- Net present value = $56.46M − $2.5M = $53.96M.
- Simple payback = $2.5M ÷ $12.5M ≈ 0.2 years.
On the numbers alone, C wins by roughly $31.8 million in NPV and requires far less capital. Any recommendation that ignores this gap has to justify the gap on grounds the analysis does not capture - which is exactly what the next question is about.
2. Alternative C has the better NPV. What factors would cause you to recommend against it or to restructure it, and how would you present that to the commander?
Read the worked answer
The decisive issue is the certifying officer function. Voucher certification and payment approval are inherently governmental: the government must retain the authority to certify that a payment is legal, proper, and correct, and cannot delegate that determination to a contractor. Alternative C as written requires the provider's staff to perform all review and approval, which is not an arrangement the government can accept regardless of its NPV. That single fact disqualifies C as proposed, and it is a legal constraint, not a preference.
Second-order concerns. Concentration risk: a single provider with no in-house capability leaves no fallback if performance fails. Data and privacy exposure for travel records. Elasticity: the contract price is per-voucher and volume may fall as the automation matures, so the fixed-fee structure could become expensive. Contract administration cost is not in the $33.5 million.
Restructured C. Retain a government certifying officer at each site or a consolidated certifying office, contract only the transaction-processing and audit-support work that is commercial in nature, and price per transaction with a volume band. Recompute the NPV with the retained government cost added - the $12.5 million annual saving will shrink, and the honest comparison may reverse.
Presentation to the commander. Lead with the constraint that eliminates the option as proposed, then show the recomputed comparison, then show B's advantages: it preserves in-house capability, converts $3.1 million of rework into a measurable target, and improves cycle time and error rate - outcomes the analysis should value explicitly rather than bury in operating cost. Recommend B with a phased transition and a stage gate after year 1.
3. The commander asks how the 8.1 percent error rate and 12.4-day cycle time should be treated in the analysis, and whether the $3.1 million of rework is already counted.
Read the worked answer
It is already counted, once - and that is the trap. The $3.1 million of rework sits inside the $46.0 million baseline and therefore inside the $32.0 million projected cost for B. If B is credited with eliminating rework on top of the $14.0 million saving, the same dollars are counted twice and the NPV is overstated. The correct treatment is to state what portion of the $14.0 million comes from rework reduction and to support it with a driver-based estimate: error rate falling from 8.1 percent to a target rate, times the average rework cost per error.
Error rate and cycle time are outputs, not savings. They should appear in the analysis as non-monetized benefits with quantified targets and, where possible, a monetized proxy: faster cycle time reduces the float the command carries and lowers the volume of interest payments and inquiries; lower error rate reduces improper payments, which carries its own reporting exposure under the Payment Integrity Information Act.
Use them as the measurement plan. Whatever option is chosen, the business process improvement effort (a DMAIC-style project is appropriate) should baseline 8.1 percent and 12.4 days, set a target - for example, under 3 percent and under 5 days within 24 months - and track them monthly as leading indicators that the projected savings are actually materializing.
Sensitivity. Run the analysis at a 15 percent volume decline and at a 20 percent error-rate improvement rather than the assumed reduction. If B's NPV turns negative under plausible volume decline, the recommendation should include a scalable staffing and automation design rather than a single-site monolith.