Turn rows into a useful view.

The sample signup values are ten, twenty and fifteen. At eight pixels per signup, adjacent bars eighty, one hundred sixty and one hundred twenty pixels long add to three hundred sixty pixels, representing forty-five signups. Dividing forty-five by three numeric rows gives an average of fifteen. These are supplied sample values, not a measured live acquisition funnel. The dashboard totals local CSV content and explicitly does not infer causation. For an illustrative column containing ten, blank, text and five, four data rows produce three finite numeric values: ten, zero and five. The source converts blank to zero and excludes the text value. At one hundred pixels per row, bars compare four total data rows with three numeric rows. Total is fifteen and average is five. This coercion can change the meaning of missing data. A column is offered if any nonblank cell converts to a number; the parser simply splits commas and does not handle quoted CSV fields. Build saves one raw CSV string locally and exports one fixed-format text summary of the chosen metric. Bars compare one saved source string with one exported document at two hundred pixels per count. Editing without Build keeps the prior dashboard and summary. An invalid Build clears parsed data but leaves old visible statistics, so Download refuses until a valid Build. Reload rebuilds saved data and selects the first numeric metric, rather than restoring the prior metric. Reset clears storage and returns to the unbuilt sample.