Where should the next unit go?

Every effort pays off less the more you pour in โ€” the tenth week on a thing rarely buys what the first did. So splitting a fixed budget for the most total value isn't an even share, and it isn't all-in on the best bet. It's water-fillingPush each funded effort until the value of its next unit is the same everywhere. Anything still worth more than that common level deserves more; anything below it deserves less.: feed each effort until the next unit everywhere is worth the same.

Splitunits of effort across these 4 efforts.
higher is better
StrategyTotal value
Best splitequalise the next unit's valueoptimum
300
Split evenly7.5 units each
265
All-in on the besteverything into Core migration
185
Value of the next unit of effort
02040budget 3051
35left on the table by splitting evenly

The even split captures 265 โ€” 88% of the 300 the best split gets, for the same 30 units. It overfeeds the efforts that have already flattened and starves the one still climbing. The fix is to level the marginal return โ€” the curve's height at your budget, 5.7 per unit.

Bug bash
2.8 u+4785% of upside
Core migration
17.0 u+12439% of upside
Onboarding
5.9 u+7174% of upside
Sales enablement
4.4 u+5978% of upside

Notice the split doesn't chase the biggest upside โ€” it stops feeding an effort the moment its next unit drops below what another effort's next unit is still worth. An effort that saturates fast gets a small, early dose; a slow burner with room to climb keeps earning more.

a diminishing-returns (water-filling) allocation โ€” value is relative, only meaningful next to itself