Research report
The maintenance dividend
We asked 140 engineering leaders where their budgets go. On average, 61% is spent keeping existing systems running. The best performers spend half that, and reinvest the rest.
About the research
Where the engineering budget actually goes
Every technology leader knows that running existing systems costs money. Few know how much. We asked 140 engineering leaders to split last year's budget between keeping systems running and changing them. The average answer was 61% on running. The best performers spent half that.
We interviewed and surveyed leaders at organizations of 500 to 15,000 people, in eight industries and nine countries, between April and June 2026. Where leaders shared budget documents, 58 of the 140, we checked their answers against them. Self-reported figures were, on average, four points too optimistic.
We call the difference between the average and the best the maintenance dividend: the money and time a team gets back when its systems are cheaper to keep. This report sets out where it comes from and how the top performers spend it.
The headline numbers
Most of the budget is spent standing still
- 61%
- Average share of the engineering budget spent keeping existing systems running.
- 31%
- The same share for the top fifth of organizations: half the average.
- 78%
- The same share for the bottom fifth, leaving little for anything new.
- 2.6x
- How often the top fifth release changes, compared with everyone else.
KIASA, The maintenance dividend 2026. Survey of 140 engineering leaders, April to June 2026.
Why the number matters
A budget that is 61% maintenance is not a failure. Systems need patching, licenses need renewing and people need support. But every point spent on standing still is a point not spent on what the business asked for this year. In our interviews, leaders in the bottom fifth described their roadmaps as wish lists. Leaders in the top fifth described them as commitments.
The gap is large and it compounds. A team that spends 78% on running has 22% for change. A team that spends 31% has 69%. Over three years, the second team can deliver roughly three times as much new work on the same budget, and it tends to spend part of that on making its systems cheaper still.
3xThe new work the top fifth can fund over three years, compared with the bottom fifth, on the same budget.
Age is not the problem
We expected old systems to be the expensive ones. They are, a little. But when we compared systems of similar size, age explained less than a tenth of the difference in running cost. Three other things explained far more: whether the system had a named owner, how much of its testing was automated, and how large its typical release was.
Systems with no clear owner cost 1.9 times as much to run as owned systems of the same size and age. Nobody fixes the root cause of a problem in a system nobody owns. The same incident is handled again and again, by whoever is on call.
“We had a twenty-year-old billing system that was cheap to run and a three-year-old portal that was a constant fire. The difference was that one had an owner who cared.”
The trap of the big rewrite
Many leaders told us they plan to cut maintenance by rewriting their largest system from scratch. Our data suggests caution. Of the leaders whose organizations completed a full rewrite in the last five years, 41% now spend more on maintenance than before. A rewrite resets ownership, test coverage and know-how to zero, and builds them back slowly.
The organizations that cut their maintenance share most did it steadily. They replaced one part at a time behind stable interfaces, retired what they replaced, and paid for the next step with the savings from the last.
Where the dividend goes
The top fifth do not bank their savings. On average they reinvest 72% of what they free up in new product work, and most of the rest in further simplification. Few of them describe this as a strategy. It is a habit: every quarter, some of the time a team gets back is spent on making next quarter cheaper.
Where to start
- Measure the split honestly, by system, using budget data rather than estimates.
- Give every system a named owner with the authority to fix root causes.
- Automate the regression tests on your five most-changed systems first.
- Shrink releases until a bad one can be rolled back in minutes.
- Agree in advance where the freed budget will go, so it is not quietly absorbed.
The top fifth
Four practices of the teams that spend least on standing still
Every system has an owner
A named person who can change it, decide its future and fix the cause of a problem, not just the symptom.
Tests run themselves
Teams with most of their regression testing automated spent 18 points less of their budget on maintenance.
Releases are small
A change small enough to roll back in minutes is cheap to make and cheap to undo. The top fifth release 2.6 times as often.
Retirement is planned
Every new system arrives with a date for retiring what it replaces, and the date is kept.
Key findings
Five findings from the 2026 study
Most budgets are spent standing still
Across 140 organizations, 61% of the engineering budget went on keeping existing systems running. Leaders in insurance and the public sector reported the highest shares; software companies the lowest.
Only one leader in five said the split had improved in the last two years.
Ownership matters more than age
Systems without a named owner cost 1.9 times as much to run as owned systems of similar size and age. Age on its own explained less than a tenth of the difference in running cost.
Automated testing pays for itself first
Teams that automated most of their regression testing spent 18 points less of their budget on maintenance. It was the single strongest predictor in the study.
Most said the investment paid back within a year.
Big rewrites rarely reduce the bill
Of the organizations that completed a full rewrite of a major system in the last five years, 41% now spend more on maintenance than before. Steady, part-by-part replacement did better in every industry we studied.
The dividend is reinvested, not banked
The top fifth reinvested 72% of the budget they freed in new product work. The rest went mostly on further simplification, which is why their advantage grows each year.
Authors
The research team
Aiko Tanaka
Director of Research
Aiko runs KIASA's research programs and designed the survey behind this report.
Osagie Bakare
Partner, Application Modernization
Samuel leads our work replacing core systems one part at a time.
Daniel Okafor
Chief Delivery Officer
Daniel is responsible for how KIASA delivers and hands over every engagement.
Related
Keep reading
Perspective
Why the second year of a cloud migration matters more than the first
The move itself is the easy part. Costs, habits and ownership settle in year two, and that is where most of the promised value is won or lost.
Research report
Platform pulse: what engineering leaders will fund in 2027
Budgets are flat; expectations are not. Our annual survey of 320 technology leaders shows where the money moves next year.
Research report
Security debt: the quiet risk on the balance sheet
Unpatched systems behave like unpaid loans: the interest compounds. A way to measure the exposure, and a plan for paying it down.
Case study
How Halden Mutual settles claims in two days, not nine
A 90-year-old insurer replaced its claims core without a freeze, one product line at a time.
Find your own maintenance dividend
In two weeks we measure where your engineering budget goes, system by system, and show you which changes would free the most of it.