**Distribution summary for the nine numeric columns** The table from the last profiling query (step 8) shows min, quartiles, mean, p95, max and standard deviation for each column. Most columns are stored as text, so each was parsed to a number. Values that failed to parse are counted as missing. **Skewed columns** - **Monthly_Income** is by far the most skewed. The median is 520 and the mean is 2,329. The p75 is about 1,618 and the p95 about 6,399, but the max is 250,000. The standard deviation of about 14,897 is mostly driven by that outlier. The median and quartiles describe a typical user much better than the mean. - **Monthly_AI_Cost** ranges from 0 to 100. The median and p25 are both 0, so most people pay nothing. The mean is 9.76, the p75 is 20, the p95 is 30 and the standard deviation is 13.65. - **AI_Usage_Hours_Per_Day** ranges from 0.5 to 23. The median is 2.2, the mean 3.01 and the p95 7.55. A max of 23 hours a day is implausible and is probably a data-entry error or outlier. - **Age** ranges from 16 to 97. The median is 27 and the mean 30.18. The middle 50% fall between 22 and 35, and the standard deviation is 12.47. The right tail (p95 of 54.9, max of 97) pulls the mean above the median. - **Time_Saved_Hours_Per_Week** ranges from 0 to 19.4. The median is 5.05 and the mean 5.81. The interquartile range is 2.7 to 7.75, the p95 is 13.54 and the standard deviation is 4.11. It is mildly right-skewed. **Roughly symmetric columns** - **Work_or_Study_Hours_Per_Day** has a median of 7.2 and a mean of 7.29. The interquartile range is 6.1 to 8.4 and the standard deviation is 2.14. The range of 2.1 to 20 includes a few extreme high values. - **Productivity_Score** ranges from 1.6 to 10, with a median of 5.0 and a mean of 5.23. The interquartile range is 3.9 to 6.5 and the standard deviation is 1.82. - **Satisfaction_Score** ranges from 3 to 10, with a median of 6 and a mean of 6.15. The interquartile range is 5 to 7 and the standard deviation is 1.52. No one scored below 3. - **Accuracy_Rating** is a 2 to 5 scale with a median of 4 and a mean of 3.43. The interquartile range is 3 to 4 and the standard deviation is 0.86. The mean sits below the median, so it is slightly left-skewed, with most ratings at 3–4 and a tail of low ratings. **Missing data** - Valid counts out of 300 rows are: - Satisfaction_Score: 293 - Accuracy_Rating: 285 - Productivity_Score: 284 - Work_or_Study_Hours_Per_Day: 279 - Monthly_Income: 284 - AI_Usage_Hours_Per_Day: 271 - Time_Saved_Hours_Per_Week: 272 - Monthly_AI_Cost: 273 - Age: 263 - Age has the most missing values (37 of 300), and Satisfaction_Score the fewest (7). - Monthly_Income and Monthly_AI_Cost are the only columns where my parsing attempts disagreed. A more lenient parse gave 290 valid incomes and 288 valid costs, with a mean income of 2,304 and a mean cost of 9.84. The stricter parse in the table gives 284 and 273. The conclusions don't change, because the median income is about 520 either way and the median cost is 0. **Bottom line:** the scores and rating columns are well-behaved and centred in the middle of their scales. Income, AI cost and daily AI hours are right-skewed with outliers, so use their medians and percentiles rather than their means.