Yes. The strongest natural grouping I found is the dataset's own `Overall_Impact` label (Beneficial / Neutral / Negative). It splits the 4,500 students into three groups that differ sharply on almost every measure. | Segment | Share of students | Avg daily hours | Avg sleep (h) | Avg stress | Avg mental health | Avg GPA | Late-night use | |---|---|---|---|---|---|---|---| | Beneficial | 81.8% (3,681) | 4.42 | 7.03 | 11.16 | 84.26 | 3.55 | 53.5% | | Neutral | 14.5% (654) | 8.50 | 5.48 | 22.30 | 63.42 | 3.00 | 82.3% | | Negative | 3.7% (165) | 12.25 | 4.23 | 31.12 | 47.13 | 2.54 | 81.8% | (Source: the `Overall_Impact` aggregate over `raw.Social_media_impact_on_life`, steps 4–6.) - **The groups form a gradient.** Moving from Beneficial to Neutral to Negative, daily use rises (4.4, 8.5, 12.3 hours). Sleep, sleep quality, mental health and GPA all fall, and stress rises. Every measure moves in the same direction across the three groups, which suggests real structure rather than noise. - **Usage hours are the clearest separator.** Daily hours differ clearly between all three groups. Late-night use only separates Beneficial from the other two (about 53% vs about 82%) and does not distinguish Neutral from Negative. - **Weekend extra hours do not help.** They are almost identical across groups (1.73–1.80), so they are not a useful segmentation axis. - **The groups are very unequal in size.** Negative is only 165 students, so any further breakdown of that group (by platform or age, for example) will rest on small counts. **Caveats** - `Overall_Impact` is a label already in the data. I don't know how it was assigned, and it may partly be derived from the same measures, which would make the separation partly circular. - I did not test whether the other categorical fields (age, gender, academic level, platform, device type, social comparison frequency) produce comparably distinct groups. - This is observational data, so the groups describe association, not cause. For a segmentation not tied to that label, a sensible alternative is usage-hour bands (for example under 5, 5–9 and over 9 hours per day), since usage separates the groups most clearly.