// Recruitment Technology, Migrations
What to Ask For Before You Give Your CRM Notice
Published: 17 November 2026,
9 min to read
The bottom line
Loading data into a new platform is the easy half of a move. The hard half is extraction. It goes badly when an agency asks for its database after giving notice rather than before.
Put the export request in writing while you are still a paying customer. Name the objects and file formats you expect, and set a delivery date the vendor can be held to.
The exit clause nobody reads until the week they need it
Most migration plans are built from the wrong end. They start with the new platform: the demo, the implementation call, the training schedule and the go-live date. The vendor being left behind gets treated as an administrative detail. Then notice is served, and a support desk that answered in twenty minutes for four years starts taking a week.
Published crm data migration best practices are written from the receiving side. They cover cleansing, mapping, validation and user training. Very little of that addresses the part you stop controlling once notice is served. That part is whether your data arrives at all, and in what shape.
More than a hundred agencies have described the same sequence to us on recorded calls. The request goes in after the decision is announced. Replies slow down. An invoice appears for something described as a data extraction service. What eventually lands is a contacts sheet and a companies sheet. Twelve years of context stays behind in the legacy system.
Why is getting data out harder than loading it into a new CRM system?
Because nobody at your current provider is measured on it. Import is a sales function, so it gets engineering time, documentation, roadmap space and a named team. Export sits with retention, and retention is not rewarded for making departures smooth.
The result shows up in what arrives. A database backup in a proprietary format, readable only with a license you are about to cancel. A CSV of contacts with notes truncated at 255 characters. Attachments delivered as forty thousand files named by hash, with nothing tying them to their records. A backup nobody can open is a receipt for a database.
Fragmentation compounds the problem. 56.16% of agency recruiters describe their tech stack as functional but fragmented. The full record of a client relationship rarely sits in one system. Some of it is in the CRM. The rest sits in a shared mailbox or a spreadsheet on one consultant’s desktop, and none of that appears in a standard export.
Exit friction is starting to attract regulation. The EU’s Data Act became applicable on 12 September 2025. It sets a framework for customers switching between providers of data processing services. Enforcement will take years to reach recruitment software. Until then, the contract you already signed is the thing that protects you.
What belongs in a complete CRM data export, and what usually goes missing?
A complete export is the database itself. Your incoming provider should be able to reconstruct which candidate went forward for which role, and when. Anything less is a summary.
Ask for all of it in a single request:
Attachments are where most handovers quietly fail. CVs, signed terms, contracts and right to work documents sit apart from the records that reference them. The bulk file dump then turns up with no manifest. Without that mapping, someone spends two weeks opening files to identify each one.
Custom fields are the second gap. Years of desk convention live in picklists, status labels, tags and free text fields your consultants invented. A generic export template drops them silently. That history also has a shelf life, and data decay speeds up once records land somewhere they cannot be searched.
- People and company records with every custom field, including the ones your team built and the labels only they understand
- The relationships between records: which contact sits at which company, which candidate went forward for which role, which consultant owns the account
- Activity history, meaning call notes, logged emails, interview feedback, submission dates, placements and fall-offs
- Every attached file with a manifest mapping it back to the record ID it belongs to
- Job and placement data including rates, fees, margins, rebate periods and start dates
- A data dictionary listing field names, data types, picklist values and what each status code actually means
How do you word the request while you are still a paying customer?
Send it in writing before you give notice. Name the formats, and attach a date. A request made during a live term reads as a service question. Sent during a notice period, the same request reads as an inconvenience.
Five things to confirm in that email:
Your contract usually backs you up. Under GDPR, a processor must return all personal data to the controller after the end of the provision of services. Your candidate database is personal data. Your agency is the controller, so the provider holding it has to give it back. Quoting the clause changes the tone of the reply.
Then test the answer. Request a sample export of a few hundred records while post migration support is still contractually yours. Open it. Check that notes survived and that dates parse as dates. Confirm that attachments arrived, and that a candidate still links to a submission. An empty field found now costs five minutes. The same field found in week three of a migration project costs a rebuild.
- The complete list of objects and tables available for export, and every field within each one
- The file formats offered, and whether attachments come with a manifest that maps each file to a record ID
- The cost, if there is one, and the clause in the signed agreement that authorizes it
- The number of working days from written request to delivery, and how long the download link stays live
- The retention and deletion timetable after termination, and whether a second copy can be requested
Which CRM data migration best practices matter once the export lands?
Counting comes first. Reconcile record counts in the export against reports from your current CRM system. Cover people, companies, jobs, placements, notes and files. A gap spotted in week one is recoverable while your old provider still has to answer the phone.
Mapping comes next, and it belongs to someone who runs a desk. Status labels, engagement types, ownership rules and custom attributes need a human decision before they move. A project administrator cannot know that a status called Hold means something specific on the contract team. Working through an ATS migration checklist keeps the sequence honest. Most of the costly migration mistakes agencies report surface after the old contract has lapsed.
Handled properly, the receiving side carries most of the weight. That is how migrations run at Atlas, an AI-powered recruitment platform that removes admin with agentic AI. Your database loads into a live, queryable environment at no cost once you have settled with your provider. Engineers write transformation scripts for your specific dataset. Then 300 or more automated checks run across a full dataset before anything reaches production. Nothing goes live until you sign off staging.
Day one is the real test of data integrity. Records that arrive indexed and searchable let consultants work the database that afternoon. That is the entire point of carrying history across. Call notes and logged emails stay attached to the person they belong to inside an AI-powered CRM. A searchable recruitment database turns twelve years of history back into placements.
Frequently asked questions (FAQs) on CRM data exports and migration
Before you give notice, while the contract is live and support is still responsive. A written request during an active term is treated as a service question rather than an exit task. Ask for a small sample export first. That confirms the format and the field coverage before the commercial relationship changes.
Some do, usually under a professional services or data extraction line rather than a published price list. Check what your signed agreement authorizes before you accept an invoice, and ask which clause the charge sits under. Loading your database at the receiving end should cost nothing. Atlas states that this step is free once you have settled up with your outgoing provider.
CSV or JSON per object covers most recruitment datasets, with a full SQL dump if the provider offers one. Attachments should arrive as files with a manifest that maps every document to a record ID. Request a data dictionary alongside the data, so field names and status codes are readable without guesswork.
Go back in writing, list the objects and fields missing against the provider’s own documentation, and give a delivery date. While the contract is live this is a service request rather than a favor. Keep the whole exchange in email, because a documented request supports any escalation later.
Most agency migrations run two to three weeks from receipt of the database to go live. Transformation typically takes a few days. The rest goes on field mapping, quality assurance, your review and final sign-off in staging. Dataset size and structural complexity move that timeline more than agency headcount does.
No. The work happens on a copy of your data in a separate environment. Your team keeps billing on the current system throughout. The cutover is usually scheduled outside working hours, so consultants return to a working database instead of a maintenance window.
The handover that decides how your next CRM performs
Most crm data migration best practices begin after the export lands. The ones that decide the outcome begin earlier, while you still hold a renewal the vendor wants. Name the objects, the formats, the attachments and the date. The rest of the move becomes a mapping exercise rather than a recovery operation.
What makes that history worth carrying is having one place for it to live. Candidate records, client activity, every conversation your team has had, documents and submissions sit together in Atlas. Nothing stays stranded in a mailbox or a folder of spreadsheets nobody has opened in years.
If your database currently lives in four places, it is worth seeing what one looks like.



