An architect has more work than capacity and every item has an advocate. What method produces a defensible allocation?

The method

• Sort by cost of delay rather than by value. Two items of equal value differ enormously if one has a deadline and the other does not. The item whose value decays fastest goes first, which is a different ordering from value alone and a more defensible one.
• Account for dependencies and sequencing, since a low-value item may unlock three high-value ones and a purely ranked list misses that.
• Express capacity in teams, not people, because the constraint is usually one team's bandwidth and moving people does not move capacity in the short term.
• Include the run cost of everything shipped, since each addition brings permanent maintenance, on-call surface and cognitive load. A plan allocating all capacity to new work is planning for a slowdown it will then be surprised by.
• Include stop decisions. A process that only adds is not prioritisation.

The category that must be protected rather than ranked

Work that removes future cost — paying down a constraint, replacing a throughput-limiting system, automating a manual process. Its value is real, diffuse and unattributed to a customer, so it loses every individual comparison against a feature with a sponsor, reasonably, and the aggregate outcome is nobody's decision.

A fixed protected allocation, owned by the team, is the only intervention that changes this, because the item-level comparison will always be lost.

The seasonal reality in commerce

Capacity is not uniform across the year. A large share of the annual outcome depends on a few sale events, and the weeks before them are effectively frozen. A plan that ignores the freeze windows over-commits the calendar and produces a predictable end-of-quarter shortfall.

The conversation that makes it defensible

Show what is not being done and why, rather than only what is. An allocation presented as a list of commitments invites the question "why not also X"; one presented with the explicit trade-off — "X displaces Y, which we chose because of Z" — makes the reasoning inspectable.

The disagreement then moves to the estimates and the weights, which is a productive place for it and one where evidence can be gathered.