Philippines staffing guide · Published
Set boundaries for CRM data cleanup
A delegated cleanup workflow that fixes formatting and duplicates without letting an assistant overwrite ownership, consent, or commercial terms.
On this page
CRM cleanup is often handed over as a list of records to fix. The work looks clerical until a merge changes an owner, a consent flag, or a deal amount. A bounded cleanup lane separates safe normalization from decisions the business must make.
The assistant can standardize formats, flag duplicates, and prepare a merge proposal. The CRM owner keeps the decisions about identity, consent, ownership, and revenue.
Define the safe field set
List the fields the assistant may normalize: capitalization, spacing, country and state formats, phone formatting, job-title cleanup, and empty-field detection. Exclude fields that carry legal, consent, ownership, pricing, or forecasting meaning unless the company has explicitly approved a rule.
Record the field-level permission in the cleanup brief. A field that looks like text may still be a controlled picklist. Changing a picklist value can move a record into a different report, so the assistant should not add new values without approval.
Detect duplicates before merging anything
Build duplicate groups using the keys the company approves: email domain, normalized company name, phone, tax identifier, or a combination. Record which key produced each group and the confidence rule. A group is a proposal, not a merge.
Keep every record in the group visible with its source, last activity, owner, and open opportunities. The assistant should not choose the surviving record when the choice affects ownership or revenue. Prepare a neutral comparison and route the merge decision to the CRM owner.
Preserve history and attribution
Before any merge, export the fields and activity history that will be affected. Record the source record, the target record, the fields that will be kept, and the fields that will be discarded. If the CRM cannot preserve the history, stop and ask the owner whether the merge should proceed.
Never rewrite the created date, original source, or campaign attribution to make a record look cleaner. Those fields may be used for commission, reporting, or consent history. A cleanup that improves formatting but destroys attribution creates a different problem than the one it solved.
Keep consent and communication state untouched
Consent, opt-out, and communication preferences are not formatting fields. Do not infer consent from a previous purchase or from an email reply. If a record has conflicting consent values, preserve both and route the conflict to the owner who can resolve it under the company's policy.
Suppression and do-not-contact flags need the same care. A cleanup that removes a suppression because the record looks stale can create a legal and reputational risk. Mark the conflict and stop rather than choosing the value that makes the list larger.
Test on a protected sample
Clean a small sample in a sandbox or with a reversible change set. Include a record with an open deal, a record with a recent complaint, a duplicate with different owners, and a record with a missing consent value. Predict the result before running the change.
Record the test records, the rule version, the result, and any unintended change. If the test shows that a merge would alter an owner or an open opportunity, revise the rule and rerun. Do not expand to the full file until the sample passes and the owner accepts the approach.
Close with a reconciliation and a rollback note
After the approved changes, reconcile record counts, owner assignments, open opportunities, consent values, and suppression flags against the pre-change export. Report what changed and what did not. A count of cleaned records is not enough if the reconciliation misses a changed owner.
Keep the pre-change export and a rollback note for the period the company requires. Record who approved the change, the rule version, the date, and the sample evidence. If the CRM cannot support rollback, the owner should decide whether the change is worth the risk before it runs.
Questions and operating boundary
Can an assistant merge duplicate CRM records?
Only if the company has approved the merge rule and the merge does not change ownership, consent, or revenue fields without an owner decision.
Which fields are unsafe to normalize?
Consent, opt-out, suppression, owner, stage, amount, close date, and any field that drives reporting or legal obligations.
What if two records have different consent values?
Preserve both values, mark the conflict, and route it to the owner. Do not infer consent from activity or choose the value that allows more contact.
What evidence should the cleanup leave?
The pre-change export, rule version, sample results, approvals, reconciliation counts, and a rollback note.
Source
nist.gov provides authoritative background and does not endorse this service.