Clean a messy data export before import

Exports from ticketing systems, time trackers and older finance packages rarely arrive in the shape an importer wants. Rows may be glued together with semicolons, a header may sit on the wrong line, or a column of reference codes may have lost its leading zeros somewhere along the way. This job is the repair routine: turn the raw block into columns, choose the separator the target system expects, then look at the first and last rows before anyone imports the file. Checking beats hoping, because a silent mis-split is far more expensive than a minute of inspection. Everything is parsed locally, so customer names, salaries and order references never leave the machine you are working on, and you can keep re-splitting until the table reads correctly. A quick suspicion test saves an import window. Count the rows in a spreadsheet, then count them again in the repaired preview; a mismatch usually means an embedded newline split one record in two or a stray quotation mark swallowed several lines. Look for a header that appears twice, which happens when an exporter concatenates several monthly files, and for a trailing total row that an importer will treat as data. Confirm the encoding before anything else, because a Latin-1 export read as modern text produces mangled names that look like corruption but are merely mislabelled. Keep the repair reproducible by writing down the delimiter, quote style and header row you settled on. Import a small sample before the whole set, so a mistake costs seconds rather than a rebuild. Dates and currencies are the quiet casualties of a careless repair. A column that mixes day-first and month-first formats will import without complaint and produce nonsense downstream, so standardise the order before handing the file on. Currency symbols embedded in a number field break arithmetic, and thousand separators vary by locale; decide the target convention and apply it deliberately. Record those decisions in a short note beside the file, and keep the untouched original so a disputed value can be checked against the raw export.