You are the architect in an annual portfolio review. Eight initiatives are funded for the year out of forty proposed; none of them is yours. Five of the eight need changes to the same entitlement component, which is owned by a three-person team that also absorbs unplanned support. What do you bring to that meeting that the product directors cannot, and how do you get enabling work funded when it has no business case of its own?

What the interviewer is testing

Whether you understand that a portfolio is a set of claims on shared capacity, and that nobody in the room can see those claims except you. Product directors each hold a correct view of one initiative. The overlap is invisible from every seat but yours, and an architect who turns up with a target-state diagram has brought the wrong artefact.

The clarifying questions that change the answer

• Is the capacity figure gross or net of run work? A team losing 30% to unplanned support has 70% of what the plan assumes, and that single correction often removes two initiatives.
• Which of the eight touch the same component, team or datastore?
• What is each initiative's cost of delay — is the money deferred by a quarter or destroyed because a competitor or a regulatory date arrives?
• Who is the single approver for each, and do any two share one?

A strong answer's arc

Bring one page: the eight initiatives against the components and teams they need, with the five-way collision on entitlement marked. Then say what the collision costs in the plan's own terms — if all five start now, the entitlement team context-switches across five consumers and finishes roughly three of eight by year end; sequenced so entitlement ships first, six of eight land. That is a sequencing claim with a number, and it is the only kind of claim that survives the room.

Then solve the funding problem by not asking for a separate budget. Enabling work with no customer in front of it loses every individual comparison against a feature, which is why it should not be compared individually. Attach it as the first slice of the largest dependent initiative: that sponsor now owns it, it has a date, and the four other initiatives become cheaper in their own cases. Where that is not possible, ask for a protected share of capacity — commonly 15% to 20% — so the aggregate decision is taken once rather than lost forty times.

Common weak answers

• "We need a technical debt budget." It creates a line item that is cut first in any squeeze and signals that the work has no business owner.
• Ranking all forty by weighted shortest job first. Useful arithmetic, wrong assumption: the formula treats items as independent and the whole problem here is dependency and shared capacity.
• Building the enabler quietly inside another initiative. It happens, it works once, and it ends the architect's ability to make this argument in public.
• Escalating to the CTO. Wins the item and loses the mechanism.

What a strong answer adds

Name the exit condition. Write the expected outcome before the work starts — "entitlement changes for a new product go from six weeks to one week" — and book a benefit checkpoint with the sponsor at month six, so next year's version of this meeting has evidence rather than a story. Say also what you will stop: a portfolio argument with nothing removed from it is a request, not a plan.

The senior move is to name the failure in advance. Choose one of the eight to deprioritise explicitly, and say which signal in production — entitlement change lead time, measured monthly — would show the sequencing claim was wrong by month four, so the plan fails loudly and early rather than quietly in Q4.