Interview: 'what's the biggest mistake you've seen in a real clustering project?'
mediumAnswer
- (Answering as a candidate) '(1) Not scaling features — one feature dominated distances → nonsense clusters.
- (2) Trusting elbow / silhouette alone without stability check → non-reproducible clusters.
- (3) Skipping outlier removal for k-means.
- (4) Presenting clusters without profiling — stakeholders couldn't act on them.
- (5) Not monitoring drift — clusters silently degraded over 6 months.
- Lesson: cluster is 30% algorithm, 70% feature engineering + evaluation + stakeholder communication.'
Check yourself — multiple choice
- Nothing
- Not scaling + trust one metric + no outlier handling + no cluster profiling + no drift monitoring — 70% is feature-eng + eval + comms, not algorithm choice
- Random
- Only algorithm
Common mistakes: no scaling / single metric / no outlier / no profile / no drift — 70% non-algo.
#interview#clustering
Practise Unsupervised Learning
214 interview questions in this topic.
Related questions
- Interview: 'you're asked to segment 10M customers — how?'
- Interview: cluster 5M e-commerce customers for a marketing campaign.
- Interview: how would you group log messages from a large distributed system?
- Interview: business asks 'why is customer X in Cluster 3?' — how do you explain?
- Interview: 'you have to pick k for the executive team — walk me through it.'
- Interview: after segmentation, one segment has 60% women but each region shows 40%. Why?