A duplicate record was created
A second record appears when a person arrives by a route that could not be matched to the first — a new channel with no linked identity, a number stored without its normalised key, or an import that did not match on anything. Merging joins the records and their identities, so the check before merging is whether every identity really belongs to one human.
The three ways a second record appears#
- A new channel with no link. An identity is one address on one channel. Somebody who has emailed for months and then sends a WhatsApp message arrives as an unrecognised identity until the two are linked, and creating a record is the correct response to somebody unknown.
- A number that could not be matched. Matching a caller uses the normalised form of the number. A contact created without one cannot be found by the number it came from, so the next call creates another record. Historically most contacts created by lead capture on one host had no normalised key at all; capture now sets it from the start.
- An import that matched nothing. An upload or a list import adds what it cannot match. A spelling difference, a personal address instead of a work one, or a company name written differently is enough.
What a duplicate costs before it is merged#
Two records mean two histories. A next action, a summary or a brief reads one of them, so Connect can answer a question knowledgeably from the smaller half of what it knows. The commercial versions of this are worse: an outreach sequence can address somebody as new when they are already a customer, and a suppression recorded against one record does not obviously cover the other.
That last point deserves care. Blocking is not a column but a memory row tagged against the contact, which is what makes it hold across channels and across future conversations. Held against one of two duplicates, it holds completely for that one and not at all for the other.
What Connect completed#
Both records are real, both are correctly scoped to your workspace, and both carry the history that arrived through them. Nothing was lost when the second was created — the messages, calls and commitments that arrived on each are attached to the record that received them, which is why merging can reunite them rather than having to reconstruct them.
What Connect did not complete#
It did not merge them for you. Duplicate candidates are surfaced, and acting on them is a person's decision, because a wrong merge joins two people's histories and is far harder to undo than a missed one. It also did not retrospectively apply the first record's memory, suppression or standing instructions to the second.
What you can do#
Open both records and compare their identities before anything else.
Result If an address or number on one plainly belongs to a different human, stop: this is not a duplicate, and merging would create the wrong-recipient problem permanently.
Decide which record is the survivor on evidence, not on age.
Result Where a number resolves to two contacts the older one wins automatically, which is deterministic and is not always the right answer. Merging deliberately is how you override that.
Merge.
Result Identities and history come together, and future arrivals on any of those identities resolve to one person.
Re-check any block or suppression afterwards.
Result A directive held against one record needs to be true of the merged one; this is the step people skip.
What an administrator can do#
- Check that contacts created by lead capture carry the normalised phone key. Without it, the same number keeps producing new records however many times you merge.
- Review the duplicates view periodically rather than only when somebody notices. Duplicates are cheapest to merge before either side has accumulated much history.
- Before a bulk import, check what the import matches on. An import that matches on nothing will faithfully create a duplicate for every row it is given.
When to escalate#
Escalate when duplicates keep reappearing for the same person after a merge — that means the matching key is missing rather than the merge failing — and when two records genuinely belong to two people who share an address, since that cannot be resolved by merging at all. Bring both record identifiers and the channel the second arrived on.
Questions#
Can Connect merge duplicates automatically?
Candidates are surfaced; the merge is a person's decision. Joining two people by mistake is much more expensive than leaving two records for one person, so the asymmetry is deliberate.
Does merging lose anything?
History and identities are joined rather than discarded. What needs re-checking afterwards is the directives — a block, a suppression or a standing instruction held against one record should be reviewed against the merged person.
Why was a record created with no name?
Because what was offered was not a name. A form of address, a pronoun or a phrase like 'the caller' is refused at every door a name can enter through, which keeps the records honest at the cost of some thin ones.