Look beyond row counts to names, field types, source IDs and links to other records.
The import reports success, but a candidate’s name contains broken characters and the note belongs to the wrong record. A completed upload doesn’t establish that the information is usable.
Candidate CSV imports need checks at three levels: the file must be read correctly, each column must reach the right field, and references must lead to the right record. This guide covers those file-level checks. A full ATS migration also involves permissions, history and deciding when the team switches systems.
First, establish what one row represents. A candidate, a contact event and an application are different things. A person appearing several times doesn’t necessarily mean there are duplicate candidate profiles. The file may contain separate contact events.
Record the selection exported and when it was produced. An active-candidates export can’t be compared directly with the total number of people in the old ATS. Give the working copy a recognisable version and keep the original separately.
A CV filename isn’t the CV itself. Check whether documents are supplied separately and how they connect to candidates. A reference to a file on an old computer is of little use if the new system can’t read that file.
Check the export’s delimiter and the importer’s expected format. If an entire row appears in one column, stop checking individual fields: the file first needs to be read correctly.
A note may contain commas, quotation marks or line breaks. RFC 4180, an informational description of a common CSV format, describes enclosing such fields in quotation marks and doubling internal double quotes. The importer needs to handle the format actually used.
Count records with a suitable CSV reader, rather than counting visible text lines alone. A multiline note may belong to one candidate. Manually removing every line break can damage the original content.
Correct columns don’t prove the characters were read correctly. Include names with accents in the trial file and compare them with the source. Microsoft explains how to open UTF-8 CSV files in Excel, including an import route when ordinary opening doesn’t work as intended. Use settings suited to your export.
Then check how values are interpreted. Excel’s text-import documentation explains that defaults affect date and number interpretation. Importing a column as text can preserve leading zeros.
A source ID isn’t a quantity. 000427 and 427 may be different values for your record links. Agree an unambiguous date format with the person running the import and check a date that exposes a day/month mix-up. Recover conversion errors from the preserved source rather than guessing missing characters.
A mapping describes more than the destination field label. Write down the meaning of the source value, where it belongs and any required conversion. Have someone who knows the recruitment process check that meaning.
| Source column | Agreement before import | Result to check |
|---|---|---|
| Candidate ID | Retain as a source reference | Exactly the same characters remain |
| Name | Full name or agreed separate fields | Accents and name particles survive |
| Contact date | Date of contact, with source format | Hasn’t become the record creation date |
| Owner | Reference to the correct user | Unknown users are handled explicitly |
| Note | Candidate or contact event, as defined by the source | Content and relationship are correct |
| CV reference | Separate file with an agreed link | Document opens from the correct record |
Agree what an empty value means too. Is it ‘unknown’, ‘not applicable’ or ‘keep the existing value’? When updating records, you need to know whether an empty cell clears a populated field. Import functions differ, so test that behaviour instead of assuming it.
Take a fictional candidate, Zoë, with source ID 000427. The export contains a note with a comma and two lines of text, plus a reference to a separate CV. Another file links a contact event to the same source ID.
Define the expected result before importing: the name remains readable, the ID is unchanged, the note stays together and the contact event belongs to Zoë. Check the CV separately. Importing the candidate row successfully says nothing about whether that document moved with it.
Include a missing reference in the trial, such as a contact event linked to an unknown candidate ID. See whether you can identify the unprocessed record and the reason. The desired response is an agreed test outcome, not a claim about what every importer already provides.
Keep the results report and record what was created, skipped or rejected against each source ID. If the importer doesn’t provide a useful report, agree how the person running it will demonstrate what happened.
Don’t immediately run the same file again. First establish whether existing records will be recognised, updated or created again. A failed relationship doesn’t mean all previously imported candidate records have been reversed.
Ask about recovery before the trial. Can the affected batch be identified? Who decides which changes can safely be undone? Preserving the source file matters, but it doesn’t itself provide an undo function in the destination ATS.
Proceed when the agreed checks have passed and differences have been explained or consciously accepted. Keep the mapping, file version and results together. A changed export format or mapping calls for another check of the affected fields.
When moving to Simply, our team reviews the export with you. We agree which records should move, help prepare the selection and set up the migration for you. This migration assistance is temporarily free. The file checks above help us agree what needs to arrive correctly; you do not have to manage the whole transition yourself.
We choose the processing route around your data and preferences. Our migration team handles the import for you, or you import it yourself with the CSV importer and set up your workspace together with Simply Ask. In Simply you decide which data AI may process. Bring a description of your export and a few difficult fields to a migration discussion, then we can agree how to carry out the work.
About the author
Remo Vloet is a co-founder of Simply, the AI Operating System for recruitment agencies: inbox, meetings, sourcing, CV parsing, search and matching, documents and automation in one system. With a background in building complex software, he contributes to the technical vision behind Simply.
LinkedInArticles by Remo VloetThe file type doesn't establish completeness. A CSV may contain only some records and fields. Documents, relationships, permissions and history need separate checks.
Work on a copy and check how columns are read. Automatic interpretation can change values before you save them again. Preserve the original export.
First establish what a row represents. Repeated candidate IDs can belong to separate contact events or other related records. Removing rows based only on a name can remove useful information.
No. Check values and relationships using recognisable examples too. An unwanted duplicate can hide a missing record in the overall count.
A general checklist can't promise that. A preserved source, explicit mapping, trial import and documented checks help reveal errors and prepare for recovery.
A QBR (Quarterly Business Review) is the quarterly review with your client. How to run one that retains and grows the account, and why data quality decides it.
Remo Vloet9 min.
In recruitment, ICP means two things: Ideal Customer Profile and Ideal Candidate Profile. Here is how to build both, and why data quality makes the difference.
Remo Vloet9 min.
AI screening tools read and rank candidates in minutes. Here is how the main types work, where bias and compliance risk creep in, and how to choose one.
Remo Vloet9 min.
Find out how Simply can completely evolve your workflow. No slides, just product.