## Distribution summary for the numeric columns Most numeric columns are stored as text in the raw table, so I parsed them to numbers first. The figures below come from the full profile in step 6, with the same parsed counts as steps 1 and 4. Only Productivity_Score, Accuracy_Rating and Satisfaction_Score are typed numeric; the rest are VARCHAR. Some values didn't parse, and they are excluded from the stats, so "n" is below 300 for every column. | Column | n parsed | Min | Median | Mean | P95 | Max | Std dev | |---|---|---|---|---|---|---|---| | Age | 263 | 16 | 27 | 30.18 | 54.9 | 97 | 12.47 | | Monthly_Income | 290 | 0 | 521.5 | 2,304.38 | 6,399 | 250,000 | 14,743.44 | | AI_Usage_Hours_Per_Day | 271 | 0.5 | 2.2 | 3.01 | 7.55 | 23 | 2.45 | | Monthly_AI_Cost | 288 | 0 | 0 | 9.84 | 30 | 100 | 13.75 | | Productivity_Score | 284 | 1.6 | 5.0 | 5.23 | 8.6 | 10 | 1.82 | | Accuracy_Rating | 285 | 2 | 4 | 3.43 | 5 | 5 | 0.86 | | Satisfaction_Score | 293 | 3 | 6 | 6.15 | 9 | 10 | 1.52 | | Time_Saved_Hours_Per_Week | 272 | 0 | 5.05 | 5.81 | 13.54 | 19.4 | 4.11 | | Work_or_Study_Hours_Per_Day | 279 | 2.1 | 7.2 | 7.29 | 10.52 | 20 | 2.14 | ### Distribution shapes - **Roughly symmetric:** Productivity_Score, Satisfaction_Score and Work_or_Study_Hours_Per_Day have means within about 5% of their medians. Productivity is centred near 5 on a 1.6–10 range, satisfaction near 6 (3–10), and work/study hours near 7.2. - **Mildly right-skewed:** Age (mean 30.2 vs median 27, up to 97), Time_Saved_Hours_Per_Week (mean 5.81 vs median 5.05) and AI_Usage_Hours_Per_Day (mean 3.01 vs median 2.2, a 1.37× ratio). The 95th percentile of daily AI use is 7.55 hours, but the maximum is 23. - **Heavily skewed:** - **Monthly_Income:** the mean is 4.4× the median (2,304 vs 521.5). The standard deviation of about 14,743 is far larger than the mean, and the 95th percentile is only 6,399 against a maximum of 250,000. A few extreme incomes drive the mean and spread, so the median and percentiles describe a typical user better. - **Monthly_AI_Cost:** the median is 0, so at least half of users pay nothing, while the mean is 9.84 and the maximum is 100. This is a point mass at zero plus a paying tail, not a bell shape. - **Accuracy_Rating:** it only ranges from 2 to 5 with a median of 4, so it sits above its mean of 3.43. The mean-to-median ratio of 0.86 is the only one below 1, which points to a left-skew, with most ratings at 4–5 and a few low ones. ### Data quality caveats - **Unparsed values:** - Age: 37 of 300 values didn't parse. - AI_Usage_Hours_Per_Day: 29. - Time_Saved_Hours_Per_Week: 28. - Work_or_Study_Hours_Per_Day: 21. - Monthly_AI_Cost: 12. - Monthly_Income: 10. - The typed columns are missing instead: Productivity_Score 16, Accuracy_Rating 15, Satisfaction_Score 7. - **Parsing sensitivity:** a stricter parse of the two currency-like columns (steps 3 and 5) kept fewer rows. Monthly_Income had 284 rows (mean 2,328.96, median 520) and Monthly_AI_Cost had 273 (mean 9.76, median 0). The conclusions are the same, but the exact figures depend on how the text is cleaned. - **Suspicious extremes:** an age of 97, 23 hours of AI use per day and 20 work/study hours per day look like outliers or entry errors. Worth checking before using the means for decisions. - **Privacy:** Monthly_Income is a person-identifying column, so only aggregates are shown here.