Duplicates in the grid
Connect surfaces rows that look like the same thing — two people with one address between them, two companies with the same name — and leaves the judgement to you. Merging is performed by the relationship service, which keeps one record and moves the other's identities and history onto it. Nothing merges on its own, and every merge is recorded.
Why duplicates happen at all#
A workspace collects the same person from several directions: a mail thread, a phone call, a form, an imported spreadsheet, a prospect that later became a customer. Connect resolves an address on a channel to a person where the evidence is clear, and declines to guess where it is not. Declining leaves a second record, which is the honest outcome — a wrong merge is much harder to unpick than a duplicate.
Prospecting has the same problem in a sharper form, since a discovered organisation may already be a customer under a slightly different name. The reasoning there is set out in telling a new prospect from an existing customer.
What surfaces as a candidate#
- Two person records sharing an identity — the same email address or the same number resolving to both.
- Records whose names match closely enough to be worth a look, without being treated as proof.
- A company appearing twice under variants of one name.
- A prospect record and a relationship record that describe the same organisation.
A candidate is a question, not a finding. Two people at one company can genuinely share a switchboard number, and two companies can genuinely have similar names. The list exists so that somebody who knows the business can look, which is why it is surfaced rather than acted on.
What a merge does#
| Thing | What happens |
|---|---|
| The surviving record | Keeps its identifier, so everything already pointing at it still points at it |
| Identities on the merged record | Move across, so future messages on those channels resolve to the survivor |
| History and timeline | Come with the identities — one person, one history, which is the point of Customer 360 |
| Field values | Resolved during the merge; you choose where the two records disagree |
| The audit trail | Records the merge, what was merged, and who did it |
The merge runs through the relationship service, like every other change made from the grid. There is no direct-to-table variant that skips the identity handling, which is precisely the part that would be easiest to get wrong and hardest to notice.
Before you merge#
Open both records and read their histories, not just their names.
Result Two similar names with entirely unrelated conversations are two businesses, and the list has done its job by making you look.
Decide which record survives, and why.
Result Usually the one with the longer history and the identifiers other systems already know.
Resolve the fields that disagree.
Result The survivor ends up with the values you chose rather than whichever record happened to be newer.
Merge, and check the survivor afterwards.
Result Identities, history and the fields you chose are on one record, and the audit says what was joined.
What merging will not do#
- Cross a workspace boundary
- Two records in different workspaces are not duplicates. They are two businesses' data and there is no operation that joins them.
- Delete a conversation
- History moves; it is not discarded. A thread that existed still exists, under the surviving record.
- Clear a suppression
- A do-not-contact entry travels with the person. Merging is not a way to lose one.
- Happen in the background
- No scheduled task merges records. Every merge has a person behind it in the audit trail.
Questions#
Why does Connect not merge obvious duplicates automatically?
Because obvious is doing a lot of work in that sentence. An automatic merge that is wrong joins two customers' histories into one record and cannot be cleanly undone, while a duplicate left standing costs somebody a minute. The asymmetry decides it.
Can the Assistant merge records?
It can find candidates and lay the evidence out. The merge itself is a change with no clean reversal, which is the kind of decision kept on the person's side of the line — see what the Assistant may not do.
Does an import create duplicates?
It reports rows that match existing records rather than merging or overwriting them, so an import can add candidates to this list. That is deliberate: the import knows the rows look similar, and it does not know which of them are the same business.