ROI Model

An ROI model is the calculation underneath a business case. It takes a set of costs, a set of benefit drivers, a set of assumptions and a time frame, and produces the figures a buyer will be judged on: return, payback, net present value and the cash flow by period. Its job is not to produce a large number. Its job is to be defensible when someone who did not build it takes it apart.

This entry defines the parts of an ROI model. For how to turn one into a sales ROI calculator that buyers and CFOs trust, and where it fits from discovery to post-sale, read The Sales ROI Calculator Playbook. For the buying-side scrutiny those assumptions have to survive, see our 2026 business case research.

What goes into one

Costs

Everything the buyer pays, which is rarely just the subscription.

One-off: implementation, data migration, integration work, training, internal project time.

Recurring: licence or subscription, support, additional seats as usage grows, the cost of the team that administers it.

Value drivers

The categories of benefit, split by how firmly they can be evidenced.

Hard: headcount avoided or redeployed, licences retired, hours released at a known rate, deals won at a modelled rate.

Soft: risk reduced, compliance improved, decisions made faster. Worth stating, worth keeping out of the headline number.

Assumptions

Baselines, improvement rates, ramp, adoption. This is where a model is won or lost, and where the buyer's own data belongs.

Outputs

Return as a percentage, payback in months, NPV over the term, and the cash flow profile. Time-phased, because a benefit that arrives in month thirty is worth much less than the same benefit in month three.

Three formulas

ROI = (Total benefits - Total costs) / Total costs x 100

Payback = Total investment / Annual net benefit

NPV = Sum of (Cash flow / (1 + r)^n) - Initial investment

For security investments, where the return is avoided loss rather than realised gain, the equivalent is Return on Security Investment (ROSI), explained in how to quantify cybersecurity ROI for the CFO. To see these outputs from real inputs, the free SaaS ROI calculator reports three-year ROI, payback and NPV, and the cybersecurity ROI calculator reports ROSI.

A worked example

A team evaluates a platform with a £60,000 implementation and a £120,000 annual subscription, so £180,000 in year one and £120,000 in each year after. The frame is three years and the discount rate is 10%.

Three drivers, all taken from the buyer's own figures:

Retired licences: two incumbent tools at £45,000 and £35,000, so £80,000 a year.

Productivity: 25 people release three hours a week across 46 working weeks, which is 3,450 hours. At £50 an hour that is £172,500 of capacity, counted at 60% realisation because not all released time converts to output. That is £103,500.

Revenue: 150 qualified opportunities a year at an 18% win rate is 27 wins. A 1.5 point improvement gives 29.25, so 2.25 additional deals at £40,000 each, or £90,000.

Total benefit is £273,500 a year. Year one nets £93,500, a 52% return, with payback at 7.9 months. Over three years the model produces £820,500 of benefit against £420,000 of cost, a 95% return, and at 10% the NPV is £147,200.

Now the test that matters. Drop the productivity realisation assumption from 60% to 30%, which is what a sceptical reviewer will do first. Productivity falls to £51,750, total benefit to £221,750, and year-one return to 23%, with payback moving to 9.7 months. The investment still works. That is the finding to report: the case survives a halved assumption, so the argument is about the assumption rather than about the conclusion.

Building one that holds

Start from the buyer's numbers, not from an industry benchmark, and write down where each input came from. Be conservative on improvement rates. Include the ramp, because benefits accrue after adoption and not at go-live. Show three scenarios, and make the conservative one the headline if you want to be believed. Keep the model small enough that a finance reviewer can follow it in one sitting, because a model nobody can audit is treated as an assertion.

How this shows up in a real deal

Finance asks for the model rather than the summary, opens the assumptions tab, and changes one cell. If the whole case collapses, the conversation is over and the credibility loss extends to everything else the seller has said. Models built with a single sensitivity switch, visible on the summary page, get the same treatment and produce a different conversation: here is the driver that matters, here is what happens if it is half right, and here is why we still recommend proceeding. That is the moment an ROI model earns its place in a deal.

Reading further

The ROI calculator playbook for sales teams covers the calculator layer that sits in front of a model like this. Why ROI models collapse under CFO review is about the failure modes in detail, and from narrative ROI to explainable economics is the argument for making every figure traceable. The Value Modeling Playbook is the practical build guide. For how the model fits the wider practice, read what value engineering means in B2B sales. To test a finished model against the questions finance will ask, run it through the ROI Defensibility Checker.