The number that has killed more transformation programs than any failed project is a single fraction, and a CFO can run it in his head: technology spend over revenue. Ours against theirs. Picture the slide: we spend a shade more than the sector median, so by the only metric on it we are running IT expensive. That one comparison can end a funding conversation before it starts. Nobody demonstrated the spend was wasted. A ratio said it was above average, and above average reads as something to cut.
I do not sign the whole IT P&L. I run security and DevOps, a real slice of it, and I report the return on that slice to a board, so take this as analysis from close range rather than a CIO's memoir. You do not have to own the whole number to watch how it gets used, and "IT spend as a percent of revenue" is the most quietly destructive benchmark in the mid-market. It rewards the companies that underinvest, it punishes the ones that are growing, and it blends a bet on the future into the same line as the electricity bill. It does its worst damage at the exact moment a business is most fragile.
The denominator is doing the arguing
The ratio has two moving parts and everyone treats it as a statement about one of them. We read it as a verdict on the numerator, are we spending too much, and forget that it moves just as hard on the denominator, which IT does not control at all. Revenue is set by the market, the sales motion, the macro. Put a number IT owns over a number it doesn't, and half the movement in the "IT metric" is not about IT.
Watch what that does to two companies with identical discipline. One is investing to grow: it stands up the platform, hires the team, and buys the capacity before the revenue that capacity enables has arrived. For a year or two the numerator is up and the denominator has not caught up, so the ratio looks bad, precisely because the company is investing in growth. The other is harvesting a mature line: it spends nothing new, revenue holds, and the ratio looks lean and efficient. The benchmark calls the first company profligate and the second disciplined, and it has the causation exactly backwards. The lean-looking one may be the company quietly eating its seed corn.
So the ratio is not measuring discipline. It is measuring where you sit in an investment cycle, and it flatters whoever invested least recently. A metric that rewards standing still is a strange thing to hand a board and call a benchmark.
The metric spikes exactly when you can least afford the cut
Now run the downturn. Revenue falls, so the denominator shrinks. IT's cost base does not shrink with it, because most of it is fixed: the contracts you signed, the licenses you renewed, the platform you run, the people who keep it up. The ratio jumps, mechanically, without a single new dollar of spend. Nothing about the technology got more expensive. The business got smaller, and the fraction did what fractions do.
Then the CFO opens the benchmark, often last year's peer median because the fresh one has not published, and sees the company sitting well above peers. The story writes itself: we are overspending on IT, trim it. And look at where the trim has to land. Keep-the-lights-on is load-bearing; cutting it breaks a promise already made to a customer or an examiner. So the cut flows to the only discretionary line, which is the transformation spend, the bets on a different future. In a downturn the percent-of-revenue metric functions as a machine for converting a revenue problem into a decision to stop investing in the recovery. That is countercyclical in the worst direction. You disinvest in digital capability at the precise moment competitors who kept investing pull ahead, and at the precise moment the resilience that spend was buying matters most.
In a regulated business the trap is tighter, because a chunk of the fixed base is controls you are examined against, and you cannot cut those without inviting a finding. So the ratio spikes higher, and the "excess" the CFO sees is even more concentrated in the one place you least want to touch. The metric points the axe at the future and calls it prudence.
The median is not your strategy
The number only becomes a weapon when it is compared to something, and the something is almost always a sector benchmark. Gartner's IT Key Metrics Data and similar syndicated studies publish IT spend as a percent of revenue by industry, and they are genuinely useful for one narrow purpose I will get to. The trap is treating the median as a target rather than a description.
Comparing your ratio to the sector median assumes you share the median company's strategy. But the median is, by definition, a company doing what everyone else is doing: defending the present, not building a different future. Benchmarking your transformation spend against a peer median is benchmarking your ambition against the average of everyone else's lack of it. If your plan is to change your position in the market, the median is the number you are trying to escape, not the one you are trying to match.
The comparison also hides the thing that actually matters. Two companies at the identical ratio can be in opposite condition: one spending it on a modernization that will compound for a decade, the other spending it servicing debt and keeping fragile systems breathing. The ratio cannot tell those two apart, and neither can a CFO who has only the ratio. Same number, opposite futures, and the benchmark scores them the same.
A blended number steers every cut toward the future
Underneath all of this is the deepest failure: one blended ratio cannot tell a bill from a bet. Gartner's run/grow/transform split, and I have argued elsewhere for reporting the three as separate lines, names the distinction the ratio erases. Run is keep-the-lights-on, the spend that breaks a promise if it stops. Grow scales what already works. Transform is the bet with no run-rate yet, the only line buying a different future rather than defending the present one.
Fold all three into a single percent-of-revenue number and every dollar looks the same, which means the cheapest dollar to cut is always the one that is not load-bearing today. Run you cannot touch without breaking something. Grow tracks revenue that, in the scenario that triggered the cut, is already falling. Transform is the line with no immediate consequence to cutting, so a metric that treats all IT dollars as fungible will always point at it first. The benchmark is not neutral about what gets cut. It has a structural bias toward eating the future, because the future is the only thing on the ledger you can stop paying for without anyone noticing this quarter.
When my team built AgentOS, our governed internal agent platform, it was a transform bet by definition: new capability and no revenue line of its own to point at. Under a percent-of-revenue lens a project like that is pure added cost. The numerator goes up and nothing offsets it in the denominator this year, because the return shows up as capability rather than a booked dollar you can name in the same quarter. A benchmark that only sees the ratio flags that investment as bloat. A board that understands it is a bet asks the one useful question, is the thesis real and is it paying off, instead of whether the ratio ticked up a tenth of a point.
What to put in front of the board instead
You do not defeat a bad benchmark by refusing to be measured. You defeat it by bringing better numbers into the room before someone brings the bad one. Four moves, none of which require a platform or a reorg.
- Split run from grow from transform, and benchmark only run. The percent-of-revenue comparison is legitimate for exactly one layer: commoditized keep-the-lights-on, where you genuinely should be near or below peers because that spend is a utility and cheaper is better. Benchmark run against the sector. Refuse to benchmark grow and transform against it, because those are strategy, and strategy is not a peer average.
- Report the slope, not the level. The useful question is never "what percent of revenue," it is "which way is the cost to serve a unit of the business moving." A run line falling as a share of revenue over time is discipline you can show. A transform line you are deliberately growing is a decision you are defending. The trend carries an argument the snapshot never can.
- Price the restraint, not just the spend. Every proposed cut has a cost that never appears in the ratio: the modernization that slips a year, the competitor who ships the thing first, the examiner finding you now remediate under a deadline instead of on your own schedule. Put a number, even a hedged range, on the cost of not spending. A CFO can weigh a return against a cost. A CFO cannot weigh a return against a blank.
- Give every transform dollar a named outcome and a thesis. Not "IT spend," but "this buys the capability to onboard the next thousand institutions without adding proportional headcount," or "this reduces the expected loss from a control gap the examiner has already flagged." A dollar with a thesis survives a downturn. A dollar dissolved into a blended ratio does not.
Report the strategy, not the ratio
Bring the board the split before someone brings them the fraction. Benchmark the utility layer against peers and defend the strategy layer against the plan, never the median. Price the restraint as carefully as you price the spend; the cost of the cut belongs on the same slide as the cost of the investment. And when the downturn comes and the ratio spikes on its own, be the person in the room who can explain that the number moved because revenue fell, not because IT got expensive.
I am reasoning toward the full version of this from a slice of the P&L, not the whole of it. The stories worth studying are the downturn ones: the benchmark that got someone cut, or the better number that defused it in the room.