Measuring software development ROI

Q: How do you measure the ROI of software development?

Return on investment is value produced divided by cost. For software, cost is usually known and value usually is not, which is why most attempts stall. Measuring it properly means establishing the value side independently: what each team and vendor actually produces, read from the work itself, set against what it costs, and tracked over time. The return on the AI investment is measured the same way, by the value the AI-built work delivers rather than the volume it generates.

Q: Why is software development ROI so hard to measure?

Because the output has never been legible. Activity metrics measure effort, not value, and AI has broken the link between the two. The result: 49% of CIOs name the assessment of technology ROI as a point of contention with their CFO (KPMG, 2025), and fewer than a third of leaders who see AI productivity gains can connect them to a business outcome (Deloitte, 2025-2026). You cannot divide by a number nobody can see.

Q: Can you measure it without slowing teams down or surveilling them?

Yes. The right unit is the work, not the worker. Reading output and value from the codebases directly, read-only and audit-trailed, needs no self-reporting, no time-tracking and no ranking of individuals. Teams are measured on what the system produces, and nothing is taken from delivery to produce the measure.

Q: How is this different from DORA or engineering productivity metrics?

DORA and engineering metrics measure how work flows, which helps teams run well. They do not measure what the work was worth to the business, and they were not designed to. Software development ROI is a value measure for the budget owner, layered on top, and read independently rather than reported by the teams being assessed.

Q: Is the AI investment actually delivering a return?

For most organisations, not yet in a way they can prove. Only 5% achieve substantial AI ROI, 86% of enterprises expect their AI budget to rise while only 25% see the returns they expected (Master of Code, 2026; NVIDIA, 2026; IBM, 2025). The gap is not that AI cannot pay off, it is that almost no one can currently see whether theirs does. Measuring it is the prerequisite to improving it.

Software is often the largest line in the budget and the only one with no agreed return. The cost is known. The value has never been legible. Here is why, and what a real measure of software development ROI actually requires.

See your own numbers

The return no one can prove

Of enterprises achieve substantial ROI from their AI investment. (Master of Code, 2026): 5%; Of enterprises achieve substantial ROI from their AI investment. · Master of Code, 2026
Of AI initiatives deliver the returns organisations expected, even as budgets keep climbing. (IBM, 2025): 25%; Of AI initiatives deliver the returns organisations expected, even as budgets keep climbing. · IBM, 2025
Of leaders who see AI productivity gains can link them to a business outcome. (Deloitte, 2025-2026): <33%; Of leaders who see AI productivity gains can link them to a business outcome. · Deloitte, 2025-2026

Budgets are rising on faith. Spending goes up, confidence in the return does not, and the reason is the same everywhere: the cost side is clear and the value side has never been measured.

Why it is hard

You cannot divide by a number nobody can see.

ROI is value over cost. For software the cost is on the invoice and the value is a mystery. The measures organisations reach for, story points, tickets, velocity, count effort, and AI has severed the link between effort and value: more is produced and less of it can be accounted for.

So the numerator stays blank. 49% of CIOs name the assessment of technology ROI as a point of contention with their CFO (KPMG, 2025). Any ROI figure built on that foundation is a guess wearing a decimal point.

The fix is not a cleverer formula. It is establishing the value side independently, from the work itself.

What a real measure requires

Four parts, and only one of them is the cost.

Output against cost

What each team, division and vendor produces, on one independent measure, divided by what each of them costs. The numerator most ROI attempts never establish.

AI uptake against delivery

The AI-built share of the work set against what it actually delivers, so the return on the AI spend is measured rather than assumed. Only 5% of enterprises see substantial AI ROI (Master of Code, 2026).

Risk and debt as a liability

Technical debt, with duplicated code blocks up 81% since AI (GitClear, 2026), and quality risk carried on the balance sheet of the code. Ignoring it overstates the return until it comes due.

Trend, not snapshot

The same measure over time, so ROI is a direction you can steer, not a number produced once for a business case and never revisited.

How the value side gets measured

Read the work, ground it in cost, and let no one grade their own return.

Arrio reads what your organisation produces from the codebases themselves, at the level of teams, divisions and vendors, and sets it against what each of them costs. Because the read is independent, the return is not the number a tool vendor or a delivery team wanted it to be. It is measured once, across the whole organisation, and tracked over time. See how measurement works, or the AI productivity paradoxbehind the numbers.

Questions

The questions worth asking

Related reading: the independent software audit, and measuring software developmentin full. Defined plainly:software development ROI.

How do you measure the ROI of software development?

Return on investment is value produced divided by cost. For software, cost is usually known and value usually is not, which is why most attempts stall. Measuring it properly means establishing the value side independently: what each team and vendor actually produces, read from the work itself, set against what it costs, and tracked over time. The return on the AI investment is measured the same way, by the value the AI-built work delivers rather than the volume it generates.

Why is software development ROI so hard to measure?

Because the output has never been legible. Activity metrics measure effort, not value, and AI has broken the link between the two. The result: 49% of CIOs name the assessment of technology ROI as a point of contention with their CFO (KPMG, 2025), and fewer than a third of leaders who see AI productivity gains can connect them to a business outcome (Deloitte, 2025-2026). You cannot divide by a number nobody can see.

Can you measure it without slowing teams down or surveilling them?

Yes. The right unit is the work, not the worker. Reading output and value from the codebases directly, read-only and audit-trailed, needs no self-reporting, no time-tracking and no ranking of individuals. Teams are measured on what the system produces, and nothing is taken from delivery to produce the measure.

How is this different from DORA or engineering productivity metrics?

DORA and engineering metrics measure how work flows, which helps teams run well. They do not measure what the work was worth to the business, and they were not designed to. Software development ROI is a value measure for the budget owner, layered on top, and read independently rather than reported by the teams being assessed.

Is the AI investment actually delivering a return?

For most organisations, not yet in a way they can prove. Only 5% achieve substantial AI ROI, 86% of enterprises expect their AI budget to rise while only 25% see the returns they expected (Master of Code, 2026; NVIDIA, 2026; IBM, 2025). The gap is not that AI cannot pay off, it is that almost no one can currently see whether theirs does. Measuring it is the prerequisite to improving it.

Sources

Master of Code, 2026 Only 5% of enterprises achieve substantial ROI from AI.
IBM, 2025 Only 25% of AI initiatives delivered the expected ROI (2025 CEO Study, 2,000 CEOs).
Deloitte, 2025-2026 79% report productivity gains; fewer than 33% can link them to business outcomes.
KPMG, 2025 49% of CIOs (against 39% of CFOs) name the assessment of technology ROI as a point of contention.
Gartner, 2026 Worldwide IT spending forecast, $6.15 trillion for 2026.
GitClear, 2026 Maintainability Gap study, 211M lines of code: the share of new code rewritten within two weeks rose from 3.3% to 7.1%; duplicated code blocks up 81%; cross-file reuse down 35%; refactoring moves down 70%.

Put a return on the largest line in your technology budget.

Let's talk