Keep a table footnote outside the numeric grid

How footnotes get into the grid

Statistical tables, price lists and financial statements put notes right under the table: sources, definitions, exclusions, numbered remarks. When the page is turned into cells, those lines can be read as additional rows, often with all the text in the first cell. Footnote markers can also stick to values, so an amount reads 12.50* or carries a small number after it.

A footnote matters for interpreting the table, but inside the grid it gets in the way: it breaks the width of the table, it shows up as text in an amount column, and it can be counted as a record.

Recognizing footnote rows

• In the browser inspector, a footnote line pasted with the data usually has one field, so it adds to Rows with a different width.
• In the numeric normalizer, an amount with a marker attached, such as 12.50*, is held under Values left for review rather than converted.
• In the Tablora app, check the last rows of the grid against the Source · page view; text below the table can be picked up as rows.
• Rows starting with a symbol or a small number followed by a sentence are rarely data.
• In pasted text, a footnote that contains a comma may split into two fields, so it can look like a short record rather than a note.

Separating notes from data

• Keep the source page open while you work.
• Find where the data rows end on the page, and mark the first footnote line.
• In the app, remove footnote rows with Delete row; tap an amount with a marker attached and remove the marker with Save.
• In pasted text, cut the footnote lines out of the table and keep them in a separate note.
• Record which values carried markers, since the marker usually means something about that value, such as an estimate or a revised figure.
• Only then normalize amounts or check totals.

Placeholders that are not footnotes

Some tables print a dash, n/a or a letter code instead of a value, and explain it in a footnote. The Tablora app's XLSX export does not count such placeholders when it decides whether a column holds numbers, and leaves them as text. In the browser normalizer they are held for review like any other non-number. Neither tool turns them into zero, and you should not either: a missing value and a zero are different facts.

After separating everything, compare the row count with the source once more. A table that ends exactly where the page's data ends is ready to export.

Tools

• delimited-inspector