By · Updated

AI-generated photograph: arranging paper wireframes to scope a web journey
Practical method

Choose a CRM around real workflows

A CRM should support relationships and processes the team understands. Before comparing features, describe how an enquiry arrives, who handles it and how follow-up ends. Assess data quality and the ability to change tools too.

01

Describe a relationship workflow

Choose a representative case: received enquiry, qualification, appointment, proposal and follow-up. At each step identify the owner, necessary information and next event.

Separate contact, organisation and opportunity. One contact can participate in several enquiries. Understandable fields avoid mixing identity, history and commercial status.

02

Clean data before importing

Inventory existing files and owners. Define genuinely useful fields, permitted values and duplicate-matching rules. Test a small copy before importing everything.

Do not automatically merge people solely by name. Retain decision records and discarded-data handling within authorised boundaries. Define who can correct a record.

03

Let users perform real tasks

Ask future users to perform the same scenario in competing solutions. Observe search, entry, reassignment and error recovery. Record observed difficulties rather than advertised feature counts.

Check role permissions, exports and essential connections. Identify functions requiring paid options or intervention. Demonstrations should distinguish supplied configuration from development still needed.

04

Plan adoption and exit

Prepare short entry rules, scenario-based training and a quality owner. A rich tool remains of little use if people interpret statuses differently or cannot find enquiries.

Test an export including useful relationships and history rather than only contacts. Document costs over the same period, volume limits and data handling on exit.

05

Exercise permissions across teams

Model two roles: one qualifies an enquiry and another prepares the response. Check what each can see, edit and forward. Distinguish internal comments from information intended for the contact.

Use separate test accounts and reassign the enquiry. Verify the intended loss of access for the outgoing owner and useful history for the new owner. An administrator-only demonstration cannot establish these boundaries.

06

Decide how similar contacts are handled

Identical names can describe different people, while one person may use several addresses. Define evidence permitting a merge and cases requiring human review. Preserve provenance rather than overwriting conflicting information.

Before a large import, test fictional namesakes, shared contact details and employer changes. Inspect linked enquiries and history after matching. Test separation if merges are reversible; otherwise require stronger approval beforehand.

A record to keep with the decision

Minimum evidence for a review
ItemEvidence
ScenarioSteps, owners and information
QualityEntry and duplicate rules
ExitReadable export with retained relationships

Frequently asked questions

Should all old contacts be imported?

Start with the need and quality of each source. History without context can hamper search and create duplicates. Have retention and use rules reviewed before importing.

Reference material

CNIL · Principes clés

The practical checklist is an editorial synthesis to adapt to your service. It does not constitute a certification or an audit result.

Continue with a related decision

Source note

Choose a CRM by testing the data journey

A CRM demonstration often shows the ideal customer record. Instead, test three real operations: create a contact, correct a value and export what is needed to change supplier. Identify collected data, purposes, permissions and processors. The CNIL explains that personal-data processing should be documented in an appropriate record.

Request a trial with a small fictional dataset and the teams who will use the tool. Check duplicates, search, histories, permissions and usable exports. Budget for migration, training, integrations and exit; a per-user price alone does not represent project cost.

Build a scenario with a duplicate contact, an address correction and an export request. Have future users perform these tasks without continuous help from the demonstrator. Record difficult steps and what disappears during merging: history, company relationships or attachments.

Open the export in an ordinary tool and check relationships between records, not just row counts. A contact list missing information needed for handover is different from a usable export. Clarify which manual steps remain necessary during migration.

Use
Three everyday tasks tested by the actual teams.
Data
Purposes, access, processors and export checked.
Cost
Subscription, integration, migration and exit estimated.

Does a CSV export prove a CRM is easy to leave?

Not alone: also check object relationships, attachments and custom fields needed for the transition.

Fictional example: a duplicate contact

A team creates two records for one person, merges them and exports the result. It checks that useful exchanges and the company relationship remain retrievable before considering the trial successful.

Try the reasoning

A CRM imports contacts but its export has never been tested. What trial should precede selection?

Reasoning: Import a small fictional dataset with relationships, notes and accented characters, then retrieve it. Check retained fields and relationships, not only file counts.

Primary source : CNIL — Le registre des activités de traitement