I interviewed with Depop for a backend engineering role and reached the pair-programming round. The exercise involved matching buyers to sellers based on the items sellers have in their inventory. The problem itself was reasonable, but the interview felt far less collaborative than advertised. The panel remained mostly silent, and there was barely any discussion around design choices during the session. The rejection feedback later focused on “problem decomposition,” specifically because I didn't prioritise the algorithm before minimal domain modelling. That expectation was never communicated during the interview. If a particular approach is strongly preferred, it would be more fair and productive to discuss it during the exercise rather than only mentioning it afterward. Without that dialogue, the process can feel more like a silent evaluation with hidden criteria instead of "pair programming".