Seven fields, in Word, Excel and Google Docs, with no email form in the way. Here is what the template does, and the five points where it stops.
This is a skeleton for the document a seller writes when a buyer asks why they should pay for software. Seven fields, about three pages once the tables are filled in. The Word, Excel and Google Docs versions carry the same seven fields, so you can start in whichever one your team already uses.
It is not a business case. A case is a set of numbers that belong to one buyer, with sources, ranges and a named owner. A template gives you places to put those numbers. It holds no facts about your deal on its own, and it cannot tell you whether the numbers you put into it will survive a finance review.
That distinction is the argument of this page. Take the template. It is genuinely useful, and it is free. Then read the fourth section, which is about the parts of the job a document cannot do.
All three files carry the same seven fields. Pick the one that matches where the work will live.
You need to send the case to someone else. Prints on three pages, opens anywhere, survives email.
The numbers will change more than once. Counts internal hours as a cost line and totals year one.
The case will be edited by several people. Upload it to Drive and Drive turns it into an editable Doc.
No form and no email address: the three links go straight to the files. For the Google Docs version, choose File → Open → Upload in Drive and pick the downloaded file; Drive converts it to an editable Doc.
Seven fields, in the order a finance reviewer reads them rather than the order a salesperson writes them.
The decision being requested
One sentence: what is being approved, how much, and by whom. If it needs a paragraph, the request is not settled yet.
Cost of the current state
One line per cost, and beside each line the system, report or person the number came from. Where there is no source, write estimate and say so in the field.
What changes with the software
A range, not a point. If the team saves eight hours a week, write five to eleven and say what drives the spread.
Total year-one investment, including internal effort
License, implementation, integration, training, and the hours your own people will spend. A $60,000 license that needs 400 internal hours at $75 an hour is a $90,000 first-year cost, and someone in finance will do that arithmetic whether or not you do.
Payback period and three-year view
The month cumulative benefit crosses cumulative cost, and the three-year total under the base case.
Assumptions, listed and dated
One per line, with where it came from, who owns it, and the date it was last checked.
Risk, and what would make this wrong
The two or three things that would break the case, what each would do to payback, and the signal you would see first. Not a disclaimer.
Filling all seven takes an afternoon when the numbers already exist. If it takes three weeks, the problem is not the document.
A template is the thing people ask for and the thing that most often lets them down. Not because the structure is wrong — the seven fields above are the right fields — but because a document cannot do five parts of the job.
No. A blank field asks to be filled, and when the baseline is not to hand the quickest route to a finished document is an estimate that looks like a fact: an average ticket volume, a plausible hours-saved figure, a license count copied from the last deal. The template will not object. Finance will, usually with one question — where did this number come from? From that point the conversation is no longer about the size of the benefit. It is about whether the person who wrote the document can be trusted with numbers. A filled box and a sourced box look identical on the page, which is why the source column in field 2 matters more than the total at the bottom of it.
Why spreadsheet business cases failNo. A template is filled in private, so nothing in it is tested until the review. The numbers that survive finance are usually the ones the buyer side supplied: ticket volume from their service desk, invoice count from their accounts payable system, the loaded hourly rate their finance team uses for internal time. A document cannot pull those in, and it cannot record which figures the buyer confirmed and which the seller assumed. That is a data problem wearing the costume of a writing problem, and no layout fixes it.
The state of B2B business cases in 2026Only the appearance of one. A single column of numbers is a claim; three columns — conservative, base and upside — are an argument, because they tell a reviewer which inputs the answer depends on. A template can give you three columns, but it cannot derive them. The spread comes from the assumptions behind the model, and a template has no model: move adoption from 60% to 40% in a Word table and every downstream number has to be recalculated by hand, which in practice means it is not recalculated at all.
Why ROI models collapse in reviewPartly. It can list them in field 6. What it cannot do is show which benefit line depends on which assumption, so that a reviewer can trace a number on the first page back to the assumption that produced it. That trace is what a CFO means by a defensible case. Where the trace is missing, the reviewer does the safe thing: assumes the worst about the largest line and discounts the rest. Listing assumptions is not the same as exposing them, and only one of the two changes a decision.
ROI Defensibility CheckerSilently. A template has no version, no owner and no record of what changed. The case that went to the CFO in March is not the case in June, once the pilot shrank from four teams to two and the license count grew to cover a team nobody had mentioned. Two files with the same name and different numbers is how a quarter disappears: budget was approved against one set of figures, delivery is measured against another, and nobody can say which version the committee actually saw.
The CFO-Proof Business Case guideNone of this is an argument against the template. It is an argument for knowing which part of the job it does, and for having something in place for the other five.
Take the document away and a defensible case is four properties. All four are about where the numbers came from, and none of them is about how the page looks.
Every baseline should be traceable to a buyer system, a buyer document or a named person on the buyer side, with the date it was taken. Vendor benchmarks are not baselines. A benchmark can tell you the shape of a benefit; it cannot tell you what this organization spends today, and finance knows the difference.
An appendix full of assumptions is better than nothing and worse than a case where each number carries its own. A reviewer should be able to point at any figure and see the assumption behind it, the person who owns that assumption, and when it was last confirmed. The owner matters: an assumption with a name attached gets checked, and one without gets inherited.
Conservative, base and upside, with the two or three inputs that move the answer most. The point of the range is not precision. It is to show which assumption the decision actually turns on, so a committee can argue about the right thing instead of the payback decimal.
Every revision recorded, with what changed and why, so the version approved is the version delivered against. Deals move; a case that cannot show its own history will be rebuilt from scratch at the first change, and the second rebuild is where champions lose confidence.
A template gets you a document. A governed model gets you a decision. The same seven fields, filled in a browser session, where each number keeps its source attached, the assumptions sit on a record with an owner and a date, and the case comes out with conservative, base and upside views instead of one column.
If you already have a draft, the fastest next step is the free business case readiness diagnostic: nine questions, three minutes, and a list of the gaps finance will find before finance finds them.
What we found when we looked at how software cases are actually built and reviewed.
The seven ways a model breaks under questioning, and what survives instead.
The failure mode that has nothing to do with the math being wrong.
Sorting assumptions by evidence quality, and defending each one on three layers.
Nine questions, three minutes, a score out of 27 and the gaps to fix first.
Where the improvement rates came from, and how much the return depends on them.
The commercial fork: keep the workbook, or move the case into a governed system.