Back to all articlesCRM migration

Validate Imported CRM Records Beyond a Successful Import Message

Alactic Editorial10 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

A CRM import can finish successfully while assigning contacts to the wrong companies, replacing meaningful values with blanks, or turning historical activity dates into import dates. Completion tells you that processing ended. It does not establish that the resulting records are correct.

To accept an imported batch, reconcile its processing outcomes, compare important fields against the source, check relationships by stable identifiers, and inspect representative records in context. Give every check an owner, an evidence location, a pass condition, and a decision for failure. Keep the batch out of dependent work until blocking mismatches are resolved or explicitly contained.

The CRM import validation checklist below is a recommended operating design, not a description of controls available in a particular product. Use the exports, logs, and inspection methods available in your workspace; check current plan availability before depending on a particular capability.

1. Define what acceptance means before checking records

Start with a batch boundary: the source files, their extraction time, the included record types, and the intended destination. Preserve an unchanged copy of those files. If the source continues changing, compare the destination with the frozen import inputs rather than with a live source that now contains newer edits.

Write down the intended treatment of each important field. A comparison cannot distinguish a defect from a planned transformation without that information. Define whether a blank means unknown, deliberately cleared, or leave the existing destination value unchanged. Define how dates, time zones, currencies, category values, and identifiers should be handled.

Also distinguish new records from updates and intentional exclusions. A source row that updates an existing record should not be counted as a newly created record. An intentionally excluded row should not disappear into an unexplained difference.

  • Critical fields: identifiers, relationship keys, ownership, access-related assignments, and communication restrictions where included.
  • Operational fields: status, stage, amount, next action, and other values that determine how the team handles a record.
  • Historical context: original dates, activity authors, note content, and links needed to interpret earlier work.
  • Presentation fields: formatting and labels that may matter without changing the underlying meaning.

Agree on these categories before inspecting results. Otherwise, acceptance can become a negotiation about how much inconvenience the team is willing to tolerate after discovering a defect.

2. Reconcile processing outcomes without confusing rows and records

Collect the import job summary, row-level results where available, rejected-row details, and destination identifiers. Tie them to the same batch reference. A screenshot of a completion message is supporting evidence, not a substitute for this reconciliation.

Use mutually exclusive outcomes for source rows: created, matched and updated, intentionally skipped, rejected, or unresolved. Every in-scope source row needs a disposition. If one row creates several destination objects, reconcile those objects separately rather than forcing them into a single record total.

Then test the other direction: every destination record attributed to the batch should map back to an expected source row or a documented transformation. This catches unintended duplication that a source-to-destination check alone can miss.

Pay particular attention to updates. The overall destination record count may barely change even when many rows were processed correctly—or when many updates went to the wrong records. Retain a source-to-destination identifier crosswalk, including the match rule used for updates.

If row-level results are unavailable, document the evidence gap. A destination export and identifier comparison may close it. If neither provides enough traceability, the batch should remain unverified rather than being accepted on totals alone.

3. Compare fields using explicit normalization rules

Compare source and destination values after applying only approved transformations. Preserve both raw values in the comparison evidence. Otherwise, an overly broad normalization rule can make a damaging change look acceptable.

Reasonable comparison rules may ignore agreed whitespace differences or compare timestamps as equivalent instants. They should not silently treat missing values as empty strings, remove meaningful leading zeros from identifiers, or assume all date-like text uses the same format.

For each mismatch, capture the source identifier, destination identifier, field name, expected value, actual value, transformation rule, and assigned severity. Keep sensitive values in a restricted evidence location rather than copying complete records into a broadly shared issue tracker.

Run full-batch comparisons for critical fields where the available exports allow it. Sampling is useful for contextual inspection, but it is a weak substitute for checking every relationship key or communication restriction. If a field cannot be exported, record which alternative inspection was used and what uncertainty remains.

Check distributions as a secondary diagnostic. Unexpected concentrations of blank owners or a status value that dominates the destination can reveal a mapping problem. Distribution checks help find anomalies; they do not prove that individual records are correct.

4. Filled hypothetical import-validation checklist

Fictional worked example: All quantities, results, folder names, and assignments in this section are hypothetical. A team imports 1,000 contact rows. The expected disposition is 900 creations, 80 updates, and 20 intentional skips. The migration lead maintains a restricted evidence folder called contact-batch-acceptance. The batch is not yet released for operational use.

The checklist uses three severities: blocker prevents release of the affected scope; major requires correction or a documented, contained exception; minor permits acceptance with an assigned cleanup task when meaning and use are unchanged.

  1. Row disposition reconciliation — passed.
    Check: account for every source row without overlapping outcome categories.
    Evidence: source-manifest.csv, job-results.csv, and row-reconciliation.csv.
    Owner: migration operator.
    Pass condition: 900 created, 80 updated, 20 intentionally skipped, with no rejected or unresolved rows.
    Observed result: the expected totals reconcile.
    Escalation: any unexplained row is a blocker; investigate before releasing the batch.
  2. Identity and update matching — passed.
    Check: verify the crosswalk and confirm that updated rows matched the intended existing records.
    Evidence: identifier-crosswalk.csv and restricted pre-import and post-import exports.
    Owner: data analyst.
    Pass condition: every imported source identifier has its intended destination mapping, with no unintended duplicate creations.
    Observed result: all 980 imported rows map as intended.
    Escalation: a wrong match or unexplained duplicate blocks the affected records and prompts a wider match-rule review.
  3. Operational field comparison — failed, major.
    Check: compare status and owner against the approved mapping.
    Evidence: field-differences.csv and mapping-register.csv.
    Owner: CRM administrator.
    Pass condition: all imported records have the expected status and owner.
    Observed result: 12 records have an unmapped status converted to a default.
    Escalation: hold those records, correct the mapping, and rerun the status comparison across the entire batch.
  4. Company relationships — failed, blocker.
    Check: compare each expected company key with the actual linked company identifier.
    Evidence: relationship-check.csv and company identifier crosswalk.
    Owner: data analyst.
    Pass condition: every required company relationship points to the intended company; intentionally unlinked contacts remain unlinked.
    Observed result: four contacts link to a similarly named but incorrect company.
    Escalation: block relationship-dependent use of the batch until the matching rule is understood and all company links are rechecked.
  5. Communication restrictions — passed.
    Check: compare imported opt-out and permission-related fields without inferring permission from missing data.
    Evidence: restricted communication-field-check.csv.
    Owner: communications operations lead.
    Pass condition: recorded restrictions are preserved, and unknown values remain unknown.
    Observed result: no differences found in the compared fields.
    Escalation: any weakened restriction blocks messaging use of the affected scope pending investigation. This field check alone does not authorise outreach.
  6. Historical context — failed, major.
    Check: compare original activity dates, authors, note text, and parent links for included history.
    Evidence: history-comparison.csv and record-inspection-notes.
    Owner: sales operations lead.
    Pass condition: included history retains its intended meaning, sequence, and parent record.
    Observed result: six activities show import dates instead of original activity dates.
    Escalation: hold historical use of affected records, repair date mapping, and compare all imported activity dates again.
  7. Record inspection — passed within stated coverage.
    Check: inspect 24 records covering creations, updates, blank fields, multiple relationships, long notes, and non-ASCII characters.
    Evidence: inspection-register.csv with record identifiers and observations.
    Owner: business process owner.
    Pass condition: visible values, linked records, and history agree with the source and mapping rules.
    Observed result: no additional defects found.
    Escalation: any new pattern expands inspection to the relevant population; this sample does not override failed full-batch checks.
  8. Acceptance decision — blocked.
    Check: review all defects, containment boundaries, and retest evidence.
    Evidence: defect-register.csv and acceptance-decision.
    Owner: migration lead, with business process owner approval.
    Pass condition: blockers closed and major issues corrected or explicitly accepted within a verifiable containment boundary.
    Observed result: release withheld because company relationships remain incorrect.
    Escalation: schedule repair and retest; do not describe the batch as accepted merely because processing completed.

5. Inspect relationships and history as connected information

A relationship is not correct just because its destination exists. Verify that it is the intended destination. Name-based matching deserves particular scrutiny when organisations, contacts, or opportunities can share similar names.

Test for missing required parents, unexpected parents, duplicate links, and incorrect relationship roles. Compare stable source keys through the identifier crosswalk rather than comparing display names alone. Check intentional exceptions too: an optional relationship left blank may be correct.

Historical context needs more than a note count. Inspect whether activities belong to the right contact or opportunity, retain their original sequence, and distinguish original event dates from import timestamps. If authorship cannot be represented directly, agree on an explicit preservation approach rather than silently attributing everything to the importing user.

Record inspection should answer a practical question: could someone opening this record understand the prior work correctly? Compare the source and destination side by side. Look for truncated notes, broken character encoding, misplaced attachments where included, and history that appears under the wrong parent.

6. Assign severity by consequence and containment

Do not classify defects solely by how many rows they affect. A single wrong relationship can matter more than widespread cosmetic spacing. Consider what the mismatch changes, whether dependent work can be prevented, and whether the cause could affect records not yet inspected.

  • Blocker: identity, relationship, access, or communication-restriction defects that make affected use unsafe or unreliable; unexplained loss; or insufficient evidence to establish correctness.
  • Major: incorrect operational or historical information that impairs work but has a demonstrable containment boundary.
  • Minor: presentation differences that do not alter meaning, matching, access, or operational decisions.

Partial acceptance is reasonable only when the clean population can be identified and separated from the defective population. A filter based on an unverified imported status is not reliable containment. Where separation cannot be demonstrated, hold the broader batch.

An exception needs an accountable approver, affected scope, reason, temporary handling rule, and correction owner. Acceptance should not mean that an unresolved difference has stopped being visible.

7. Repair narrowly, then retest beyond the changed records

Before rerunning an import, establish whether it will create, update, or duplicate records. Preserve current destination values, especially if people have edited records since import. A broad replay can overwrite valid work while fixing the original defect.

Choose between a targeted correction, a controlled reimport, or rollback based on the defect and available recovery methods. Do not assume rollback exists or restores everything. Verify its scope, including relationships and history, before relying on it.

Retest the corrected records, the rule that caused the defect, and adjacent fields or relationships that the repair might affect. Repeat batch reconciliation if records were created, deleted, merged, or rematched. Keep earlier evidence alongside the new results so the acceptance decision has a traceable history.

For the next imported batch, assign a validation owner before dependent work starts. Produce a row-disposition reconciliation, a critical-field difference file, a relationship check, and a documented inspection set. The measurable finish line is not a completion banner: it is no unresolved blockers, explicit decisions for remaining exceptions, and evidence that another reviewer can follow.

CRM migrationData validationImport acceptanceData quality

Put the next conversation to work.

Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.

Explore Communications