Software ROI is one of those measurements that everyone demands and almost nobody does well. The cost side is easy. The benefits side is messy: revenue lift is contingent, productivity gains are partial, optionality is genuine but hard to monetize, and the time horizons that matter for software are usually longer than the planning windows the organization runs on. A workable ROI framework has to accommodate that messiness rather than paper over it.
The Four-Bucket Frame
MIT Sloan's CISR research codifies a useful frame: software investment returns fall into four buckets — cost reduction, revenue acceleration, risk reduction, and option value. Each is measurable. Each requires a different measurement technique. And each gets undercounted in different ways when teams default to a single ROI number.
Cost Reduction Is The Easiest And The Smallest
Cost reduction is the bucket that fits the spreadsheet cleanly: hours saved, FTEs avoided, infrastructure decommissioned. Useful, measurable, and almost always the smallest of the four when the program is good. Programs that frame their entire ROI on cost reduction tend to underbuild — they leave the harder buckets on the table.
Revenue Acceleration Requires Causal Discipline
Revenue lift is the bucket where measurement gets interesting. The right technique is causal: experimentation where possible, holdouts and difference-in-differences where not. Harvard Business Review's analytics archive has a useful body of writing on the discipline of attribution and counterfactuals. The shortcut of attributing revenue to whatever launched recently is how programs claim wins they did not create.
Risk Reduction Is Quietly Underrated
Risk reduction is the most underrated bucket because the unit of return is something that did not happen. The right way to monetize it is by the expected value of the avoided incident: probability times severity, before and after. Compliance investments, security hardening, and reliability work all live here. McKinsey's risk and resilience research is a useful reference for the quantification techniques that hold up under scrutiny.
Option Value Is The Hardest And The Largest
Option value is the return on capabilities the investment unlocks but does not yet exercise. A modern data platform unlocks personalization, ML experimentation, and operational analytics that the business can run later. A modular monolith unlocks future extraction. A clean API unlocks partner integrations the team has not yet sold. The right valuation technique is real-options analysis, and Stanford's business school faculty research is a long-running anchor for the underlying mechanics. Most enterprise ROI calculations zero this bucket out entirely, which is why the calculations consistently under-fund the highest-value investments.
The Operating Habit That Pays Back
The single most useful operating habit is to set measurement targets in all four buckets at investment approval, then review them at fixed intervals. The act of choosing a target forces a clear theory of change. The act of reviewing forces honesty about whether the theory held. Programs that do this consistently outperform programs that don't, by a margin that gets larger every year as the portfolio compounds.
Key Takeaways
- Software ROI lives in four buckets: cost, revenue, risk, and option value
- Cost reduction is easiest to measure and almost always the smallest of the four
- Revenue lift requires causal discipline — experimentation, holdouts, difference-in-differences
- Risk reduction is quietly the highest-leverage bucket; monetize avoided incidents
- Option value is hardest and largest; most ROI calculations zero it out and under-fund the right work
- Set measurement targets in all four buckets at approval; review at fixed intervals