A CRM export with broken accents and wrong emails

Goal Take over the old CRM's contacts with a repaired name, an international phone and a valid email, and name those that cannot be reached.

The sales team leaves an old CRM. Its contact export mixes:

capitals and broken accents (BÃ©nÃ©dicte);
two phone columns, one often empty;
mistyped emails.

The manifest turns it into a usable list.

lumnik Open
Sales
CRM
Connector
Entity

contacts.csv

CUST_ID;CUST_NM;CONTACT_NM;EMAIL;TEL_MOB;TEL_FIXE
C001;DUPONT  SARL;BÃ©nÃ©dicte MARTIN; B.Martin@Dupont.FR ;;04 72 00 00 01
C002;ATELIER LEFÃˆVRE;JEAN-LUC PETIT;jl.petit@lefevre;06 12 34 56 78;
C003;BOULANGERIE MOREL;Sophie Morel;sophie.morel@morel-pain.fr;06 98 76 54 32;04 78 11 22 33
C004;GARAGE ROUX ET FILS;MARC ROUX;MARC.ROUX@GARAGE-ROUX.FR;;
C005;CAFÃ‰ DU PORT;AMÃ‰LIE FAURE;contact@cafeduport.fr;+32 470 12 34 56;
C006;NGUYEN TRAITEUR;Linh Nguyen;linh@nguyen-traiteur.fr;07 11 22 33 44;04 72 99 88 77
C007;PHARMACIE GARNIER;PAUL GARNIER;p.garnier@@pharmacie.fr;;04 74 10 20 30
C008;FLEURS & CO;CLAIRE DURAND;claire.durand@fleursetco.fr;06 55 44 33 22;

manifests/10-contacts.yaml

apiVersion: connectors.lumnik.io/v1kind: Connectormetadata:  name: scenario-crm-b  connector_type: csv-file  scopes: [sales-b]spec:  transport:    kind: local    path: /tmp/lumnik-scenario/crm-contacts-reachable/b/contacts.csv  parser:    delimiter: ";"  after_process: none  mapping:    on-error:      fields: { email: dlq }               # a contact with an unreadable email goes to the dead letters    row:      - map-columns:          rename: { CUST_ID: customer_id, CUST_NM: company, CONTACT_NM: contact, EMAIL: email,                    TEL_MOB: mobile, TEL_FIXE: landline }      - coalesce: { column: phone, from: [mobile, landline] }    # the mobile, else the landline    cell:      - fix-mojibake: { columns: [company, contact] }            # BÃ©nÃ©dicte -> Bénédicte, before the case      - titlecase: { columns: [company, contact] }               # DUPONT  SARL -> Dupont Sarl      - parse-email: { column: email, lower: true }      - parse-phone: { column: phone, default-region: FR }       # without it, 06 12 34 56 78 is read as a US number

manifests/20-contact.yaml

kind: Entityname: contact_blabel: Contact clientdescription: "A customer contact taken over from the old CRM: company, contact person, one email and one phone in international format."match: [customer_id]scopes: [sales-b]columns:  customer_id:    label: "Code client"  company:    label: "Société"  contact:    label: "Contact"  email:    label: "Email"  phone:    label: "Téléphone (international)"sources:  - table: t_scenario_crm_b    map:      customer_id: customer_id      company: company      contact: contact      email: email      phone: phone

$ lm apply -f 10-contacts.yaml
{"name":"scenario-crm-d","connectorId":"6969b750-4353-4bcc-afd4-5f4d1504bc13","endpointsApplied":1}

$ lm connector run scenario-crm-d
run scheduled: a0adfcb9-87fa-4ff8-99b2-a1a7385ac2a5

$ lm run get a0adfcb9-87fa-4ff8-99b2-a1a7385ac2a5
ID        CONNECTOR       ENDPOINT  STATUS     IN  OUT  SKIPPED  DEAD-LETTERS  STARTED           ENDED             EVALUATED
a0adfcb9  scenario-crm-d  default   Completed  8   6    2        2             2026-10-23 18:49  2026-10-23 18:49  no
⚠ 2 rows quarantined — see: lm dlq list --run a0adfcb9-87fa-4ff8-99b2-a1a7385ac2a5

$ lm entity apply -f 20-contact.yaml
Applied entity "contact_d" → connector."v_contact_d"

$ lm entity get contact_d
CUSTOMER_ID  COMPANY              CONTACT           EMAIL                        PHONE         _CONFLICTS
C001         Dupont Sarl          Bénédicte Martin  b.martin@dupont.fr           +33472000001  {}
C003         Boulangerie Morel    Sophie Morel      sophie.morel@morel-pain.fr   +33698765432  {}
C004         Garage Roux Et Fils  Marc Roux         marc.roux@garage-roux.fr                   {}
C005         Café Du Port         Amélie Faure      contact@cafeduport.fr        +32470123456  {}
C006         Nguyen Traiteur      Linh Nguyen       linh@nguyen-traiteur.fr      +33711223344  {}
C008         Fleurs & Co          Claire Durand     claire.durand@fleursetco.fr  +33655443322  {}

$ lm dlq list --run a0adfcb9-87fa-4ff8-99b2-a1a7385ac2a5
ID        CONNECTOR       RUN       REASON         ATTEMPTS  RESOLUTION  MESSAGE
550db09d  scenario-crm-d  a0adfcb9  MappingFailed  1                     email: parse-email could not read the value
c22f12e4  scenario-crm-d  a0adfcb9  MappingFailed  1                     email: parse-email could not read the value

$ lm dlq get 550db09d-b9bf-455b-a548-641e23d42a4e
ID:        550db09d-b9bf-455b-a548-641e23d42a4e
Connector: scenario-crm-d (6969b750-4353-4bcc-afd4-5f4d1504bc13)
Run:       a0adfcb9-87fa-4ff8-99b2-a1a7385ac2a5
Reason:    MappingFailed
Attempts:  1
Message:   email: parse-email could not read the value
Detail:    {"field":"email","message":"parse-email could not read the value"}
Payload:   {"email":"p.garnier@@pharmacie.fr","phone":"04 74 10 20 30","mobile":null,"company":"PHARMACIE GARNIER","contact":"PAUL GARNIER","landline":"04 74 10 20 30","customer_id":"C007"}
Resolution: open

Limits

titlecase writes Sarl and Et: it knows neither legal forms nor short words.
The dead letter keeps the row from before the cell stage, columns renamed.
Synthetic contacts.

Sales · CRM takeover

Can we reach these customers?

The old CRM has their contacts, damaged.

Broken accents, two phones, mistyped emails.

Synthetic contacts

The problem

The export is not usable as is.

BÃ©nÃ©dicte MARTIN.
The mobile here, the landline there.
p.garnier@@pharmacie.fr.

Nothing is changed

The export stays as is.

lumnik repairs the names.
It keeps one phone, in international format.
It validates the email.

What it took to write

- fix-mojibake:
    columns: [company, contact]

Broken accents, repaired.

The whole scenario fits in 2 declaration files, 51 lines.

The answers

6 contacts reachable, 2 named.

Bénédicte Martin, +33472000001.
Café Du Port, Belgian number kept.
2 unreadable emails as dead letters.

The limit

The casing stays mechanical.

titlecase writes Sarl and Et.
A contact without a phone stays without one.

Try it

The scenario replays end to end.

One connector, one entity: each step is published with its output.

This scenario runs on lumnik Open, free and open source.

Try lumnik Open →