New value versus keeping the lights on
Most orgs can't say what fraction of their capacity built something new last quarter versus kept the lights on. Look at the split occasionally. The moment it becomes a target, it starts lying to you like every other metric does.
Ask a team lead what fraction of last quarter's capacity went to something genuinely new, against keeping the existing thing running.
Watch what happens.
Most either answer with a number that has no evidence behind it, or admit they've never looked.
Both more common than a considered estimate. Even at companies that otherwise track their engineering numbers carefully.
Headcount and roadmap decisions run on an assumption about this split, whether anybody states it or not.
That's why the gap matters.
A plan assuming most capacity is available for new work, when the real number is closer to half, isn't an aggressive plan. It's a wrong one, wrong in the same direction every quarter, until somebody actually measures the split rather than continuing to guess at it.
A rough cut, not a taxonomy
No formal system needed. Building one is usually a way of avoiding the simpler, more useful version.
Three buckets, applied loosely, by the team itself rather than imposed from outside.
New value. Work not required by anything already shipped.
Keeping the lights on. Maintenance, upgrades, the unglamorous work of keeping existing commitments alive.
Recovery. Fixing something that shouldn't have broken.
That third one is worth separating from ordinary maintenance, because if it's large it's the bucket most likely to be hiding a real problem.
Run it for a month.
Whatever the team already tracks, categorised loosely.
The exact boundaries between buckets matter far less than the order of magnitude does.
Fifteen percent on new work, when leadership has been planning around fifty? That team just learned something worth acting on, even if the real number is off by five points either way.
Where it goes wrong immediately
Measuring an engineering org without lying to yourself lays out what happens to any number the moment it becomes a target.
It stops measuring the thing it was built to measure.
It starts measuring how well people can move a number.
This one is especially exposed, because the categories are judgment calls rather than hard counts, and judgment calls are easy to game the instant somebody's incentive depends on the split looking a particular way.
So the rule here is stricter than for most things on a dashboard.
Look at this number.
Don't manage it.
No target attached to the new-value percentage, ever. No comp or headcount decision rewarding a team for reporting a higher one. The moment either happens, every future number from that team is worthless, because the incentive to categorise generously arrives instantly and never leaves.
What it's actually for
Used as a diagnostic, this sits next to throughput beats utilisation. A different axis on the same underlying question.
Not how full the calendar looks.
What category the work in it actually falls into.
A team that's fully booked and shipping constantly can still be spending nearly all that time keeping the lights on, and a calendar full of activity will never surface it. Only looking at the split will.
The one useful conversation this starts
Run it once, honestly.
The conversation that follows is rarely about the number. It's about whether the org's plans have been quietly assuming a split that was never true, and how many of this year's commitments got built on top of that assumption without anybody ever naming it out loud.