Choose what will not fit this week

A founding engineer interview may ask how you handle competing urgent requests. The useful answer makes the constraints visible and chooses a sequence. Saying everything is important does not allocate time. Saying revenue always wins can ignore a security defect or a failure that prevents existing customers from receiving value.

Identify the decision owner and the available capacity. Separate a confirmed obligation from a prospect's request, an active incident from a speculative risk, and a learning task from implementation. Estimate enough to compare options, then state uncertainty. A two-hour investigation can be the correct first step when the cost or cause is unknown.

Use impact, reversibility, and dependency rather than a decorative score that hides judgment. A defect exposing private data needs containment before a cosmetic improvement. A customer-critical export failure can block both retention and a planned demo. A requested integration may be valuable, but its urgency depends on a specific commitment and whether a smaller workaround can test the need.

Explain what deferral means. Name the user effect, temporary workaround, next review point, and evidence that would move the item earlier. Deferred does not mean forgotten. In a small team, an explicit decision record helps prevent the same argument from restarting whenever someone mentions the request.

A fictional team has four engineer-days. The backlog contains a confirmed cross-account access defect requiring one day to contain and verify, an export failure affecting three paying pilots requiring one day to diagnose and fix if the current hypothesis holds, a prospect integration estimated at five days, and a demo improvement estimated at one day.

The chosen plan contains the access defect first, export recovery second, then the demo improvement if capacity remains. Reserve the final day for uncertain recovery work. The integration gets a short discovery conversation and a manual sample if safe, not a five-day commitment hidden inside a four-day week. The decision records the prospect's needed outcome and a review date.

A priority answer should show which facts are firm and which estimates can move. Treat the one-day export repair as conditional on its current diagnosis. If the hypothesis fails, the plan needs a checkpoint before consuming the reserve or starting the demo work. Do not describe uncertain work as a guaranteed delivery date.

Four days can cover access containment, conditional export work, the demo, and reserve only if estimates hold. The five-day integration cannot fit inside that plan. The discovery conversation also consumes time; fit it inside an explicitly allocated small part of the reserve or defer it rather than adding invisible work.

A concise recommendation could be:

1Decision: contain access defect first.2Next: one-day export diagnosis/repair timebox.3Checkpoint: after the diagnosis, confirm repair estimate and user state.4Then: demo only if recovery is verified and reserve remains.5Deferred: full integration; offer a bounded discovery conversation.6Revisit: confirmed contractual date, changed harm, or failed repair hypothesis.

This is an original interview practice record, not a claim about a specific employer's prioritization rubric. Its value is that another person can change an input and see the consequence.

Distinguish containment from a complete fix. Disabling the affected sharing route might stop exposure within hours, while the permanent authorization repair takes longer. State what the temporary control prevents and what useful behavior it suspends. A plan can become feasible by containing harm first, but not by silently calling a disabled feature fully repaired.

The integration does not only cost five engineering days. It displaces current-user recovery, the demo, investigation, or reserve. Name the displaced outcome. A prospect's possible revenue can matter, but do not treat an informal statement as a signed commitment with a guaranteed value. State the evidence and ask what narrower proof could reduce uncertainty.

A manual sample may test the integration's desired output without building a production connection. It still needs appropriate data permission, a bounded effort estimate, and an honest explanation that it is manual. If the prospect needs automated daily synchronization to evaluate the product, a one-off sample tests only part of the need. The smaller test must match the uncertainty rather than simply be easier to build.

Suppose the export investigation discovers a data repair requiring three more days. The original four-day plan no longer fits. Keep the access containment, report the changed scope and current customer workaround, and renegotiate the demo or integration timing. Do not consume hidden evenings to preserve a false plan without exposing the consequence.

A correct answer need not choose one universal order for every business. It should preserve active-harm containment and explain the tradeoff to the decision owner. Once harm is contained, a confirmed deadline or critical customer commitment can change the next sequence. The evidence and capacity table make that change understandable.

One misconception is that a numerical priority score removes judgment. A score can conceal uncertain revenue estimates or erase a nonnegotiable access boundary. Another is that adding a reserve means the team is undercommitted. Uncertain diagnosis and verification consume real capacity; omitting them makes the plan less accurate.

Exercise: access containment finishes in half a day, but permanent repair needs two more days. Export recovery remains one day, and only three and a half days remain. Propose a feasible plan that separates temporary and permanent outcomes. One answer completes the two-day access repair and one-day export work, leaves half a day for verification or investigation, and defers the demo. Another can be valid if containment is verified and the permanent repair is safely sequenced under explicit ownership. Award one point for capacity arithmetic, one for the containment/repair distinction, one for verification, and one for honest deferral.

In an interview, finish by saying which new fact would change your recommendation. This shows judgment under uncertainty without pretending that the first schedule is immutable.

The prospect offers to sign only after the integration, but the access defect is still active. Does the order automatically reverse? No. Contain the privacy harm, then discuss a narrower safe integration proof and realistic commitment. Award one point for retaining the harm boundary, one for exploring a smaller test, and one for stating the trade-off to the decision owner. Do not invent a guaranteed sale from an informal promise.

How do you disagree with a founder who wants all four items? A strong answer presents capacity and consequences, offers a sequence, and asks which outcome should be sacrificed if the priority changes. Follow up with a wrong estimate. A weak answer works silently until the deadline fails. Prioritization is an explicit choice about outcomes under limited capacity, and the reasoning should remain visible when facts change.