Check CSV data before analysis

Key takeaways

• Check delimiter and column count before you calculate.
• Docsi requires UTF-8 or UTF-16 with BOM as character encoding.
• Leading zeros and numbers over 15 digits stay text.
• Docsi reads only unambiguous ISO dates as dates automatically.
• A CSV export contains exactly one sheet.

Columns first, metrics second

A semicolon-separated file may contain decimal commas. Check headers and a sample row. If an address is split across several columns, resolve the import or source format before analysing it.

Identifiers such as 00127 are not quantities. Leading zeros need to survive. An empty field and numeric zero also differ; explicitly define how missing data should be handled.

Request a reviewable sample

1. Add the CSV and name the meaning and units of relevant columns.
2. First request headers, row count and five representative records.
3. Check decimals, dates and identifiers against the original.
4. Only then define filters, totals and output.

Document corrections

Keep the original CSV separately. When correcting flawed rows, visibly document affected fields and the rule used. A tidy result should be reviewable rather than merely look complete.

Do not silently replace missing values

Define what an empty field means in every affected column. Missing revenue may mean “not recorded yet” and must not silently become zero. For a count, zero may be a valid observation. Request separate lists of empty and invalid fields. Inspect currency symbols, whitespace and percent notation before treating them as numbers in calculations.

When several CSV files belong together, inspect their columns and units separately. Matching headings do not guarantee matching meaning; one export may contain gross figures and another net. Keep originals with a description of the analysis rules. A small unresolved set is preferable to an apparently complete table of guessed values. Calculate metrics after those decisions.

Delimiters and regional settings

CSV stands for “comma-separated values”, but German-language programs often use the semicolon as delimiter, because the comma serves as the decimal separator. Docsi recognises comma, semicolon and tab and allows decimal commas such as “12,50” in semicolon files.

According to RFC 4180, fields containing delimiters or line breaks are enclosed in quotes; quotes inside a field are doubled. Docsi follows these rules. If everything still ends up in one column, the delimiter usually does not match the export.

Recognising character encoding

Docsi reads text files in UTF-8 or UTF-16 with a byte order mark and rejects faulty encodings instead of passing on wrong characters. Older Windows programs sometimes save CSV in a Western European encoding; umlauts then appear as “ä” or question marks.

Open such a file in an editor or in Excel and save it again as “CSV UTF-8”. UTF-8 can represent every Unicode character and is the standard for data exchange. Docsi then reads umlauts, accents and special characters correctly.

Telling numbers and identifiers apart

Postal codes, article numbers, IBANs or customer numbers look like numbers but are identifiers. A leading zero must not disappear, and an 18-digit number must not be rounded. Docsi therefore keeps values with leading zeros and numbers with more than 15 digits as text.

The 15-digit limit matches Excel’s calculation precision; longer numbers would be altered there. Never calculate with identifier columns, and ask Docsi to use them in analyses only for counting or grouping.

Reading dates unambiguously

A date such as “03/04/2026” can mean 3 April or 4 March. Docsi therefore reads only unambiguous ISO dates such as “2026-04-03” as dates; regional notations remain text. That prevents a silent error that would later distort monthly analyses.

If you know your file’s format, state it in your request (“date in DD.MM.YYYY format”). Docsi can then interpret the values for the analysis and add a column in ISO format that you spot-check.

Writing CSV from Docsi again

A CSV export from Docsi contains exactly one sheet. Choose the delimiter to match the target program: semicolon for German-language Excel, comma for most international tools and databases. The file is UTF-8 encoded.

If you need several sheets, formatting or charts, Excel is the better format. CSV is ideal for pure data exchange but loses everything except the values.

A worked scenario

An example: you export member data from a club management program as CSV. Docsi first reports the file as unreadable; in an editor you see that accented letters appear as question marks. You save the file again in Excel as “CSV UTF-8”, and now Docsi reads it without errors.

The column overview shows that membership numbers have leading zeros and the joining date is in DD.MM.YYYY format. You state the format in your request, Docsi adds an ISO date column, and you check five samples. Only then do you have the joins per year counted – with correct dates and unchanged numbers.

Checklist before sharing

Before you pass on results from a CSV file, check the data basis one last time. Most errors arise when reading the file, not when calculating.

1. Delimiter and column count are recognised correctly.
2. Accented letters and special characters display correctly.
3. Identifiers have kept their leading zeros.
4. Date values have been spot-checked against the original.
5. Empty fields are counted and not included as zero.

Key terms explained

Delimiter

The character between the fields of a CSV line: comma, semicolon or tab.

Field

A single value in a CSV line, equivalent to a cell.

Character encoding

The rule for storing characters as bytes, such as UTF-8.

Header row

The first line of a CSV file with the column names.

ISO date

A date in year-month-day format, such as 2026-10-29, unambiguous in every language.