Data profiling
See the type, the empty cells and what needs fixing in every column before you change anything.
What each header shows
Each column header shows a histogram and three counts, worked out from the first 1,000 rows. Types are inferred separately, from a random sample of 1,000 rows: about 20 are recognized, including numbers, dates, emails and currencies, and a value that does not fit its column's type counts as invalid.
| Indicator | What it shows | Use it to |
|---|---|---|
| Histogram | The shape of the column: where values cluster and where they do not | Spot a skewed or odd distribution |
| null | How many cells are empty | Decide whether to drop or fill them |
| invalid | How many values do not fit the column's type | Find values to fix |
| unique | How many distinct values exist | Catch duplicates in an ID column |
- 1 Raw import, before profiling.
Two groups: To fix and Tidy up
Open Health from the toolbar. It sorts what it finds into two groups. To fix holds values that do not fit their column, such as a word in a number column. Tidy up holds clean-ups that take one click, such as Repeated rows, Cells that say N/A or similar, Empty rows and an Almost empty column.
Act on a line
Click a line to jump to the first affected cell, then step through the rest. Each line has one button, such as Fix or Rewrite. Health then shows one message with the count and an Undo. The step is added to your template.