Writing a Business Case Finance Will Sign

The first draft of Meridian's business case had a headline figure: "$1.8 million a year in productivity benefit, payback in four months." The finance partner read it and asked one question. "Which budget line goes down by $1.8 million?" None did. The figure multiplied 46,000 saved hours by $40 an hour, but the 450 staff would still be paid whether or not they saved time. Saved minutes scattered across a working day do not become cash on their own.

That question is the difference between a business case that gets politely filed and one that gets signed. Finance does not doubt that saved time is valuable. It needs to know which part of the value will change a real number in a budget, when, and how sure the team is.

This lesson rebuilds Meridian's case the way finance reads one: hard and soft benefits kept apart, options compared honestly including doing nothing, sensitivity on the assumptions that matter, and funding released in stages against evidence. It produces part B of MER-11.

Hard and soft benefits

Meridian had a hard benefit available, and the first draft missed it. Hardship volume is growing 30% a year, and operations had already planned to hire 24 more staff next year to absorb it. If the assistant creates capacity, some of those hires are not needed. That is a budget line.

To turn hours into avoided hires, divide by productive hours per person, about 1,700 a year, and then apply a realisation factor: the share of theoretical capacity that actually becomes usable. Minutes saved in fragments across a day never all combine into whole people. Meridian's finance team uses 60% for this kind of change. So 46,000 hours is about 27 people of theoretical capacity, and about 16 avoided hires in practice. At a loaded cost of $62,000 each, that is about $992,000 a year.

The second hard benefit is fewer bad letters. About 17,500 letters a year go out. Moving the quality failure rate from 7% to the 2% target avoids about 875 failures, and each costs about $180 in rework, complaint handling and remediation: about $158,000 a year.

Options, including doing nothing

A business case with one option is a request, not a case. Meridian compared three. All figures are in thousands of dollars; net present value (NPV) is over three years at a 10% discount rate, with year-one benefits at 60% while adoption ramps up.

The result surprised the team. Option C has a higher yearly net benefit but a lower three-year NPV than B. Letters add 14 of the 184.5 daily hours saved, so they barely change the hiring figure. They add $264,000 of build and $120,000 a year of run cost, most of it review drills, template upkeep and compliance time. Over three years, letters alone have an NPV of about minus $95,000; over five years they turn slightly positive.

An honest architect puts that in the case rather than burying it. The time argument for letters is weak. The argument for letters is the 7% of customers in financial difficulty who today receive a letter with a wrong figure or missing mandatory wording. That is a conduct and customer-outcome argument, and it is the risk committee's call, not finance's.

Sensitivity

Finance wants to know which assumptions could break the case. Change one at a time and show the steady-state net benefit for option C.

The table tells finance where to look. Token price, the thing engineers worry about most, barely matters. Adoption and realisation matter enormously. So the case should spend its words, and the plan should spend its money, on adoption: training, the review design, the Atlas panel's usability, and team leads who use the assistant themselves.

Stage gates

Because adoption drives the case and letters carry the most risk, Meridian's recommendation releases funding in stages.

1. Stage 1, approve now — build option B: policy answers and summaries. Release 1 at month 7.
2. Gate 1, three months after launch — at least 70% of staff using it weekly and the correctness objective met. If adoption is under 50%, pause and review before further spend.
3. Stage 2, letters — released only if gate 1 passes and a letter pilot on 300 drafts shows quality failures at 3% or less with a review drill catch rate of 90% or more.
4. Gate 2, six months after letters launch — letter quality failures at 2% or less in the existing QA sample, or letters return to template mode.

Stage gates turn "trust us" into "fund us one step at a time". They also give the sponsor something valuable: a clean point to stop, with little lost, if the evidence says stop.

With stage 1 funded, section 12 records the risks that remain, who owns them, and how the system is governed once it is live.

Check your understanding

0 of 3 answered

1.Why did the finance partner reject "$1.8 million a year in productivity benefit"?

AThe hourly rate was too highBIt multiplied saved hours by pay, but no budget line would actually fall by that amountCProductivity benefits are never allowed in business casesDThe hours saved were not measured

2.Option C has a higher yearly net benefit than B but a lower three-year NPV. What should the architect do?

ARemove letters from the design without telling anyoneBInflate the letter benefits so C winsCShow it plainly and explain that letters are justified mainly by customer outcomes and error reduction, a decision for the risk committeeDRecommend option A instead

3.The sensitivity table shows adoption matters far more than token prices. What should the plan do with that finding?

ANegotiate harder on token pricesBInvest in adoption, such as training, review design and panel usability, and gate funding on measured weekly useCIgnore it, because sensitivity tables are only for financeDAssume 100% adoption to simplify the case