What a disciplined version includes
| Component |
What belongs in it |
The shortcut to watch for |
| Cost boundary |
Licence or build cost, implementation, internal time, training, integration, run costs, exit costs |
Costs stop at the contract price; internal effort is treated as free |
| Benefit classes |
Hard savings, revenue effects, risk reduction, capacity released, each with an owner |
Soft benefits monetised aggressively and counted as if banked |
| Baseline |
A realistic "do nothing" case, including the genuine cost of not acting |
The baseline is flattered or catastrophised to make the delta look bigger |
| Time and discounting |
Cash flows over a stated horizon, discounted; sensitivity on the key assumptions |
A single-point estimate over an optimistic horizon, no sensitivities |
| Cost to achieve benefits |
Change management, redundancy costs, parallel running, productivity dip during transition |
Benefits counted gross, as though the transition were frictionless |
When this framework is the right tool
Cost-benefit analysis suits decisions with a genuine yes/no or invest/do-not-invest shape: a system replacement, an outsourcing move, an automation programme, a market exit. It is weaker for comparing dissimilar options against strategic criteria (that is weighted scoring territory) and close to useless for decisions whose value is primarily optionality, where forcing everything into annual cash flows strips out the point of the investment.
An honest illustration
A distribution business considers automating its returns processing. The advocate's version counts the licence, the integrator quote and headcount savings from year one. The disciplined version adds what the first one omitted: six months of parallel running, a productivity dip while the team relearns the process, the internal project team's time, and the redundancy costs required before any headcount saving is real. The benefit side is trimmed too: capacity "released" is only a benefit if the hours are redeployed or removed. The project still clears the bar, but two years later than the original deck claimed, and now the board knows the true shape of the commitment it is approving.
The standard distortions
The same bends appear in cost-benefit cases across every sector, and they nearly all push in the proposal's favour.
- Benefits arrive gross of the cost of achieving them: the transformation is priced, but the disruption it causes is not.
- Soft benefits do the heavy lifting: "productivity uplift" and "improved decision-making" are monetised at rates nobody will ever be asked to bank.
- The do-nothing baseline gets gamed. When the sponsor needs a big delta, doing nothing becomes a catastrophe; when a rival proposal threatens, doing nothing becomes surprisingly viable.
- Sensitivity analysis is either absent or ornamental. The case is presented at the single point where it clears the hurdle rate, and nobody shows the version where benefits arrive a year late.
- Optimism compounds: each individual assumption is defensible at the ninetieth percentile, and the multiplied result is a case that was never realistic as a whole.
Questions the approver should ask
- Which single assumption, if moved 20 per cent, kills this case, and what evidence supports it?
- Who owns each benefit line, and will their budget be reduced when it lands?
- What does the case look like if benefits arrive twelve months late? Most do.
- What are the exit costs if this fails at the halfway point? Or is this effectively irreversible once committed?
Why an outside review changes the arithmetic
Everyone who touched the case inside the organisation has a relationship with its answer: the sponsor needs it approved, finance has negotiated the assumptions, the vendor supplied the benefits benchmarks. Selected senior operators who have delivered comparable programmes can tell you what the benefits curve actually looked like, where the hidden costs sat and how long the productivity dip lasted. That review costs a fraction of approving an overstated case.
