Build or buy a business case platform: how to decide
When building your own ROI calculator is the right call, when it is not, and the cost most teams leave out of the comparison: maintaining the value model.
The build-or-buy question for business cases is usually asked about the calculator, which is the cheap part. The expensive part is keeping one value model current and consistently used across a whole sales team for years, and that is the cost most build estimates leave out entirely.
The thing you are actually building
Ask an engineer to build an ROI calculator and you will have something working in a week. Inputs, assumptions, a formula, a presentable output. With a coding agent, faster.
That is not the product. The product is the answer to all of the following, two years from now, with forty people depending on it:
- When pricing changes, who updates the assumptions, and how do reps find out?
- When a rep disagrees with a default, can they override it, and does anyone see that they did?
- When a CFO asks where the 23% figure came from, can anyone reconstruct it?
- When the champion leaves and their replacement asks for the original case, does it still exist?
- When you want to know which arguments actually won, is the data in a shape that answers?
A calculator is a weekend. The above is a system, and it keeps needing attention long after the person who built it has moved to something else.
Applying the four axes
Minoa's own build-vs-buy framework scores a decision on robustness floor, lifespan, strategic differentiation and reversibility. Run this particular decision through them and the answers are unusually clear.
Robustness floor: high. The output goes to a customer's CFO. A spreadsheet that is wrong in the fourth decimal place is a nuisance internally and a credibility problem externally. Things with customer-facing failure modes have high floors, and high floors are where the last 20% of the build gets expensive.
Lifespan: long. If value selling works for you at all, you will be doing it in five years. Short-lived things are good candidates to build. This is not one.
Strategic differentiation: low, with one exception. How you calculate ROI is not your moat; what you sell is. The exception is a genuinely unusual value model, for example regulated pricing or unit economics nobody else shares, where off-the-shelf assumptions do not fit. That exception is real and it is rarer than the teams invoking it believe.
Reversibility: asymmetric. Buying and leaving costs you an export and a migration. Building and abandoning costs you every case built in the interim, because they live in a format nothing else reads. The asymmetry runs against building.
Three of four point the same way for most teams. Which is exactly why the honest cases for building are worth stating plainly.
When building is the right call
You have fewer than about ten sellers and one person builds every case. At that size the consistency problem does not exist yet, because consistency is one human being. Buy when the bottleneck becomes that person, not before.
The calculator is a marketing asset, not a deal tool. A public ROI estimator on your website that captures leads is a different product with a much lower robustness floor. Build it.
Your value model is genuinely unusual. Regulated pricing, consumption economics nobody else shares, a metric specific to your industry. Test this claim honestly: ask whether a competent outsider could follow your model from a one-page description. If yes, it is not unusual.
You are still learning what to measure. If you do not yet know which drivers matter, a spreadsheet is the correct instrument for finding out. Commit to a platform once the model stops changing weekly.
When buying wins
Past roughly twenty to thirty sellers. This is the most reliable signal. Below it, variance between reps is manageable by conversation. Above it, every rep develops a private version of the model and nobody can tell which numbers a customer was shown.
When the case has to survive into the renewal. A pre-sale calculator produces a document. Proving delivered value eighteen months later needs the original assumptions retrievable and comparable against actual outcomes. Almost no homegrown calculator is built for retrieval, because retrieval is not a problem on the day you build it.
When you need governance without becoming a bottleneck. The requirement is that reps move fast and claims stay defensible. That means defaults someone owns, overrides that are visible, and an audit trail. It is buildable and it is a meaningful amount of work.
When the person is the bottleneck. If deals wait on one value engineer's calendar, the constraint is capacity, and tooling that only that person can operate does not relieve it.
The cost comparison people actually run, and the one they should
The common comparison is a vendor's annual price against a few weeks of engineering. That comparison is wrong in both terms.
It overstates the build by ignoring that the first version is genuinely cheap now. It understates the build by ignoring years of maintenance, the internal support burden, the cost of the quarter where it silently produced wrong numbers, and the opportunity cost of the engineer who is now the owner. The full cost breakdown is its own piece, because it is the part of this decision that is consistently estimated worst.
A cleaner test: if this had been running for three years and broke tomorrow, who fixes it, and how long until the next deal has a case? If you cannot name the person, you have not built a system. You have built a dependency.
Where AI changes the answer and where it does not
Coding agents make the first 80% of the build cheaper, which is real. They do not lower the robustness floor, shorten the lifespan, or make the thing more reversible. The axes that favour buying are precisely the ones AI does not move.
Using a general model to draft the cases themselves is a separate question with a separate answer, covered in what ChatGPT and Claude can and cannot do here. The short version: drafting is the part they are good at, and consistency across a team is the part they are not.
FAQ
Is it cheaper to hire a value engineer or buy value selling software?
For a single hire covering a large team, software is usually cheaper on a three-year view, but that framing misses the point. A value engineer and a platform solve different halves of the problem. The engineer decides what is worth measuring and how hard to push attribution, which is judgement work. The platform removes the repetitive assembly that stops one engineer from covering forty sellers. Teams that replace the person entirely tend to produce cases that are consistent and unconvincing. Teams that buy nothing tend to produce a bottleneck with a calendar.
We have a spreadsheet that works. Why change?
If it genuinely works, do not. The signals that it has stopped working are specific: reps keeping private copies, nobody able to say which version a given customer saw, the file breaking when its author is away, and nobody able to answer which value arguments won. If none of those are true, your spreadsheet is fine and this decision can wait.
How long does it take to build an ROI calculator reps will actually use?
Days to build, months to get used. The gap is the whole problem. Adoption depends on the calculator being faster than the rep's own workaround, correct on the edge cases that matter in their segment, and maintained when pricing changes. Most internal builds clear the first bar and fail the third, which is when reps quietly return to private spreadsheets.
What is the real risk of letting reps build their own ROI numbers?
Two distinct risks. The commercial one is that a number you cannot reconstruct ends up in front of a CFO, who finds the flaw before you do. The compounding one is that you lose the ability to learn: if every case is bespoke, no pattern is visible across them, and the hundredth business case is no sharper than the first. The second risk is slower and more expensive.
At what team size should we stop building and start buying?
There is no clean threshold, but the useful signal is not headcount. It is whether anyone can answer "which numbers did we show this account, and who approved them" without opening a file and asking a colleague. Teams usually lose that ability somewhere between twenty and thirty sellers, and the loss is gradual enough that it goes unnoticed until a deal is damaged by it.
Ready to get started? Book a demo to see Minoa in action.