A finance transformation business case usually fails at one of two moments. It fails at approval, because a sceptical board does not believe the number. Or — more damagingly — it passes at approval and fails at realisation, eighteen months later, when someone asks what happened to the savings.
The second failure is the expensive one. A rejected case costs a quarter. An unrealised case costs the credibility a CFO needs to fund the next programme, and that credibility is not quickly rebuilt.
Both failures usually have the same root: the case presented one aggregate number, built largely from headcount assumptions, at a confidence the underlying evidence never supported.
The governing rule
Aggregate the value and you destroy the case. Finance transformation produces value from four sources that differ in kind, in timing, in who receives them and in how firmly they can be evidenced. A single blended figure is less defensible than its weakest component, because the first challenge to any part of it discredits all of it.
The four value sources
These are not categories of convenience. They behave differently enough that mixing them is the analytical error.
1 · Released cash — the strongest, and the most under-claimed
Working capital freed from the cash conversion cycle: receivables collected sooner, payment terms aligned, cash visible enough to stop being buffered defensively. This is genuine cash arriving in the business, it is largely one-time, and it is measurable against a baseline the organisation already reports.
It is the strongest source and it is chronically under-claimed, because it belongs to treasury and receivables rather than to the transformation programme, and because it is one-time rather than recurring — which makes it look less impressive in a model built to show an annual run-rate.
2 · Avoided cost — defensible, but only against a stated plan
Cost the organisation would otherwise have incurred: the finance headcount not added as the business grows, the audit escalation avoided, the system not bought because the ERP was improved instead. Avoided cost is real, and it is defensible only when the alternative is a plan of record. "We would have hired four people" is credible when four roles were in the approved plan and are now withdrawn. It is not credible as a hypothetical.
3 · Released capacity — real, but not cash until someone acts
Hours returned to the finance team by automating manual effort. This is where most business cases place most of their value, and where most realisation failures occur — because capacity is not cost. It becomes cost saving only if headcount is actually reduced, or value only if the capacity is deliberately redeployed to work that matters.
If neither happens, released capacity is absorbed silently. The work expands, the team feels slightly less pressed, and no line in the P&L moves. That is not a modelling failure; it is a management decision that was never taken.
4 · Reduced risk — the least quantifiable, and often the actual reason
Control weaknesses closed, audit findings resolved, restatement and penalty exposure reduced, a balance sheet that survives due diligence. Almost impossible to quantify honestly, and frequently the real reason the programme is being run — particularly ahead of a listing, a sale or an audit the CFO expects to be difficult.
The temptation is to quantify it anyway, using a probability-weighted cost of a hypothetical failure. Resist it. A risk number invented to fill a column is the single easiest thing for a sceptical board member to attack, and attacking it discredits the three columns that were sound.
| Source | Nature | Timing | Evidence available | What you may commit to |
|---|---|---|---|---|
| Released cash | One-time cash | 6–18 months | Reported DSO and cycle-time baselines against peer benchmarks | A range, with the baseline stated and the mechanism named |
| Avoided cost | Recurring, counterfactual | 12–36 months | The approved plan the cost sits in | A committed figure — but only where the plan of record is withdrawn in writing |
| Released capacity | Hours, not money | 3–12 months | Process measurement before and after | Hours only. Never a saving, unless a headcount or redeployment decision is committed alongside it |
| Reduced risk | Exposure avoided | Continuous | Control findings, audit history | A qualitative statement with named findings closed. Never a probability-weighted number |
The evidence ladder
The second half of a defensible case is not the size of the number but the firmness of it. Every figure should carry a grade, and the grade should govern the language used.
| Grade | What it rests on | Permitted language | Board treatment |
|---|---|---|---|
| Measured | Your own instrumented baseline — measured, not estimated | "Our close is 11 days; the target is 5" | Commit |
| Benchmarked | Your baseline against an independent peer set | "Our cost of finance is above the peer median; closing half the gap is worth £X" | Commit to a range |
| Modelled | A structural model with explicit assumptions | "If invoice volume grows as planned, this avoids three hires" | Present with assumptions visible |
| Judged | Experience and reasoning, no measurement | "Control remediation reduces the likelihood of a qualified opinion" | State qualitatively; do not size |
The discipline this imposes is uncomfortable and worth it: a number may not be promoted up the ladder because the programme needs a bigger total. If the honest grade is Modelled, the language is "if these assumptions hold" — and if the board wants a commitment, the answer is that measurement must come first.
Our benchmark reading guide covers how to place a baseline against a peer set without fooling yourself, and The Cost of Finance Is the Wrong Target explains why the most commonly-used headline metric is the weakest foundation for a case.
Sequence the case so it funds itself
The structural mistake in most business cases is not the arithmetic. It is the shape: one large ask, one large approval, one large risk, and a payback that arrives at the end.
The sequence framework we publish separately has a direct consequence here. Because capabilities have prerequisites, value does not arrive evenly — and because released cash arrives earliest and is the best-evidenced source, the case can usually be structured so that early work pays for later work.
The practical shape:
- Fund the foundation from released cash. Receivables and cash-cycle improvement need
comparatively little technology and produce measurable cash within two quarters. That cash funds the data and standardisation work nobody wants to sponsor on its own merits.
- Fund standardisation from avoided cost. Once volume growth is met without adding
headcount, the withdrawn plan is the evidence.
- Only then commit to platform spend. By this point the prerequisite is met, the case
rests on measured rather than modelled figures, and the decision archetype has moved from standardise before automating to adopt dedicated software when scaling.
A programme structured this way asks the board for a smaller initial decision, produces evidence before the largest commitment, and — critically — gives the CFO something to report at the two-quarter mark. Programmes that report nothing for four quarters are the ones that get cancelled in the third.
| Do not | Because |
|---|---|
| Present one blended number | Its weakest component discredits all of it |
| Convert released hours into savings without a headcount or redeployment decision | The capacity is absorbed and the saving never appears in the P&L |
| Use a vendor's ROI model | It was built to close a sale, and a board member will know that |
| Probability-weight a hypothetical control failure | It is the easiest number in the pack to attack |
| Promise recurring savings from a one-time cash release | It is the fastest way to lose the next approval |
| Claim benefits that require a decision nobody has agreed to take | The programme is then accountable for someone else's choice |
The strongest case against this framework is that it will produce a smaller number, and smaller numbers do not get funded. A CFO competing for capital against a revenue programme promising aggressive growth may find that a carefully-graded, deliberately-conservative finance case simply loses — and that the disciplined approach is a good way to be right and unfunded.
This is a real risk and it deserves a real answer rather than a reassurance. The answer is that the framework optimises for a different objective: not maximum approved value, but maximum realised value across a sequence of programmes. A CFO who over-promises once and under-delivers has borrowed against every subsequent case. A CFO who commits to a range and lands inside it can return.
There is a narrower objection that also lands. In organisations where capital allocation is genuinely competitive and comparably rigorous, presenting a conservative case beside optimistic ones from other functions is not honesty — it is unilateral disarmament. Where that is true, the right response is not to inflate the finance case; it is to argue for a consistent evidence standard across all cases. That is a harder conversation and a better one.
What to do with this
Take your current business case and split the total into the four sources. Most cases, when split, turn out to be seventy per cent released capacity — the source with the weakest path to a P&L line. That single observation usually changes the conversation more than any refinement of the arithmetic.
Then grade every figure. Any number sitting at Judged or Modelled that the case treats as a commitment is where the realisation failure will originate.
- Never present one number. Released cash, avoided cost, released capacity and reduced
risk differ in kind, timing and evidence. Blended, the case is only as defensible as its weakest part.
- Capacity is not saving until someone decides it is. Most unrealised business cases fail
here — the hours are freed, no headcount or redeployment decision is taken, and the P&L never moves.
- Grade every figure and let the grade govern the language. Measured and benchmarked
numbers may be committed; modelled numbers carry their assumptions; judged risk is stated qualitatively and never sized.