A CSV import should let users understand column mappings, validation results and the records that will be created or changed before committing. Define duplicate handling and item-level outcomes. Accepting a file does not establish that its rows have become valid business records.
Let’s plan your data import workflow.
Bring your workflow and the decisions you need to make. We’ll help you clarify requirements, review the trade-offs and define a practical next step.
Publish a meaningful template
Explain required columns, identifiers, allowed values, date conventions and units. Include a clearly labelled example without real customer data. Avoid a template whose headings differ from the product's terminology.
For a hypothetical inventory import, item code, location and quantity meaning are essential. A spreadsheet containing 10 boxes should not become 10 individual units because the import omitted the unit field.
Specify the import stages
| Stage | User decision | Evidence to show |
|---|---|---|
| Upload | Is this the intended file? | Name, size and permitted preview |
| Mapping | Which column serves each field? | Required and unmapped fields |
| Validation | Which rows violate rules? | Row number, field and useful reason |
| Duplicate review | Create, update, skip or investigate? | Matching basis and proposed effect |
| Commit | What population will change? | Counts and important consequences |
| Results | What succeeded or failed? | Stable job and item-level outcomes |
Define whether the import is atomic or supports partial success. Both can be legitimate, but the user must understand which applies.
Make retries predictable
Keep a job reference and a consistent matching rule. Reuploading the same file should not accidentally create a second copy of every record. Where updates are allowed, define what happens when the destination changed after preview.
Validate on the server and apply record permissions. A successful browser preview must not bypass business rules or permit cross-tenant updates. Treat cell contents as untrusted data throughout display and any generated spreadsheet output.
Support correction without losing context
Provide an error report that preserves row references and explains corrections without exposing unrelated records. Allow users to understand whether they should retry failed rows or submit a complete revised file.
Test different encodings, quoted values, empty cells, long identifiers and representative file sizes. Verify leading zeros and date interpretation where they matter to record identity.
CSV import questions
Should import silently fix invalid values?
Only where the transformation is explicit and safe under agreed rules. Show consequential changes before committing them.
Does a matching name prove a duplicate?
No. Use the approved identifiers and matching policy, with review for ambiguous records.
Use the data acceptance checklist for broader reconciliation, and discuss import scope with Drema's enterprise software team.
