Investment cases / February 2026
Rebuilding the IT budget from zero
Set every technology line to nothing and reinstate it only where a named business outcome for the coming year requires it. The purpose is a defensible budget, not a smaller one. What finance wants from an executive team is the evidence under each line and a way to keep producing it after the review closes.
The Problem
The incremental budget now fails a review it used to pass easily.
Most technology budgets are still assembled by taking the prior year and adjusting for growth. A function that spent a given amount last year argues for the increase and is almost never asked to defend the base. That method depends on three assumptions. The first is that the previous spend was earned. The second is that the coming year is a scaled copy of the last one. The third is that the marginal dollar buys roughly what the first dollar bought.
None of the three survives a board that has watched interest expense become a managed line item and has approved a wave of technology pilots on expectation instead of measured return. Those pilots are the clearest case. A pilot funded mid-year arrives at its first real budget review with no prior base to grow from, which is exactly the input an incremental method cannot process.
Spending is not shrinking, which is part of why the scrutiny feels new. Gartner's worldwide IT spending forecast published in April 2024 projected about $5.1 trillion for that year, an increase of about 8 percent over 2023. Growth at the top of the budget and hard questioning underneath it are happening in the same review, and most finance functions have no method that answers both at once.
The Method
A rebuild is not an across-the-board cut.
The reflex answer to a cost problem is a flat percentage applied to every function. It is quick, it looks even-handed, and it produces the opposite of fairness. A flat cut takes the same proportion from the team that already trimmed its request as from the team that padded its request in anticipation of this exact exercise. It rewards the padding and penalizes the discipline.
A rebuild asks a different question. It never asks what moved since the last cycle. It asks what this spend produces over the coming twelve months. A function that genuinely needs its full prior budget should recover all of it, and asking should carry no penalty. A function that inflated its number has nowhere to put the inflation, because there is no baseline left to hide inside.
Zero-based budgeting is the established finance discipline for this, and the two-test sort in the next section is what we add on top of it.
The Sort
Sort every line before you defend it, on two tests.
Rebuilding an eight or nine figure budget in a single sitting, one row at a time, is not realistic. Sorting it is. The first test is whether the spend is essential. Essential means it protects a revenue-producing system, a regulatory obligation, or a security control an auditor will name. Popularity inside the organization is not the test.
The second test is whether a cut can be undone. A reversible line can be switched off now and turned back on in three months with nothing broken behind it. A locked-in line cannot, because a contract penalty applies, because rehiring takes months, or because a data move cannot be run backwards once it begins.
Two tests produce four treatments, and the treatments are the point.
- Essential and reversible. Fund at the current run rate and re-examine the number quarterly instead of fixing it for twelve months. Elastic consumption and contracted capacity behind a revenue system usually belong here.
- Essential and locked in. Protect first and review last. Reinstate with the lightest scrutiny in the exercise, and require executive sign-off for any exception instead of budget-owner discretion.
- Optional and reversible. Zero it now. The owner earns it back by naming the outcome it produces this year. The fastest uncontested savings are here, and none of them need a vendor conversation.
- Optional and locked in. Do not cancel it. A multi-year agreement running at half its committed volume needs a renegotiation, a rescope, or a controlled run to the break clause.
The common failure is applying the third treatment to all four boxes. Zeroing a security control because it looked like every other row on a spreadsheet is how a cost reset becomes a control gap two quarters later.
What Finance Checks Next
The flagged number and the recovered number are not the same number.
A first pass flags a gross figure. Finance will ask what portion of it converts to cash this year, and that is where most exercises overstate themselves. Locked-in agreements usually come back with the committed minimum untouched and a partial adjustment at renewal. Present the recovered figure as the commitment and the gross figure as the pipeline. A committee that budgets the gross number misses its own target by exactly the difference.
Ownership decides the quality of the sort. A rebuild run inside finance alone flags the wrong lines. A spreadsheet shows how large a row is. It says nothing about whether cutting it can be undone. A rebuild run inside the technology organization alone protects its own base. Put one seat from finance beside one from architecture or engineering and hold both to a single scorecard. Give them spend grouped by business unit and workload instead of by account structure, or the sort will score vendor invoices in place of business outcomes.
A professional services firm paid for a stack of overlapping software subscriptions and several suppliers doing similar work, with each year's budget set by adjusting the prior year. We rebuilt the spend line by line with one finance owner and one architecture owner, separated the optional and reversible spend from the protected core, and consolidated the duplicated agreements over two renewal cycles. Technology operating cost fell 21 percent and every remaining line had a named purpose.
Keeping It
A rebuild done once a year is an incremental budget with extra steps.
The most common way this discipline dies is quiet. A budget rebuilt in one month and then trended normally from the next month onward has deferred the problem by eleven months. Every commitment made in the spring becomes an untouchable base by the following winter, which restarts the habit the rebuild was meant to end.
Re-run the sort quarterly, not the whole rebuild. A line that was optional in February can become essential by May once the business builds a dependency on it. A line that looked locked in can become cancellable the week a break clause opens. The sort is cheap to repeat. The rebuild is not, and it does not need to be.
What To Do Next
Six moves, in this order.
- 1. Restate every technology line against a named business outcome for the coming year, not against what it cost last year.
- 2. Apply the two tests to each line. Essential or optional, then reversible or locked in.
- 3. Zero the optional and reversible box in the same week. It needs no negotiation.
- 4. Route every optional and locked-in agreement to a renegotiation queue and budget the recovered figure, not the flagged one.
- 5. Reinstate the essential and locked-in box first, with executive sign-off required for any exception.
- 6. Put the sort on a quarterly cadence with two named owners, one from finance and one from technology.
The executive teams that get the most from a reset do not finish with the smallest number. They finish able to explain, row by row, what the number produces.
Related Service
IT Strategy and Investment Cases
We build the investment case underneath the budget and stay in the room when finance tests it.
Start A Conversation