Drop in a workbook — or use the deliberately messy sample — and see exactly how CapStack turns it into a model: where it finds the header row, which columns it matches and on what evidence, what it refuses to guess, and the cap table that comes out.
The sample is generated here, in this page. It is deliberately awkward: a four-row title block above the headers, a merged header cell, ownership as percentages with no share counts, one share figure stored as text with commas, dates as both Excel serials and MM/DD/YYYY, an #N/A and a TBD, two tables on one sheet, and a sheet whose name is a lie.