The ROI case for the finance director

An ROI case sets out what the buyer spends and what the buyer gets back, in the buyer's terms. It's a calculation she can check, not a brochure.

The first rule: build it from the buyer's own numbers. A figure from an article ("language training returns several times its cost") is the first thing a finance director crosses out: whose training, in which company, measured how? The same goes for Klasna's best-ever client result, or neat round numbers chosen because they're easy to remember.

For Veloria the inputs come from inside Veloria: Stepan's incident log, Pavlo's estimate of what an hour of a dispatcher's time costs, the translation bills. Each is labelled with who gave it.

That labelling does two jobs. Yaryna can check every line against her own records, and if a number turns out wrong, it's corrected together instead of undermining the whole case.

Klasna's invoice is only part of what a program costs the buyer. When employees learn during paid hours, their time is a cost to the company, even though no extra invoice arrives.

For Veloria: 300 lessons a month for twelve months is 3,600 lesson hours a year. About half fall in paid hours, so 1,800 hours. An hour of a dispatcher's time costs Veloria roughly 300 UAH with taxes (Pavlo's estimate). That's 1,800 × 300 = 540,000 UAH a year.

It's tempting to leave this out, because it makes the program look more expensive. Don't. A finance director will add it anyway; if you add it first, she sees that you understand her side and stops looking for what you left out.

Use the buyer's figure for the hourly cost and say whose it is. And count only the hours that really fall in paid time: lessons in a learner's own time don't cost the company those hours.

Total cost of ownership is everything the buyer spends to make the program work, not just the seller's invoice.

For Veloria's year:

• Klasna's fee: 300 lessons × 12 months × 800 UAH = 2,880,000 UAH;
• dispatchers' paid hours in lessons: 1,800 × 300 = 540,000 UAH;
• Pavlo's coordination and the IT check: about 80,000 UAH.

Total: 3,500,000 UAH.

Two things don't belong here. Klasna's tutor payouts are Klasna's cost, already inside the fee; adding them would count them twice. The translation bills Veloria hopes to cut are on the benefit side: they're what the program may reduce, not what it costs.

Showing the full cost first can feel like making the deal harder. In practice it makes it easier: the finance director's own version would include these lines, and a case that already has them is one she can take to her board unchanged.

Benefits depend on how much the training actually removes, and that's an assumption. So give a range, not one number.

Veloria's yearly costs of poor English, from its own figures:

• incidents: about 8 a month at roughly 50,000 UAH each = 4,800,000;
• lost hours: 40 people × 3 hours a week × 48 weeks × 300 UAH = 1,728,000;
• translation: 50,000 a month = 600,000.

Training won't remove all of it. Agreed with Stepan and Pavlo: the low case cuts incidents and translation by half and lost hours by a third; the high case by 80% and two thirds.

Low: 2,400,000 + 576,000 + 300,000 = 3,276,000. High: 3,840,000 + 1,152,000 + 480,000 = 5,472,000.

Against the total cost of 3,500,000, the net result is −224,000 to +1,972,000 UAH. Subtract the full cost of ownership, not just the fee, and keep both ends of the range.

Two numbers summarise the case, and they answer different questions.

• ROI = (B − C) ÷ C says how much comes back per hryvnia spent, net of the cost. B is the year's benefit, C the total cost of ownership.
• Payback = C ÷ (B ÷ 12) says how soon, in months, treating the year's cost as one budget.

For Veloria, C = 3,500,000:

• low case (B = 3,276,000): ROI = −224,000 ÷ 3,500,000 = −6.4%; payback 12.82 months;
• high case (B = 5,472,000): ROI = 1,972,000 ÷ 3,500,000 = +56.34%; payback 7.68 months.

A program can have a decent ROI yet pay back slowly, or the other way round, so show both. And watch the formula: ROI divides the net gain by the cost; dividing the benefit itself by the cost (5,472,000 ÷ 3,500,000) gives 156%, which overstates the return by the whole cost.

When the low case comes out negative, it's tempting to show only the middle and leave it out. Don't. A finance director will find it, and then she'll doubt every other line.

Shown openly, the weak case is useful. It tells her what to watch: Veloria's program pays back within the year unless the incident line moves less than both sides hoped. That's a specific, checkable risk, not a vague worry.

It also changes how the case reads. One number ("ROI 25%") is a guess with decimals. A range with its reasons ("−6.4% to +56.34%, depending mostly on incidents") is something she can defend to her board, because she can explain both ends.

When you round for the summary, round honestly: −6.4% is "about −6%", not "about zero". Keep the exact figures in a table under the summary.

An ROI case uses a few terms, and each answers one of the finance director's questions:

• Total cost of ownership: everything Veloria spends to make the program work, not just the invoice (3,500,000 UAH).
• ROI: how much comes back per hryvnia spent, net of the cost (−6.4% to +56.34%).
• Payback period: how many months until the benefit covers the cost (12.82 to 7.68).
• The range: how far the result moves with the assumptions (3,276,000 to 5,472,000 of benefit).
• Flagged assumptions: which inputs are estimates, and who gave them (300 UAH an hour from Pavlo, 8 incidents a month from Stepan's log, the cut shares agreed in the meeting).

ROI and payback answer different questions, so neither replaces the other. The range and the flagged assumptions tell her how much to trust both.

Not every assumption matters equally. Find the one that decides the case: look at how far each line swings between the low and high cases.

For Veloria: incidents swing by 1,440,000 UAH (2,400,000 to 3,840,000), lost hours by 576,000, translation by 180,000. Incidents alone move more than the other two together (756,000). So whether the program pays off depends mostly on how far language-related incidents fall.

That tells you what to measure first: incidents in Stepan's log, reported every month next to attendance, from the first quarter. Then Yaryna can watch the case come true or not, and a weak trend shows up early, while there's still time to adjust the program. Learner satisfaction matters, but it isn't in the case; translation bills are easy to count but barely move the result.

How it fits together: the buyer's numbers, total cost of ownership, benefits as a range, ROI and payback, the weak case shown, the key assumption measured. Next lesson: what a discount really costs Klasna, and what to ask for in return.