Clean records when values are wrong, inconsistent, or duplicated. Enrich records when useful information is missing. Re-research the account when you cannot establish whether the business belongs in your campaign at all.
Treat cleaning and enrichment as different jobs
Cleaning repairs the records you already hold; enrichment adds useful missing context. The work can overlap, but the acceptance criteria should stay separate. A file with more populated columns is not necessarily a more reliable file.
Salesforce’s cleanup material addresses errors, inconsistencies, duplicates, and old data. Its data-quality guidance distinguishes completeness from accuracy. 1, 2 Use those distinctions to describe the actual problem rather than asking a supplier to “improve the database.”
Diagnose the issue before choosing the service
Use the following table as a starting point for a small audit. These are proposed operational decisions; they are not rules for automatically changing or merging your CRM.
| What you find | Likely task | Acceptance question |
|---|---|---|
| Two records for the same confirmed company | Duplicate review and cleanup | Were identifiers and history preserved? |
| A relevant account without a current buying-role contact | Contact research or enrichment | Is the new role supported by a current source? |
| A job title known to be outdated | Correction and role re-check | Is the old value preserved in the change log? |
| Several spellings of one country | Standardization | Does the export follow the agreed format? |
| A company whose business model is unclear | Account research | Can it be included under the campaign brief? |
| A record on your exclusion list | Suppression handling | Is it excluded from the campaign export? |
An example: why “fill the blanks” can be the wrong first task
Imagine an agency has 800 old records. Some are duplicate accounts, some have no contact role, and others no longer fit the offer. This is a hypothetical example, not a client result or an estimate of typical database decay.
First identify exclusions and possible duplicates. Then decide which accounts still belong in the campaign. Only after that should the team add missing contact information to the retained records. Otherwise, it may spend research effort enriching records it will later remove. Keep the original file unchanged and compare every revision against it.
Run a small field-level audit
Agree the intended use of the database and mark essential fields. Inspect a varied sample, classifying each issue as missing, conflicting, outdated, duplicated, or outside scope. Keep the source and date for any new value.
Then scope the work by field. For example: standardize country names, review suspected duplicate accounts, check current buying roles, and add missing company websites. Assign an owner for conflicts. Do not instruct a tool to merge on a company name alone without reviewing the identity rules.
Measure what changed, not just how many cells were filled
Ask for before-and-after counts for each agreed task: candidate duplicates reviewed, missing fields researched, supported updates made, and records still unresolved. Distinguish “not found” from “not attempted” so another researcher can continue the work.
For an illustrative completeness measure, divide records containing the required field by records eligible for that field. State the denominator and the definition of complete. A populated field still needs an accuracy check; completeness on its own does not demonstrate that the value is right.
Agree what happens after the first cleanup
Choose review triggers based on how you will use the file: before a new campaign, after delivery failures, or when a contact reports a role change. There is no refresh interval established by this article that suits every database.
If the accounts are still right but the records are incomplete, discuss database enrichment. If the target market has changed, start with a new ICP brief. If you need a continuing supply of newly researched accounts, recurring research is a separate scope from maintaining the old file.
Sources & methodology
This guide combines the primary sources below with CienceLeads editorial recommendations. Examples are illustrative; no original study, client outcome, or search-volume estimate is claimed. Sources checked 2 November 2026.
- Salesforce Trailhead: Data Cleanup Fundamentals ↗
Supports the scope of cleanup: errors, inconsistency, duplication, and outdated data.
- Salesforce: Guide to Data Quality ↗
Supports the distinction between completeness, accuracy, and other data-quality dimensions. The workflow and numerical example are CienceLeads editorial guidance.
Decide which records belong in the campaign before paying to add more information to them.
Explore the relevant research service