Computing • A PRACTICAL GUIDE

A CSV handover should include more than a file attachment

Preserve identifiers, explain columns and check a round trip before a small data export becomes somebody else’s confusing spreadsheet.

Computer workspace

An export is a small agreement

A file called final-export.csv can look reassuringly ordinary. It has rows, columns and values, and almost every spreadsheet application seems willing to open it. That apparent simplicity hides a collection of decisions: what each row represents, whether a number is a quantity or an identifier, which date convention was used and what an empty cell means. A good handover makes those decisions visible. The recipient should not need to guess whether a leading zero disappeared or whether two entries with the same name describe one person.

Treat a CSV export as a package containing data and a short explanation. The explanation can be a plain text note. It should identify the source, the intended recipient, the export date and the meaning of the columns. Keep personal information to the minimum the task requires. If the original collection contains private details that the recipient does not need, remove those columns in a separate export copy before preparing the handover. Convenience is not a reason to send the entire database.

Keep a master before changing formats

Retain the original workbook or application data in its native format. A CSV file is a text-based exchange format, not a complete replacement for a workbook with multiple worksheets, formulas, formatting and other features. Microsoft's documentation explains the limitations of exporting Excel content into other formats. Use the export to deliver the particular table you need, while keeping the original somewhere you can recover it. Saving repeatedly over the master in a reduced format can leave you without the features that made the original useful.

Give the export a name that describes its purpose and date. Avoid a filename that implies it will remain the final version forever. If you need to revise it after a recipient spots an error, make the revision identifiable and explain whether it replaces the previous file. Keep your working transformation notes separate from the delivered data. That separation helps prevent an internal calculation, hidden comment or unused private column from becoming part of an attachment by accident.

Decide what one row means

Write a one-sentence definition of a row. It might represent one purchase, one item, one registration or one monthly total. These are different structures, even if they share column headings such as name and amount. A recipient who assumes one row per customer may double-count a file that actually contains one row per purchase. Make the distinction before checking totals, because a correct sum does not prove that the table is being interpreted correctly.

Check whether each row has a stable identifier. Names are often poor identifiers because they can change, be duplicated or be spelled inconsistently. If the source has an appropriate record identifier, document what it identifies and whether it is unique within this export. Do not invent an identifier that looks permanent if it is only a temporary row number. A simple row number can help discuss an error, but it should not silently become the key used to join unrelated datasets later.

Protect text that looks like a number

Postal codes, product codes and account references can contain digits without being quantities. Leading zeros may be meaningful. A long identifier may need every character preserved exactly. Document these columns as text and use the recipient application's documented import process to choose the appropriate type. Double-clicking a file can invoke automatic interpretation, which may make choices before you have inspected them. A deliberate import gives you a better chance to see what is happening.

Create a few non-sensitive test records that reveal likely problems. Include an identifier beginning with zero, a code containing a letter and a value longer than an ordinary human-readable reference. Check the imported result against the original characters. Do not try to solve the issue by decorating every value with punctuation that another application will later treat as part of the identifier. Agree on the intended representation and verify it in the actual tool the recipient plans to use.

Make dates and quantities unambiguous

A date such as 04/05 can be interpreted differently across regions. Choose a documented representation that includes the year and explain it in the handover note. If a column contains times, specify whether they are local times, include offsets or refer to a named time zone. An appointment time and the time an event was recorded may need different explanations. Do not remove timezone information simply because the shortened value looks neater in a spreadsheet cell.

For quantities, include the unit in the column name or the accompanying description. A column called weight is less useful than a documented weight column with an explicit unit and a statement about how missing measurements are represented. Currency needs its own clarification, particularly when an export can include more than one currency. Avoid making a recipient infer units from the size of a number. A plausible-looking value can be wrong by a large factor while still fitting comfortably inside a cell.

Give empty cells a meaning

An empty cell can mean unknown, not applicable, not yet collected or deliberately withheld. Those states can matter to the person using the file. Choose a representation that the receiving system supports and explain it. Avoid switching casually between blank cells, zero and a word such as none. Zero is a value, and treating it as missing can distort an average or a count. The right convention depends on the task, but it should not depend on which colleague prepared a particular row.

Check whether the file contains inconsistent category labels as well. A small controlled vocabulary can be easier to analyse than several nearly identical spellings. Make changes only in the export copy, and record any consolidation that changes the meaning of a label. Do not silently merge categories just to make the table look tidy. A recipient may need to distinguish records that seem similar to you. When uncertain, preserve the original value and add a documented interpretation rather than replacing information irreversibly.

Inspect the text as well as the grid

Open a copy in a suitable text viewer and inspect a small sample. You are checking the structure, not reading every row manually. Look for a consistent separator, a header row and sensible handling of fields containing punctuation or line breaks. Use the exporter's documented options rather than constructing a CSV file by joining values with commas yourself. A note that happens to contain a comma should not push the remaining values into the wrong columns.

Character encoding deserves a check when names or descriptions contain accents or non-Latin characters. Use an agreed encoding supported by both applications and test representative text after import. If the grid looks correct in one application, that alone does not prove the file will arrive correctly in another. Also handle files from unfamiliar sources cautiously: spreadsheet software may interpret some cell content as formulas. Follow the receiving application's security guidance and do not enable features or active content merely to make an untrusted attachment open.

Perform a small round trip

Import the exported file into a separate test document or the recipient's intended application. Compare the number of records, the first and last relevant rows, a few representative identifiers and any totals that have a clear meaning. Use a copy rather than overwriting the original source. A round trip is particularly useful when an application makes automatic data-type decisions, because the resulting errors can look like ordinary formatting differences until somebody tries to match records.

Choose checks that could reveal a real problem. If every test identifier contains five ordinary digits, you have not tested the leading-zero case. If every name uses plain Latin letters, you have not tested the encoding question. Keep a short list of these edge cases with your export procedure. Over time it becomes a useful checklist grounded in the kind of information you actually handle. It should remain small enough that people will perform it before sending a file.

Send context through the right channel

Include a readable note describing the row definition, column meanings, units, date convention, text identifiers and known limitations. State whether the file is a complete collection or a filtered subset. If it is filtered, explain the selection in terms the recipient can understand. A table covering completed orders is different from one covering all orders, including cancellations. The distinction should travel with the file rather than survive only in the email conversation that requested it.

Use an appropriate sharing method for the sensitivity of the information. Check recipients and access permissions before sending. Passwords and access secrets should not be placed in the same file as the data they protect. If the delivery is recurring, agree how revisions and corrections will be announced. A recipient may have already imported an earlier version, so replacing the attachment in a shared folder without an explanation can create two conflicting versions of the same work.

Leave a repeatable export procedure

Save the steps that produced the export, including the source view, selected columns, filters and transformation rules. Record the checks you actually performed rather than writing a broad assurance such as data verified. If a recipient later finds a problem, these notes make it easier to locate the stage where the value changed. They also help another person repeat the export without depending on your memory or on a sequence of undocumented clicks.

The goal is a handover that remains understandable after the original conversation is forgotten. A small CSV file with a clear row definition, preserved identifiers and a tested import can be more dependable than a much more elaborate file with no explanation. Keep the master, make the exchange copy deliberately and verify the parts most likely to be misread. Then the recipient can spend time using the information rather than reverse-engineering what the columns were supposed to mean.

Further reading

Questions or a correction? Contact Tian Xin Kai at [email protected].

General educational guidance. Features vary; check your device documentation before changing settings.

Search the journal

Privacy choices

Google advertising is integrated. Consent settings are not available until the consent platform is configured.

When Google's consent platform is configured, this link will reopen its settings. This information panel does not record consent.

Read our Privacy Policy

Contact: [email protected]