Duplicate Detection
Before a lead is written, two normalised keys settle whether it is already here: the email, and failing that the phone. Exact once normalised, with no similarity score for anybody to argue about.
An incoming row that already exists
matched on email- Job titlereplacedOperations ManagerHead of Logistics
- Cityleft aloneOaklandempty
- Lead scorereplaced6274
- Assigned tonever touchedS. BhattM. Ortiz
- Tagsnever touchedenterprisewebinar
- Custom fieldsadded toterritoryfleet_size
Product figures from the platform’s own defaults - not customer averages
How it works.
Email first, because it identifies
An email match is taken over a phone match whenever both are available, on the grounds that two people genuinely share a switchboard number far more often than they share a mailbox. Names are never compared at all.
Normalising is what makes it exact
Addresses are trimmed and lowercased. Numbers are reduced to their digits and rebuilt behind a plus, and a leading double nought is read as the international prefix it is, so a number wrapped in brackets and dashes lands on the same key.
Checked per chunk, not per row
During an import the comparison runs once for a whole batch of five hundred rather than once per row. Within a run the row identifiers are fixed, so a crash and a retry pick up without doubling anything; a second upload of the same file is a fresh run and leans on the keys alone.
The original wins, not the newest
Where several records hold the key, the earliest created one is the match. An incoming row therefore joins the record that has been accumulating history, rather than the copy somebody made last week.
Two moments in every record.
Every record passes through the same seven. Duplicate Detection is the lit ones, and everything either side of it is a different page in this category.
- 01Capture
how it arrives, and from where
- 02Shape
the fields and stages it carries
- 03Find
getting back to the right ones
- 04Work
what happens against it day to day
- 05Advance
moving it towards a decision
- 06Protect
who may see it, and staying clean
- 07Report
what the floor is judged on
The specifics.
8 facts- Keys
- Normalised email, then normalised phone. Names and addresses are never read
- Email rule
- Trimmed and lowercased. Dots and plus-addressing are not folded together
- Phone rule
- Primary then mobile, reduced to digits behind a plus. Home and other ignored
- Method
- Exact once normalised. No similarity scoring and no confidence value
- Tie break
- The earliest created record is chosen when several carry the key
- Runs on
- Import, and any lead a call creates. Not scheduled, and not on demand
- Two runs at once
- The call path holds a transaction lock. Concurrent imports are best effort
- Normalised both ways
- The stored key and the incoming one are built by the identical routine, so a match cannot depend on which side was typed first
What else writes to this record.
A record is worked by more than one part of the system at once. These are the ones that put something on it, or read it back.
More in Leads & Sales CRM
12 capabilitiesThe system of record the phone writes to directly - no sync job.
Unlimited pipelines, each with its own stages, win probability, colour and temperature. Seven seeded, and every stage change writes itself onto the deal.
Create the stages your floor argues about, colour them, order them and rate them hot, warm or cold, while six fixed categories keep the reporting stable.
Ten sources seeded and sealed, your own added beside them, each with a colour and a live count, and one stamped across a whole import rather than per row.
Eleven field types, all of them on the picker, with options, defaults and a section of their own, on leads, deals, products and vendors. No cap on how many.
Sort the fields you add into named sections with their own colour, order and table visibility, then grant or withhold a whole section per permission profile.
Columns, sorts, filters and layout save together and hand to a CRM role, so a new starter opens the same board instead of rebuilding it every morning.
Eighteen type-aware operators, a trigram search across seven columns, and a straight account of where a custom-field filter is resolved and what it costs.
Every field change on a record with both values, beside the notes, mail, tasks and deal links against it, assembled as it happens rather than typed up later.
Calls, voicemails, emails and texts merged into one chronological history per lead, attached by the number they reached, with duration and the wrap-up code.
Tasks with assignees, reminders, comments and templates, raised by hand or by a sequence, and the only CRM object that leaves the app to find someone.
A role tree, sixteen org-level flags, per-entity CRUD and sharing rules across branches - so a BPO runs several clients on one tenant, safely.
Ten measures over six windows, with attainment read from won deals and logged work the moment you open the target rather than filed by hand or overnight.
The rest of the platform.
Five more categories, all on the same record and the same bill. Each card names three of its capabilities, so you can tell from here whether it is worth opening.
Leads & Sales CRM
See duplicate detection on your own floor.
Thirty minutes, your numbers and your data. We will set duplicate detection up live and you can decide from the thing itself rather than from this page.
14-day trial · no card · migration included