Value Selling
Why your CFO doesn't trust your ROI model
Key takeaways
- 1Finance is in more software decisions than it was a year ago. That is a buying condition, not a claim that a better case prevents a veto.
- 2A single-point estimate hides the uncertainty a finance team models for a living.
- 3Attribution is the question: how much of the change is the product, and how much is everything else.
- 4A defensible range is three readings a reviewer can open. It is a projection, not a result.
Somewhere right now, a finance reviewer is looking at a vendor ROI model for the first time. They are not admiring the formatting. They are looking for three things: where the numbers come from, whether the assumptions can be changed, and whether the method is one their own team would recognize.
31% → 46%
Finance involvement in software decisions, in a single year. The same survey says 49% of buyers reported a CFO veto of an already-approved purchase in the prior 12 months. A veto is not evidence that a better case would have prevented it.
G2 2026 Buyer Behavior Report, online survey of 1,038 B2B software decision-makers, June 2026
That rise is a buying condition. Serre does not claim to stop a veto, shorten a cycle, or raise a win rate. The useful question is whether the case can be examined at all.
The smell test
Finance teams review models for a living. Internal cases usually show stated assumptions, visible formulas and more than one reading of the outcome. A polished slide with one large number and no supporting detail does not fail because the number is certainly wrong. It fails because there is no way to tell.
An opaque model is an untrustworthy model, whatever the arithmetic would have said.
A single number pretends the uncertainty is gone
Most vendor cases present one figure. Illustrative example: "your company will save $2.4 million a year." Costs move. Adoption varies. A single number treats that as settled.
The reviewer's own models use more than one case. When yours arrives as a point and theirs arrive as a range, the gap is obvious. The fix is not a more confident point. It is a projection that says what it is.
The attribution question
Even a reasonable total leaves the question finance actually asks: how much of that is the product?
Take an illustrative claim such as a 40% reduction in time to resolution. Some of that may be the software. Some may be the process change that comes with it. Some may be the team paying attention because the work is being measured. A model that attributes all of it to the product is asking the reviewer to ignore what they know.
Serre keeps the claim, the conditions it depends on, and the evidence behind it. A person approves what enters the library. The business case is a projection of that model, not a result. Value engineering for every seller means a rep can show that reasoning without waiting on a specialist. It does not mean the number has been proven.
Say what you do not know
Wide bounds on a variable are a way of saying the uncertainty is real. They do not make the rest of the model a promise.
What transparency means
Listing assumptions on a slide is better than hiding them. It is not the same as a model the buyer can change.
- Assertion: "Based on industry benchmarks, you will save $2.4 million a year." The buyer has to take the sentence.
- Model: the formula is named, the inputs are sourced, and a number the buyer disagrees with can be replaced. The output moves because the input moved.
The first puts the burden of trust on the vendor. The second lets the reviewer try to break it. That is a different conversation. It is not a claim that the purchase will pay back.
A range, not a simulation count
A defensible range is three readings of the same model: conservative, expected and upside. Each one comes from bounds someone set on purpose. The arithmetic is deterministic. The same inputs produce the same result. A run count is not evidence, and it is not what makes the range honest.
An illustrative reading sounds like this: "In the conservative case the projection clears the hurdle you named. In the expected case it is larger. You can change either bound." The reviewer does not have to believe a single number. They can decide which set of assumptions they are willing to stand behind.
That is how a finance team already talks about uncertainty. Speaking that way does not make the case immune to a hard question.
What a reviewable model shows
The shift is structural, not cosmetic.
- Visible formulas, with named variables
- Assumptions the buyer can replace with their own
- Benchmarks that name the source, the sample and the date
- A range with explicit low and high bounds, labeled as a projection
- Value drivers tied to a specific change, not a vague productivity claim
A reviewer who can open the formula has something to interrogate. That is a different conversation from a slide that asks to be trusted. Turn every deal into a defensible case for change. The case still has to survive the questions. Serre does not claim the questions go away.
Turn every deal into a defensible case for change.
Value engineering for every seller. A person approves what enters the library.
Get Started →
