What the RDC pre-check actually complains about · cause and fix for each category
Relativity's import machinery is stricter than most, and its error messages are famously terse. The good news: almost every rejection falls into a small set of categories, and each has a mechanical cause you can fix in the file before re-importing.
Records with errors are skipped, not aborted: the load continues, a warning appears when it finishes, and the RDC offers to save a document-level error load file containing only the failed records, with every error listed. The correct workflow is: save the error file, fix those records, and re-import with an Append Load. Re-running the full load instead duplicates every record that succeeded the first time. A log of the errors also lands in the Errors tab. Import settings themselves can be saved to a .kwe file and reloaded for identically formatted follow-up files.
| Symptom | Usual cause | Fix |
|---|---|---|
| Garbled text, shredded fields, or every record failing at once | Encoding mismatch — the RDC detects encoding only from a byte order mark; a BOM-less file loads under the default encoding chosen in settings | Re-encode to one family, or add the matching BOM; see the encoding trap guide |
| Fields split in the wrong places, commas breaking records | Wrong delimiter configuration. Delimiters are selectable from ASCII 001–255; the Concordance defaults are quote 254 (þ), newline 174 (®), multi-value 59 (;) | Match the quote/newline/multi-value settings to the producing file — open it in a viewer and look |
| "Identifier" errors | The Identifier is the only mandatory field. On overlays you may point it at any field — commonly the Bates number — but it must be unique and populated | Populate the identifier; for family loads set Group Identifier to the parent's identifier on every member |
| Date parse errors | Date and time must be one field; time-only values error; dates cannot be zero; formats like 4/9/2018 10:22:00 a.m., 2018-044-09 or 4/9/18 13:30 PM are rejected outright | Normalize dates to a listed format (e.g. 6/30/2017 1:23 PM) in one column; watch the locale of the importing machine |
| Column count mismatches | More or fewer delimiter-separated values in a record than in the header — usually an unescaped comma or a literal line break inside a field | Fix the record: escape qualifier characters (doubled þ) and convert in-value breaks to ®; the DAT grammar has the details |
| Field type / length failures | Writing prose into a fixed-length or date-typed workspace field, or a multi-choice field whose values lack the multi-value delimiter | Map to the right field type, or restructure the column; multi-choice values need the multi-value delimiter between entries |
| Missing files / broken paths after load | Relative vs absolute path mismatch, or forward slashes instead of Windows-style backslashes. Folder Info is backslash-delimited; extracted text and natives accept both relative and absolute Windows-style paths | Rewrite paths to match the repository layout; check case and stray suffixes on Bates-named files |
Most of the table above is checkable client-side, without a workspace: count fields per record against the header, verify one record per physical line, confirm the encoding, normalize dates to a single format, and confirm every referenced path exists in the volume. Run those five checks in the converter and the cross-file checks in the Bates checker, and the RDC error file you save should be empty.