Run, grow, transform is a budgeting convention, not a cost model
In brief
Run, grow, transform is a useful board shorthand and a weak cost model. Where the split breaks down, what the federal DME and O&M split adds, and what to do.
Steve Elliott · · 7 min read
On this page
Run, grow, transform is a labeling convention for technology spend, not a measurement. It works as a board-level shorthand and as a way to fence off money for new work, but the labels rest on judgments that drift, and they say nothing about what the "change" spend will cost to run once it ships.
In our view the fix is not to drop the split. It is to tie each label to a portfolio decision, track how change spend turns into future run cost, and report the split on top of a TBM cost model rather than in place of one.
Where does run, grow, transform come from?
The run, grow, transform split is widely attributed to Gartner. Gartner's research pages were not accessible to us, so we rely on a secondary description here. Info-Tech Research Group, in a 2018 note, says the model "originates with Gartner's approach to IT budgeting." It describes the three categories this way:
- Run is the largely nondiscretionary work that makes up most of IT's day: systems support, maintenance and firefighting.
- Grow is more discretionary: investments in systems that serve business goals.
- Transform covers investments that break new ground, such as taking the business to a new level of revenue, market share or expansion.
The same note reports that Gartner identified an ideal mix of 50:25:25 across run, grow and transform, against an average of 70:19:11. We could not verify those figures against Gartner directly, and they are several years old. Grow and transform are often merged into one bucket, usually called grow and sometimes change, giving a two-way run/grow split, as a 2018 Computerwoche piece notes. That piece puts the average at about 70% run and 30% grow, without citing a study.
What is the split good for?
It earns its place in two ways.
It is a shorthand a board can hold. Three numbers fit on one slide. A board can ask whether run spend is growing faster than revenue, or why transform fell from one year to the next, without learning a cost taxonomy. The TBM Council's guide for CFOs has a section on run-versus-grow trade-offs, which it describes as choosing between legacy maintenance and new innovation.
It protects new work from the run budget. Run spend is hard to cut quickly. Contracts renew, systems need support and incidents don't wait. We think that, without a fenced category, money for new work tends to be what's left after run is funded. Naming a grow or transform target gives a CFO and CIO something to defend in the budget cycle.
Both uses are real. Neither requires the labels to be precise, which is where the trouble starts.
Where does the split break down?
These are our observations, not survey findings.
Classification is a judgment, and judgments drift. The categories describe intent, not a property of the cost. The same upgrade can be filed as run (it keeps a system supported) or grow (it adds features). Whoever classifies has a reason to lean one way, since grow and transform often get easier approval. Info-Tech's note makes a related point: it argues that most portfolios, especially in industries like construction, are driven mainly by requests IT takes on because someone wants them.
Grow and transform overlap. A new customer app might grow an existing line of business or transform how it sells. None of the sources above gives a test for the line, so the grow-transform split often says more about how a project was pitched than about what it does.
Delivered change becomes run. A transform project ends; the system it built does not. Licenses, cloud consumption, support staff and security work move into run, often in the next budget year. If run is reported as a share of the total, a successful transform program shows up later as a rising run share, which can read as failure on a board slide.
Persistent product teams blur the line. A team funded for a product year after year fixes defects, keeps services running and ships new features in the same sprint. Splitting its cost into run and change means estimating that mix, for example from time records or a fixed ratio. The TBM Council notes that persistent agile teams disrupt traditional capex and opex accounting. In our view they disrupt run versus change for the same reason: the funding unit is the team, not the project.
Here is how the third problem plays out. This example is illustrative and assumes all other run spend stays flat at $70M.
| Year | Transform project spend | Annual run cost of what it delivered | Total budget | Run share |
|---|---|---|---|---|
| 1 | $6.0M build | $0 | $100M ($70M run) | 70% |
| 2 | $2.0M final release | $1.5M licenses, cloud and support | $100M ($71.5M run) | 71.5% |
| 3 | $0 | $1.8M as usage grows | $100M ($71.8M run) | 71.8% |
Nothing went wrong in this example. The project delivered, and the organization now runs what it built. But the run share rose almost two points on a flat budget, and the grow and transform envelope shrank by the same amount. Unless the cost to run was visible when the project was approved, the board sees a trend it did not choose.
How does the federal DME and O&M split handle the same problem?
US federal IT budgeting uses a two-way split that is, in our view, the closest public analogue. OMB's FY 2021 IT budget guidance defined development, modernization and enhancement (DME) as projects and activities leading to new IT assets, or changing existing ones to substantively improve capability or performance, implement legislative or regulatory requirements, or meet an agency leadership request. Operations and maintenance (O&M), also called steady state, covers the expenses required to operate and maintain an IT asset in production. That includes maintenance projects needed to sustain current capability and performance, and the same guidance lists technology refreshes among maintenance activities.
Two features of the federal approach are worth borrowing.
- The main test is capability, not whether the work is a project. A large refresh with a start and end date is still O&M if it adds no capability. OMB's FY22 guidance asked agencies to report such maintenance projects within O&M. We think that is a sharper rule than "is it a project?", which is how run and change are often split.
- The split was reconciled to a cost model. OMB's FY22 guidance said investment spending across IT cost pools and across IT towers would generally match investment totals by DME and O&M. The two views were expected to add up.
Federal figures show both the level and the drift. GAO reported in July 2025 that the federal government spends more than $100 billion a year on IT and cyber-related investments, and that agencies have typically reported spending about 80% of it on operations and maintenance of existing IT. On the trend, OMB's Capital Programming Guide notes that steady-state IT investment has risen each year since 2003, while DME has trended downward.
One caveat: the 2026 edition of OMB Circular A-11, section 55, removed what it called outdated references to TBM and points agencies to a data collection site for detailed reporting instructions. The definitions above come from earlier published guidance. For more on the federal TBM effort, see why IT cost transparency projects stall.
How do you make run, grow, transform useful?
The three views answer different questions. In our view you need all three, joined.
| View | Question it answers | What it's built from | Main weakness |
|---|---|---|---|
| Run, grow, transform | How much goes to keeping the lights on versus new work? | Classification of projects and budget lines | Rests on judgment; ignores future run cost |
| DME and O&M (federal) | Does the spend add capability or sustain it? | Investment-level reporting against defined tests | Still needs an estimate for mixed teams |
| TBM cost pools and towers | What does technology cost, by type and by function? | General ledger mapped to a standard taxonomy | Doesn't say what spend is for |
Here is what we would do.
- Tie each label to a decision. A grow or transform label should point to a portfolio decision: who approved it, the outcome it is for and when it will be reviewed. If no one can name the decision, the label is a description, not a commitment.
- Write down the classification rule. Borrow the federal capability test: does the work add capability or sustain it? Apply it the same way every year, and log reclassifications rather than restating history quietly.
- Put the cost to run on the approval. Every change investment should carry an estimate of the run cost it will create, by year, in the same cost pools the TBM model uses. Then track actuals against it after go-live. This is a total cost of ownership discipline applied at approval.
- Report run growth from delivered change separately. Split year-on-year run growth into what came from newly delivered systems and what came from everything else: price rises, volume, technical debt. The first is a planned consequence of investment; the second is the number to challenge.
- Classify product teams by a stated ratio, and revisit it. For persistent teams, set a run-change ratio from time records or backlog data, publish it and review it each quarter. Precision matters less than consistency.
- Keep the split as a view on the cost model. Tag projects, solutions or budget lines with run, grow or transform, and roll the tags up from the same data as cost pools and towers. The totals should reconcile, as federal DME and O&M totals were meant to.
The point of the split is the trade-off it puts in front of the board, and that trade-off is only honest if today's change spend is connected to the run cost it creates. That connection, between a funding decision, the work and its lasting cost, is what an operating model should hold.
Sources
- Info-Tech Research Group, Travis Duncan, Run-Grow…Appease: Whither the Ideal Portfolio Mix?, July 10, 2018 (secondary description of Gartner's model)
- Computerwoche, Bernd Worlitzer, Die IT-Budgetierung folgt neuen Gesetzen, March 19, 2018
- TBM Council, Jeff Bartelli, The CFO's Guide to the TBM Taxonomy, updated May 8, 2026
- OMB, FY 2021 IT Budget – Capital Planning Guidance, Appendix C (Definitions), 2019
- OMB, FY22 IT Budget – Capital Planning Guidance, 2020
- OMB, Circular A-11 (2026), Section 55: Information Technology Investments, August 2025
- OMB, Capital Programming Guide, supplement to Circular A-11 (2026), August 2025
- GAO, Information Technology: Agencies Need to Plan for Modernizing Critical Decades-Old Legacy Systems (GAO-25-107795), July 17, 2025
Frequently asked questions
Is run, grow, transform a Gartner model?
It is widely attributed to Gartner. Info-Tech Research Group described it in 2018 as originating with Gartner's approach to IT budgeting. The split groups technology spend into run (keeping existing operations going), grow (discretionary investment in business goals) and transform (investment in breaking new ground). Gartner's own research is subscription-only, so check definitions with Gartner directly.
What is the ideal run, grow, transform ratio?
There isn't one that holds for every organization. Info-Tech reported in 2018 that Gartner identified an ideal 50:25:25 mix against an average of 70:19:11, and Info-Tech's note argued no archetypal mix exists. In our view the trend matters more than the target, especially how much run growth comes from newly delivered systems.
What is the difference between DME and O&M?
They are the federal government's two IT spending categories. In OMB's FY 2021 guidance, DME covers projects that create new IT assets, substantively improve existing ones, or meet legal or leadership requirements. O&M, also called steady state, covers the cost of operating and maintaining assets in production, including maintenance projects that sustain current capability. A refresh that adds no capability counts as O&M.
Does the TBM Taxonomy include run, grow and transform?
Not as a layer. The TBM Taxonomy classifies cost by cost pools, resource towers, solutions and consumers. The TBM Council's CFO guide does discuss run-versus-grow trade-offs. In our view run, grow and transform works best as a tag applied to the same data, so the split reconciles to cost pools and towers rather than replacing them.
Steve Elliott
Published November 22, 2026
More on TBM & IT Financial Management
Cost pools, towers, showback, unit cost.
Keep reading
Eliminate Data Entry, Not Strategic Context
New in Dotwork: agents that collect status in Slack so teams stop re-entering updates, an orchestrator agent for leaders, and Agent Studio for building governed enterprise agents.
On the TBM Council's numbers, tokens are 4.5–15% of AI cost. What that means for your budget
The TBM Council puts labor at 50–70% of AI project cost. On its numbers, tokens are 4.5–15% of the total. What the figures show and what to budget for.
Annual budget or rolling forecast for IT spend? Plan the fixed lines, forecast the volatile ones
When a fixed annual IT budget works, when a rolling forecast is better, and how to run both: plan headcount and contracts, forecast cloud and AI by driver.
ASU 2025-06 and your IT cost model: what changes before you adopt
ASU 2025-06 drops the stage model for internal-use software. What it changes for IT finance: when capitalization starts, the evidence, and the cost model.