ATS Field Mapping Workbook for Recruitment Agencies 2026

Your test import says 18,400 candidates arrived. Nobody notices that 420 contact objections became blank, salary currencies were flattened and three years of recruiter notes lost their author. The record count is right. The migration is not.

What belongs in an ATS field-mapping workbook?

An ATS field-mapping workbook should record every source field, destination field, transformation, authoritative owner, privacy classification, missing-value rule, relationship dependency, acceptance test and named approver. Reconciliation must cover candidate identities, mandates, attachments, history, objections and exception counts—not candidate totals alone.

Treat the workbook as an agreed operating contract. A field marked “do not migrate” needs the same explicit owner and reason as a field marked “required”; otherwise an accidental omission becomes indistinguishable from a deliberate retention decision.

Run the exercise on controlled samples before cutover. A real-looking dataset with former candidates, multilingual CVs, withdrawn permission and conflicting employment dates will expose more than a clean demonstration import.

Inventory objects before columns

Candidate profiles, companies, contacts, mandates, submissions and attachments represent different records with different links.

Start with objects and their stable identifiers. A person can have several email addresses, many role submissions and multiple employment entries; a company can have different client contacts and several confidential search mandates. A flat CSV may display those relationships without preserving how they connect.

Ask for a sample export containing profiles, notes, documents, activities, stage changes, permissions and rejection or objection states. Identify objects that require separate files or API extracts. Mark fields containing sensitive information, internal opinions or data that should not move at all.

• Assign one stable source identifier and one destination identifier to each object.
• Document parent-child relationships before deciding the import order.
• Separate files the agency owns from client-controlled or shared material.

Define field meaning, owner and source of truth

Matching column names does not mean the underlying values describe the same thing.

Write the operational meaning of each field. “Owner” might mean the consultant who created a profile, the colleague with the active relationship or the person accountable for the current mandate. When evaluating recruitment CRM software, map those concepts separately instead of allowing one destination dropdown to erase the distinction.

Give every important value a source-of-truth owner. A candidate-confirmed mobile number deserves different treatment from a contact detail inferred from a public profile. Add provenance, verification date and confidence where they already exist; migration must not upgrade an unverified guess into an established fact.

Map resume parser output without manufacturing certainty

Imported CV fields should preserve parser origin, original evidence and the difference between extracted text and recruiter-confirmed information.

CV parsers regularly meet compound surnames, overlapping contracts, dual-language job titles and dates written in different formats. Capture original document reference, parser output and human correction where possible. Do not overwrite a confirmed candidate statement with a newer but unreviewed extraction simply because its timestamp is later.

Create transformation examples for seniority, employment periods, current employer, location, notice period and compensation currency. A missing end date is not proof that someone still works there. Keep unknown values visibly unknown and route contradictory extracted history to a reviewer.

“Imported successfully” only describes file handling. It says nothing about whether an AI resume parser preserved the meaning of the candidate record.

Preserve relationship history and candidate choices

The most expensive fields to lose are often notes, communication chronology, restrictions and explicit objections.

Map recruiter notes with their original author, timestamp and purpose. Preserve which client saw a candidate, when a submission happened, what was agreed and whether the candidate declined future contact. A generic activity entry without the underlying actor or role can distort both relationship history and later review.

Handle objections, withdrawn consent and channel restrictions as protected states. Suppression records need priority over marketing-style activity labels, scheduled sequences and profile enrichment. If the target system cannot represent a required restriction, stop the relevant migration rather than importing everyone as contactable.

Write transformation and collision rules

Every conversion needs a predictable output and a safe failure case.

Document date formats, time zones, currency, employment status, language codes, phone normalization and custom stage translations. Keep the original value where the business meaning could change. “Available” in one system may mean interviewed last week; elsewhere it may indicate an active job search.

Plan collisions before the trial import: shared household email, reused mobile numbers, identical names and inconsistent company identifiers. Define when a record stays separate, when a potential match enters review and who may approve a merge. Automatic cleanup must not combine two people merely because a parser produced similar text.

Reconcile counts, relationships and meaningful samples

Reconciliation compares expected and observed results at the level where an error would matter.

Count candidates, contacts, companies, mandates, attachments, activity events, submissions, suppression states and rejected rows separately. Record exclusions with their approved reason. Compare totals by desk, country, responsible consultant and mandate so a missing regional dataset cannot hide inside a correct overall count.

Then inspect a small risk-weighted sample. Include a withdrawn candidate, one multilingual CV, a person with two active mandates, a corrected employment date, a confidential client relationship and an attached document. Trace each expected value from source to destination and record discrepancies with an owner and retest date.

1. Freeze the approved mapping version and capture source extract totals.
2. Import into a controlled environment and classify every rejected or transformed row.
3. Check object, segment, relationship and suppression counts against the signed expectations.
4. Walk through representative candidate and mandate histories with the responsible recruiters.
5. Block cutover until material mismatches are fixed or explicitly accepted.

Rehearse cutover, rollback and residual deletion

A migration is not complete until the team knows where new changes land and how failure is reversed.

Define the final export time, read-only boundary, incremental changes and destination owner. Rehearse what happens when a candidate withdraws consent between export and import, when a consultant adds an interview note during the freeze or when a client-facing shortlist is still active.

Keep rollback time-bounded and access-controlled. Decide who can reopen the source, how current restrictions survive a reversal and how temporary transfer files are deleted. Document supplier responsibilities without claiming that an agency-wide commercial export automatically equals an individual candidate’s Article 20 portability right.

ATS field-mapping and reconciliation worksheet

A useful worksheet makes the source meaning, destination behaviour and acceptance evidence visible to the people who run the recruiting process.

What a mapping workbook cannot guarantee

A signed workbook does not prove that a source system was accurate or that a supplier export is complete. It defines what the team expects to test; recruiters still need to validate important facts and review unresolved conflicts.

Commercial migrations, contractual export rights and individual data-portability rights are different questions. Ask qualified advisers to assess contracts, jurisdiction and specific data-subject requests rather than relying on a generic checklist.

ATS field mapping FAQ

These answers focus on the practical gaps that tend to surface after a candidate database appears to import correctly.

Should we map every historic ATS field?

No. Map every field to an explicit decision: retain, transform, restrict or exclude. Keeping information without a continuing purpose can be as problematic as losing necessary candidate or client context.

How do we migrate AI resume parser data?

Preserve the source document, extracted value, review state and any recruiter correction where available. Never treat an inferred employer, date or location as candidate-confirmed simply because the parser populated a field.

What is a good migration reconciliation test?

Check object totals, desk-level segments, linked mandates, attachment counts, stage history and contact restrictions. Follow several realistic profiles from the source export through the destination workflow.

Can we import objections after candidate profiles?

Only if outreach and automation remain disabled until the suppression data is present and verified. A candidate who objected must not temporarily become contactable during the cutover.

Official data accuracy and migration principles

These primary sources explain accuracy, individual rights and recruitment-record handling; commercial migration rules must still be reviewed for the agency’s own context.

• EUR-Lex: General Data Protection Regulation, including accuracy, storage limitation and data-subject rights
• UK Information Commissioner’s Office: Accuracy principle and correcting personal data
• European Data Protection Board: Respect individuals’ rights

Plan the rest of the migration

Plan ATS implementation and data migration · Understand AI CV parser accuracy · Plan an ATS migration without losing data · Review Yena for executive search