Test strategy, automation framework design and the test-case generation round that separates QA engineers from script writers.
The round that decides QA and SDET loops is test case generation. You are given a deceptively small feature - a login form, a discount code, a file upload - and asked what you would test. Weak answers list the happy path and a couple of invalid inputs. Strong answers work through boundaries, state transitions, concurrency, idempotency, permissions and failure injection, and then say which of those are worth automating and which are not.
SDET loops add a real coding round. It is usually easier than an SWE round but genuinely coded, and the framework design question that follows is a low-level design question in disguise: page objects, fixtures, data builders, parallel execution, flaky test isolation and reporting.
Interviewers listen for prioritisation. Infinite tests are easy to propose; a test strategy with a stated risk model, a pyramid shape and an explicit decision about what you will not cover is what reads as senior.
45 min
Given a feature, generate and prioritise cases. Boundaries, state, concurrency, failure modes.
60 min
Layering, page objects, fixtures, parallelism, flakiness handling and reporting.
45 min
Write real automation or a small utility. Lighter than an SWE round but genuinely coded.
45 min
Release gating, CI integration, risk-based prioritisation, and a bug that escaped to production.
Structure and prioritisation. Work through the input space systematically - boundaries, equivalence classes, state transitions, concurrency, permissions, failure injection - then rank by risk and say explicitly which cases you would automate and which you would leave manual, with a reason.
Real but moderate. Expect to write working automation and a small utility, plus a framework design discussion. It is closer to a low-level design round than to a competitive programming round.
Increasingly folded into one loop. Most roles now expect both: the judgement to design a strategy and the ability to automate a meaningful portion of it. Loops that are purely manual are becoming rare at product companies.
Design a test strategy and defend what you chose not to cover.