What counts as a duplicate contact?
A duplicate depends on the entity and purpose you are counting. Two rows may describe one person, one company, or different people sharing a switchboard. Define the matching rule before deleting anything. Preserve source and ownership evidence so deduplication does not remove a useful relationship or revive a suppressed record.
Microsoft’s duplicate-detection documentation describes field-based rules that identify potential matches, with limitations. A detected match is a review signal, not universal entity resolution. Exact email matches, domain matches and similar names answer different questions and should not be treated as interchangeable.
Choose the grain of the campaign: one person, one buying role or one account. Keep stable internal identifiers where available, normalize carefully and route ambiguous matches for review. Preserve a merge log so the team can explain which record survived and why.
Sources: Microsoft: detect duplicate records
How do duplicates change usable-record economics?
Apply exclusions in a fixed sequence and count each removed row once. Then divide total relevant cost by the unique records that pass the remaining campaign checks. Keep duplicate removal separate from audience mismatch, suppression and missing evidence, because each failure has a different remedy and a different owner.
Hypothetical file: 1,000 acquired rows at $2 each cost $2,000. Remove 150 duplicate rows, then 100 additional out-of-brief rows, then 50 additional records held by your internal exclusion rules. The remaining 700 are candidates for the next review gate. The invoice cost is $2,000 / 700 = $2.86 per remaining row.
If resolving duplicates takes four staff hours at $60, total cost becomes $2,240 / 700 = $3.20. These figures describe an invented workflow, not vendor quality or authorization to send. The remaining rows still need whatever evidence and use checks your actual campaign requires.
| Stage | Rows removed | Remaining |
|---|---|---|
| Acquired | — | 1,000 |
| Duplicates | 150 | 850 |
| Outside brief | 100 | 750 |
| Internal exclusions | 50 | 700 |
Sources: Microsoft: detect duplicate records
How can a research brief prevent duplicate rework?
Specify the entity grain, match keys and exclusion handling before research begins. Keep sensitive suppression information under the appropriate access controls and apply it before any authorized use. A new file should not silently overwrite existing ownership, evidence or opt-out status merely because its import timestamp is more recent.
Define whether multiple roles at one account are desired coverage or unwanted duplication. Provide permitted exclusion instructions through an appropriate controlled process, and keep the reason for exclusion out of unnecessary downstream exports. Do not treat an export of opt-out addresses as a general prospecting input.
For delivery review, save the incoming file unchanged, produce a reviewed copy and reconcile both counts. HEED’s order terms allow target criteria and exclusions to define the research request. A precise brief reduces avoidable ambiguity, while your CRM and sending controls still govern subsequent use.
- State whether the unit is a person, role or company.
- Preserve a reversible merge log and source record identifiers.
- Reapply current exclusions before authorized outreach, not only at purchase.
Sources: HEED service and record acceptance terms
Define a B2B research briefSpecify the audience, required fields and exclusions before choosing a batch.
Common questions
Are two contacts at one company necessarily duplicates?
No. They may represent distinct people or required buying roles. The campaign’s entity grain and coverage objective determine whether both belong.
Does a recent import replace an older opt-out?
No. An import timestamp does not establish a new permission state. Preserve applicable exclusions and review the proposed use through the appropriate controls.