Marketing wants warehouse segments pushed into the campaign tool hourly. What do you require before agreeing?

What the interviewer is testing

Whether you recognise that this converts an analytical pipeline into an operational dependency.

What changes

Before: if the transformation fails, a dashboard is stale and someone notices next morning. After: the campaign tool acts on the data, so a failure means wrong customers receive wrong messages, or a segment silently does not update and a campaign targets last week's audience.

The pipeline now needs an operational service level it almost certainly does not have.

What I would require

An operational SLO and monitoring. Freshness and completeness targets, alerting on absence — not just on failure — and a named owner with an on-call path. A pipeline monitored only by "did the job report success" cannot support this.

Consent and suppression enforcement at push time. This is the requirement most often missed and the one with regulatory consequence. A segment computed last night and pushed this morning must respect anyone who withdrew consent in between, and it must respect suppression lists. Checking consent when the segment was computed is not sufficient.

Data minimisation. Push the fields the campaign tool needs, not the full customer profile. Every additional attribute is personal data in another system, in another jurisdiction possibly, in scope for breach notification.

Idempotency and reconciliation. Pushes will retry; the tool must not create duplicate audience entries. And a periodic comparison confirming the tool's audience matches the warehouse segment, because silent divergence is the normal failure mode.

Failure semantics agreed in advance. If the push fails, does the campaign run against the previous segment or not run at all? Marketing must choose, before it happens.

What a strong answer adds

Asking whether hourly is genuinely needed. Warehouse batch latency means the data is already hours old at best; pushing it hourly does not make it fresh, it makes the pipeline more fragile. If the use case truly needs recency, the answer is a streaming path or a direct service call, not a faster warehouse round trip.

Common weak answers

Treating it as a scheduling change. Agreeing without the consent enforcement point.