Why most ROI business cases fail before anyone reads them

Most ROI business cases don't fail in the boardroom. They fail on page one, the moment a finance director spots a single suspiciously precise number — "this project will return £417,000 a year" — with no range, no sources and no downside. From that point on, every other figure in the document is on trial.

If you have to defend numbers to a CFO, the problem is rarely your arithmetic. It is the three habits below.

Habit one: the single-point estimate

A single number says: I know exactly what will happen. Nobody does. Your churn assumption, your deflection rate, your adoption curve — each is a judgement, and judgements have error bars.

Finance people know this, which is why a lone point estimate reads as either naivety or salesmanship. The fix is not to hedge everything into mush. It is to present three explicit scenarios — conservative, moderate, optimistic — and to say which assumptions drive the spread. A case that says "worst plausible outcome is £120k, best is £480k, and the swing is mostly the adoption rate" invites a conversation about adoption. A case that says "£417k" invites a cross-examination.

There is a second benefit. When you model the conservative case honestly, you sometimes discover the project only works in the optimistic one. Better to find that out yourself than have a CFO find it for you.

Habit two: the cherry-picked benchmark

Benchmarks are where good business cases go to die, because the numbers that circulate most widely are the ones that have been mangled most.

Take the most famous retention statistic in business. "A 5% increase in retention increases profits by 25 to 95%" is routinely presented as one finding. It is actually two separate claims from two different publications: the "more than 25%" figure is about financial services (Bain, 2001), and the "up to 95%" figure is about dot-com-era e-commerce companies (Reichheld & Schefter, HBR, 2000). Blending them into a single range, then applying it to your mid-size B2B services firm in 2026, is exactly the kind of move a sceptical reviewer will catch.

Or take email marketing's beloved "$42 return per $1 spent". The verified figure is £42.24 per £1, from a UK survey in which marketers self-reported their own returns (DMA, 2019) — the dollar version is a currency-mangled copy of the pound figure. Quoting it in dollars signals you took the number from a listicle, not a source.

Even honest benchmarks get sharpened in the retelling. Walmart's page-speed study found every one-second improvement produced up to a 2% conversion increase (Walmart, 2012) — and the "up to" is routinely dropped by people quoting it. The difference between "up to 2%" and "2%" is the difference between a ceiling and a promise.

The rule: every benchmark in your case should carry its source, its year, its sector and its caveats. If you can't state where a number came from, it shouldn't be load-bearing.

Habit three: no downside scenario

Ask why a business case has no downside scenario and you'll usually hear some version of "we didn't want to undermine the proposal". This gets the psychology exactly backwards. A case with no downside doesn't look strong; it looks unexamined.

Your reviewer's first question is always some form of "what if you're wrong?" If the document already answers it — here is the conservative case, here is the assumption that would have to fail, here is what we'd see early if it were failing — you have taken their best objection off the table. You have also demonstrated the thing that actually earns approval: that you understand your own proposal well enough to know where it's fragile.

Sensitivity analysis does this systematically. Rank the assumptions by how much each one moves the result, and you learn which two or three inputs deserve real diligence and which barely matter. Most cases have one dominant assumption. Find it before your CFO does.

What a defensible case looks like

Put the three fixes together and the shape of a credible business case emerges:

• Ranges, not points. Conservative, moderate and optimistic scenarios, with the drivers of the spread named.
• An assumptions audit trail. Every input listed, every benchmark cited, every judgement flagged as a judgement. The reader should be able to disagree with a specific line, not the whole document.
• Visible arithmetic. The path from inputs to headline number should be traceable step by step, like a receipt. If the maths is hidden, the reader assumes it's hiding something.
• A follow-up plan. State what you'll measure after go-live and when you'll compare actuals to the projection. A projection you intend to check is instantly more credible than one you don't.

That last point deserves emphasis. The single strongest signal you can send a finance audience is that the business case doesn't end at approval. Committing to measure the realised impact — same assumptions, same structure, real numbers — tells them the projection was made in good faith.

None of this requires a finance degree. It requires discipline about sources, honesty about uncertainty, and arithmetic a reader can follow.

---

This is the structure LeadersToolset's calculators are built around: every result comes as conservative, moderate and optimistic scenarios with a full assumptions audit trail and a step-by-step calculation ledger, and each Projection can later be re-run as an Impact measurement against actuals. You can see the whole format in a worked sample report, read how the numbers are constructed on the methodology page, or pick a calculator and build your own case — your first calculator is free, with unlimited re-runs.