Guides · Decide

Technology business case: a structure finance will recognise

Most technology business cases are not rejected for being wrong. They are rejected because finance cannot tell which costs are complete, which benefits are real and who will deliver them. This guide gives you the structure, the tests finance applies, and a worked example with the numbers.

Why technology cases fail with finance

A finance team reads a business case looking for reasons to trust it. Four weaknesses come up again and again in technology cases:

  • Costs that stop at go-live. The project budget is there; the internal time, the parallel running, the subscription increases and the cost of running the new system for five years are not.
  • Benefits nobody owns. "Productivity gains" written by IT, with no business leader prepared to reduce a budget or raise a target.
  • One option. The case argues for the answer instead of showing the choice, so finance cannot see what is being traded off.
  • A single number. One NPV, with no view of what happens if the project is late or the benefits are smaller.

The structure below answers each of these. It works for a new system, a platform replacement, an infrastructure refresh or an AI use case. Scale the depth to the size of the decision: two pages for a small request, ten for a major programme, but keep the same sections.

The ten sections, and the question finance asks of each

#SectionWhat to writeThe question finance asks
1Decision requestedWhat you ask for, how much, by when, and from whom. One paragraph.What exactly am I approving?
2Problem and opportunityThe business problem in business terms, with today's facts: cost, volume, errors, delays, risk.Is this a real problem, and how big is it?
3Strategic fitWhich business goal or priority it serves, and what happens to that goal without it.Why this, and why now?
4OptionsDo nothing, do minimum, recommended, and a more ambitious option, compared on the same basis.What else could we do with the money?
5CostsOne-off and recurring costs over the whole life, internal and external, with contingency.Is this the full cost, or the first instalment?
6BenefitsEach benefit with its type, baseline, owner, measure and the date it appears in a budget.Who will deliver this, and where will I see it?
7Financial summaryCash flows by year, NPV at finance's discount rate, payback, and the effect on the profit and loss account.Does it pay back, and when does it hit my numbers?
8Risks and sensitivitiesThe main risks, what each would do to the numbers, and how they are controlled.What would make this case fail?
9Delivery planPhases, milestones, the team, and the decision points where funding is released.Can we deliver it, and when can we stop if it goes wrong?
10Measures and reviewHow success will be measured, by whom, and when the benefits will be reviewed after go-live.How will we know it worked?
Write section 1 last, and put it first. If you cannot state the decision in three sentences, the case is not ready.

Options: show the choice, not just the answer

Finance compares options, so give them options. A business case should cost at least three on the same basis and over the same period; a short committee paper needs at least two, one of them doing nothing or doing less.

  • Do nothing. Not zero cost: the cost of the current situation over the same period, including support that will expire, rising maintenance, forced upgrades and the risk you carry. This is the baseline every other option is measured against.
  • Do minimum. The cheapest way to remove the most urgent risk, for example extending support or patching the current system.
  • Recommended. The option you propose, and why it beats the others.
  • More ambitious (optional). What a larger investment would add, so the board sees what it is choosing not to buy.

Benefits and costs in the case are incremental: the difference between an option and doing nothing. Money already spent is sunk and does not belong in the decision, however painful.

Costs: the whole life, not the project

Cost the full life of the solution, usually five years for business applications and three to five for infrastructure. Separate one-off from recurring costs, and internal from external. These lines are the ones most often missing:

Cost lineOften forgotten because
Internal staff timeBusiness experts and IT staff "are already paid". Their time has a cost, and someone has to do their day job.
BackfillThe temporary staff who cover the key users seconded to the project.
Data migration and testingThey are treated as part of the vendor's work, and rarely are, in full.
Parallel runningThe old and new systems both run, and are both paid for, until cut-over is safe.
Training and changeCounted as "communication", when it is often weeks of time for hundreds of people.
Subscriptions during the buildSoftware-as-a-service fees often start at signature, months before go-live.
Price increasesSubscription and support contracts usually rise each year. Use the contract terms, not today's price.
Run cost after go-liveSupport team, hosting, integrations, security, upgrades: the system has to be run for its whole life.
DecommissioningArchiving data, ending contracts, retiring infrastructure. Without it, the old costs never stop.
ContingencyLeft out to make the case look better. Size it to the uncertainty and show it as its own line.

Capital or operating expense. Ask your controller early how each cost will be accounted for, because it changes how the case lands in the profit and loss account. Software-as-a-service fees are operating costs. Under IFRS, the configuration and customisation of a cloud service you do not control are often expensed rather than capitalised; under US GAAP, the costs of building and configuring a cloud solution are generally capitalised and expensed over the contract term within operating expense, while training, data conversion and planning costs are expensed as incurred. A case built on the assumption that implementation will be capitalised can show a profit impact several times larger than planned in year one.

Benefits finance will bank

Not all benefits are equal. Classify each one, because finance will discount them differently, and say plainly which ones the financial case relies on.

TypeExampleHow finance treats it
Cash-releasingA contract ended, licences removed, overtime no longer paidBanked, if a budget is reduced by the amount and the owner agrees
Cost avoidanceA forced upgrade or a price increase that will not happenAccepted if the avoided cost is evidenced: a quote, a contract term, an end-of-support notice
Revenue or marginHigher conversion, fewer lost orders, better pricingAccepted when the sales or business owner puts it in their targets
Capacity freedHours saved in a processNot cash unless the hours are removed or used for work that is otherwise paid for. Show it, but do not bank it
Risk reductionFewer outages, a closed audit finding, regulatory complianceHard to value in cash. Describe the exposure, and treat compliance as a requirement rather than a return

For every benefit, record six things: the baseline measured today, the measure, the target, the date it is expected, the owner (a business leader, not IT), and where it will show: which budget line or target changes. If no one will sign as owner, the benefit stays out of the numbers.

NPV, payback and the rest: what each one tells

Use cash flows, not accounting profit, and use the discount rate your finance team sets for investments. Do not choose your own.

MeasureWhat it tells youWatch out for
Net present value (NPV)Value created in today's money after paying for the capital. Positive means it beats finance's required return.Very sensitive to benefits in later years and to the discount rate
PaybackHow long until cumulative cash turns positive. Easy to grasp; often what the CFO asks first.Ignores the time value of money and everything after the payback point
Internal rate of return (IRR)The return the investment earns, to compare with the required rate.Misleading when cash flows change sign more than once or projects differ in size
Total cost of ownership (TCO)The full cost of an option over its life. Best for comparing vendors or options with similar benefits.Says nothing about value; never a case on its own
Profit and loss impactHow the case affects operating profit each year, after the accounting treatment.A positive NPV can still hurt year-one profit, which is what many executives are measured on

Worked example: replacing a warehouse system

A fictional mid-size manufacturer runs a warehouse management system that loses vendor support in two years. The recommended option is a cloud replacement. Figures are in thousands, over five years after a build year (year 0), with a 10% discount rate set by finance. They are pre-tax and nominal, with no residual value after year 5. All numbers are illustrative.

In thousandsYear 0Year 1Year 2Year 3Year 4Year 5
One-off costs−1,104
Subscription and support−220−227−233−240−248
Old system costs retired100200200200200
Decommissioning−40
Upgrade avoided300
Operational savings240480480480480
Net cash flow−1,10480753447440432
Cumulative−1,104−1,024−2711766161,048
Present value at 10%−1,10473622336301268
  • One-off costs: implementation partner (600), internal project team and backfill (200), data migration and testing (100), training and change (60), plus 15% contingency (144).
  • Subscription and support rise 3% a year, as the contract allows.
  • Old system costs retired (cash-releasing): maintenance and hosting of 200 a year stop at go-live, but the old system runs in parallel for half of year 1, and costs 40 to decommission.
  • Upgrade avoided (cost avoidance): the forced upgrade of the old system quoted at 300 for year 2.
  • Operational savings (cash-releasing): overtime and temporary labour (320) and picking errors and returns (160), half in year 1. They are owned by the head of logistics and will be taken out of the warehouse budget.

Result: NPV 496 at 10%, IRR about 25%, payback in year 3 (about 2.6 years after the build year, assuming cash flows evenly through each year). Undiscounted net benefit over five years: 1,048.

Look at the profit and loss account too. If, as is common for cloud software under IFRS, most of the 1,104 is expensed as incurred, operating profit falls by up to 1.1 million in the build year, even though the NPV is positive. Under US GAAP, the build and configuration costs are generally capitalised and spread over the contract term, while training and data conversion are expensed. Ask your controller which applies, and show the year-by-year profit impact next to the cash flows.

Sensitivities: what would make the case fail

Test the few assumptions that matter most, one at a time, and show the results next to the base case. Four for the example:

ScenarioNPV at 10%PaybackWhat it tells the decision makers
Base case4962.6 yearsThe case pays back during year 3
One-off costs 25% higher (1,380)2203.2 yearsSurvives a significant overrun
Operational savings 30% lower (336 a year)153.6 yearsOnly breaks even: the case depends on the logistics savings, so their owner must commit to them
Go-live one year late−1914.4 yearsDestroys value: delay is the risk to control
Late, and the upgrade can no longer be avoided−439Not within 5 yearsIf the delay also forces the upgrade, the loss more than doubles

In the delay scenarios, the old system's costs, parallel running, decommissioning and the operational savings all move one year later; the subscription does not, because it starts at signature; the project team costs 110 more in year 1. In the first delay scenario the old system can still run without the upgrade until the late go-live; in the second it cannot.

This is the most useful page in the case. It tells the decision makers where to put their attention: here, a firm savings commitment from logistics, subscription fees that start at go-live rather than at signature, and a fixed-scope first release. A case that shows its own weak points is more credible than one that hides them, and it lets you agree the controls before the money is spent. The business case workbook in the Project Delivery Framework Kit calculates the NPV, payback and three sensitivities as you type.

The one-page summary

Executives will read one page. Put this on it, in this order:

  1. Decision requested: the amount, what it buys, and the date.
  2. Why: the problem in one sentence, with one number.
  3. Options considered, and why the recommended one wins.
  4. Money: one-off cost, change in annual run cost, NPV, payback, profit impact in year 1.
  5. Benefits, with their owners and the budgets they will change.
  6. Key risk from the sensitivities, and how it is controlled.
  7. Next decision point: when the board will see the case again, and what will have to be true by then.

Twelve tests before you submit

DoneTest
☐The decision requested fits in three sentences.
☐The problem is described with today's numbers, not adjectives.
☐"Do nothing" is costed over the same period as the other options.
☐Costs cover the whole life, including internal time, parallel running, run cost and decommissioning.
☐Contingency is shown on its own line and sized to the uncertainty.
☐Your controller has agreed which costs are capital and which are operating.
☐Every benefit has a baseline, a measure, a date and a business owner who has signed.
☐Capacity freed is shown but not counted as cash.
☐NPV uses finance's discount rate; sunk costs are excluded.
☐At least three sensitivities are shown, including a delay.
☐Funding is released in stages, with a decision point at each.
☐A benefits review date after go-live is set, with its owner.

Questions

How long should a technology business case be?

As short as the decision allows. Two pages are enough for a small request; a major programme may need ten, plus appendices for the cost and benefit detail. The one-page summary matters more than the length.

Which discount rate should I use?

The one your finance team sets for investment decisions, often called the hurdle rate. Using your own rate, even a reasonable one, gives finance a reason to redo your numbers.

How do I make a case for security or compliance work?

Present it as a requirement with options, not as a return. Describe the exposure and its consequences, show the cheapest option that meets the requirement and what more money would add, and compare the options on cost. Do not invent a return on risk you cannot measure.

Can I count staff time saved as a benefit?

Show it, but count it as cash only if the hours are removed from a budget, avoid hiring that is otherwise planned, or are moved to work that is otherwise paid for. Finance will discount it otherwise, and your credibility with it.

What about an AI use case?

Same structure, with two additions: a pilot stage with a clear go or no-go test before full funding, and the running costs that grow with usage, such as model and compute charges. Value the benefits on the pilot's measured results, not on vendor claims.

Toolkits on this topic

Related