Zero duplicates

One customer, one record. At last.

Three reps, three entries for the same customer, and nobody knows which record is authoritative any more. Lea finds the duplicates and lets you decide, record by record.

Potential duplicates
High confidenceSame emailSame phone
Camille Ferrand
  • c.ferrand@atelier-nord.fr
  • 06 12 44 89 03
  • 4 deals · 12 activities
Keep this record
Camille FERRAND
  • c.ferrand@atelier-nord.fr
  • +33 6 12 44 89 03
  • 1 deal · 2 activities
MergeDismiss
Worth checkingSame name
Atelier Nord
  • atelier-nord.fr
  • Lille
  • SIREN 812 445 903
Keep this record
ATELIER NORD SAS
  • Lille
  • 3 contacts
MergeDismiss

You never see the duplicate. You do see the damage.

Two records for the same customer, and the history splits in half: half the exchanges on one side, half on the other. You chase someone who just signed. You send the same sequence twice. Revenue per account becomes wrong. Lea scans your address book on demand and shows you the suspect pairs, along with the reason they were matched.

How it works

You stay in control from start to finish. Lea suggests, you decide.

  1. You start the scan

    From the Duplicates tab of your address book, whenever you want. The scan reads your records without ever modifying them.

  2. Lea lays out the pairs

    Each pair comes with the reason it was matched: same email, same phone, same SIREN, same name.

  3. You pick the record to keep

    Both records sit side by side, with their deal and activity counts to help you decide.

  4. You merge or you dismiss

    Merging reattaches the entire history to the record you kept. A dismissed pair is never suggested to you again.

What Lea compares

Identifiers, not impressions.

Matching runs on normalised values, so a difference in case, accent or formatting no longer hides a duplicate.

  • Email, lowercased and trimmed.
  • Phone, reduced to international format: “06 12 44 89 03” and “+33 6 12 44 89 03” are recognised as one number.
  • LinkedIn profile, whatever the subdomain or tracking parameters in the URL.
  • First and last name, accents stripped and order irrelevant: “Jean Dupont” finds “DUPONT Jean”.
  • For organizations: SIREN (the French company registration number), website domain, phone, and legal name with the entity type removed.

Against false positives

Two colleagues are not a duplicate.

This is the classic deduplication trap: matching different people because they share an address or a switchboard. Lea guards against it explicitly.

  • Role addresses are excluded

    contact@, info@, support@, sales@ and around forty others are never used to match two records. Those are company inboxes, not people.

  • Shared contact details are ignored

    An email, phone number or domain carried by three or more records is treated as a switchboard or a shared inbox, and produces no pair at all.

  • No combinatorial explosion

    Five records for the same person give you four pairs to review, not ten. You review a group, not a cross product.

  • The scan cleans up after itself

    A pair that is no longer a duplicate, because you fixed a record in the meantime, disappears from the list on the next scan.

The merge

Nothing is lost, nothing is overwritten.

The record you keep retains every one of its values. The other record's values only fill in the fields you left empty. It all happens in a single transaction: either the merge completes entirely, or nothing moves.

  • Deals, activities, emails, notes, tasks, attachments, quotes and bookings are reattached to the record you kept.
  • Labels from both records are combined.
  • For organizations, the SIRENE legal block is carried over as a whole, never mixed between two legal entities.
  • An email opt-out on either record is preserved on the merged one.
  • If both records were enrolled in the same follow-up sequence, only one stays active: no double send.

What it does not do

Better to know before you rely on it.

  • No fuzzy matching: “Dupond” and “Dupont” are not flagged as duplicates. Lea compares values that are identical after normalisation, not values that look alike.
  • A merge cannot be undone. A summary is kept in the audit trail, but the absorbed record is deleted.
  • The scan is started manually. There is no scheduled scan and no automatic scan after an import.
  • While typing, the duplicate warning only covers the email address, only when creating a contact, and it never blocks saving.
  • Each scan handles up to 500 pairs per type. Beyond that, Lea tells you, and you run it again once you have worked through them.

Included from the first plan

The quality of your data is not a paid add-on. Detection and merging are available on every plan and to every role, where several CRMs on the market reserve duplicate management for their top tiers.

Start again from a clean address book.