FAQ and troubleshooting
The questions that come up in the first week.
Do I have to set up a profile before my first run?
No. If no profile matches the file, the built-in generic profile runs the structural checks that hold for every language: broken XML, missing targets, tag and placeholder corruption, number mismatches. A profile is how you add what is specific to one client. Set up a client.
The same finding was minor on one report and critical on another. Why?
Two different profiles, or one profile and a tier above it, set that check's severity differently. Severity is a per-profile decision, not a global one. The report says which profile ran. The other cause is Which findings count in the profile's Checks. In a file uploaded in the app, a finding outside the editor's remit is demoted to minor, or removed when the profile is set to leave such findings out. In a Dropbox folder run it is left out of the report and its segment is sent to no AI review, the whole-file review included. Only a finding that shows an inconsistency with another segment stays at its usual severity. It is not sent to the AI either. Evaluate everything turns this off.
Does the AI pass send my whole file to a model?
Yes, by default: every translated segment with a target goes to the AI, because a reversed meaning or a dropped sentence usually sits in a segment where no rule fired. In a Dropbox folder run, a segment outside the role's remit is never sent. A profile can narrow it to the segments the rules flagged, which is cheaper; the report then says reviewed flagged segments instead of reviewed every segment. Your segments are never used to train models and your admin chooses the providers. Security and privacy.
What happens if the AI provider is down or out of quota?
The rule layer still runs and the report is still produced: the AI lane degrades on its own, and the report says so in a banner at the top. If no call got through, the AI status is failed; if only some did, it is partial. Either way it is never a quiet pass, and rule findings are unaffected.
Where is the Spelling report?
A file has one report and one row. The spelling pass adds its confirmed misspellings to the same report under their own Spelling error type, with their own group and filter box. While it runs the report says Spelling: checking. Spelling hits are never scored. Spelling.
My spreadsheet was rejected with "no source/target column".
The header row could not be identified: usually a title or banner row above it, or column labels the tool does not recognise. Map the columns by hand in the Confirm the columns dialog, which overrides detection entirely. Check a spreadsheet.
The run says FAILED but I can see findings. Is the run broken?
No. FAILED is the file's result, not the run's: the check completed and found errors. A run that could not complete says so, names the reason and produces no findings at all.
If I approve a Learn suggestion, does it change my old reports?
No. Approving changes the profile, and the profile applies to runs from that point on. Past reports are kept exactly as they were produced.
Can one client's terms leak into another client's checks?
No. Protected terms, glossary and AI guidance live inside the profile that owns them. A file that does not match that profile never sees them. Cascading only ever flows down the tiers you set up yourself: org, then account, then division.
Can I show a report to someone without an account?
Yes. Mint a share link on the report. It is signed, tied to that one report and workspace, expires after 30 days and is view-only. Creating one is a PM or admin action. Sharing and exporting.
A Trados package failed to parse as "malformed XML".
Re-run the file once. If it still fails, the file itself is broken: open it in the CAT tool and re-export.
Where do reports go when Dropbox, email or the API starts the run?
Into the same Reports list as a run you started yourself, tagged with the trigger that started it. Dropbox additionally writes the report back beside the file it checked.
Why does Help look shorter for a colleague than for me?
What someone sees follows their role, and so do the pages here that describe a feature their role does not have. Workspaces and roles.
Something else
Tell us the time, what you did and the report link. Never send a key or a token.