Why CRM migrations lose history
Most migrations are judged on one number: did every record arrive? It is the least informative check there is. An import can deliver every contact, company and deal and still leave the new system unable to answer the questions the old one answered, because the history was carried in fields the import did not map, in dates the import reset and in relationships the import did not rebuild.
The losses follow a pattern. Fields that should be written once, such as original source and the date a lead was created, are written again by the import, so every record appears to have arrived on migration day from an "import" source. Consent and suppression flags are mapped loosely or not at all, so people who opted out become reachable again. Stage names change without their definitions, and the stage conversion reports built on cohorts break, because the dates that placed each record in a cohort are gone. Owners who left years ago have no user in the new system and their records land with whoever ran the import. And the click identifiers and contact details that the CRM integration guide relies on are dropped as clutter, so the outcome uploads to the ad platforms stop matching without any error.
| What gets lost | How it shows up later | The step that protects it |
|---|---|---|
| Original source and created date | Reports credit the migration with every lead; cohorts collapse into one month | Map fields; carry source intact |
| Consent per channel and suppression entries | People who opted out are emailed or texted again | Carry consent and suppression intact |
| Stage definitions and stage dates | Conversion rates change overnight with no change in the business | Map pipeline and lifecycle stages |
| Owners | Records owned by former staff or by the person who ran the import | Map fields, picklist values and owners |
| Duplicates merged under pressure | Activity and deals split across records, or the wrong value surviving a merge | Clean and deduplicate in the old system |
| Workflows and their enrolment rules | The load enrols every record into a welcome or nurture sequence | Rebuild workflows paused |
| Click identifiers and the contact details captured at the form | Offline conversion uploads stop matching; bidding learns from nothing | Reconnect the integrations |
None of these failures raises an error. Each is found weeks later by someone reading a report, which is why the steps below check the meaning of the data, not only its volume.
Before you start
Gather five things. Administrator access to the old system, including the right to export every object, field and history table, not only the records a normal user can see. The automation inventory described in the automation governance guide, or the time to build one, because every workflow will need a decision. The written stage definitions from the pipeline, and if they do not exist, a working session with sales to write them before anything is mapped. A register of every integration that reads from or writes to the CRM: forms, the email or automation platform, ad platform lead forms, outcome uploads, calendar and calling tools, the warehouse; the data flow and integrations guide explains the patterns each one is likely to follow. And the retention decisions of whoever owns data protection, because the migration will ask what happens to old and inactive records and marketing should not answer that alone.
Agree two things in advance. What counts as a successful load, written as a set of reports that must give the same answers in both systems: leads by original source by month, records by lifecycle stage, open deals by stage and owner, and the count of suppressed people. And who decides on the day of cutover. A test load can be deleted and repeated as often as needed; the cutover, after which people start working in the new system and new history accumulates there, is hard to reverse. Treat it as the kind of decision the one-way and two-way doors framework asks to be made slowly and by a named person, and treat everything before it as a door that can be walked back through.
The steps
Inventory objects, fields, workflows, integrations and reports
List every object the old system holds (contacts, companies, deals, activities, campaigns, tasks, lists and any custom objects), every field on each with its type and permitted values, every active workflow, every integration and every report that someone actually opens. Mark who uses each item. The inventory is the input to every later step, and it usually shows that a large share of fields and workflows have no current user, which makes the next step easier.
Decide what moves, what is archived and what stays behind
For each object and field, decide whether it moves, is archived outside the new system or is left behind. The principles the GDPR glossary entry sets out, data minimisation, accuracy and storage limitation, are a reason to leave behind Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier. that no longer has a purpose, and a migration is the natural moment to apply them. They are not a licence to delete what the retention rules say to keep, so record each decision with the data protection owner's agreement. Two categories always move: suppression entries and the consent recordsThe stored evidence of a marketing permission: the channel, the wording the person saw, the date and the source of each opt-in, kept with the contact record. behind active contacts, for the reasons in the step on carrying them intact.
Map fields, picklist values and owners
Build a field map with one row per old field: the new field it goes to, the transformation if the type changes, and a value-by-value map for every picklistA CRM field type that accepts one value from a defined list of permitted values, so that records can be reported on instead of typed in many spellings., because a source value spelled one way in the old system and another in the new will split every report. Fields written by integrations keep a name that says which system writes them, as the fundamentals guide recommends. Build an owner map too: each old user to a new user, and for former staff a rule (the account owner, the territory owner) rather than a default, so records do not pile up with the person running the import.
Map pipeline and lifecycle stages
Map each old stage to a new one using the written definitions, not the names. Where two old stages merge or one splits, write down how each record is placed. Carry the dates that matter for reporting: when the record was created, when it entered each Lifecycle StageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer., when each deal changed stage and when it closed. If the new system will not accept historical values on its own system fields, carry them in clearly named fields of their own, so cohort reports can still be built on the original dates.
Clean and deduplicate in the old system first
Merge duplicates, fill or retire blank required fields and fix picklist strays while the data is still in the system that holds its history. Use the merge rules from the fundamentals guide, decided before the first merge: the older record's original source wins, the most recent consent wins, activities and deals are combined. DeduplicationRemoving double counts when one conversion is reported by more than one source. done after the move works on records that have already lost their activity links, and a merge is usually irreversible, so this is a step to do carefully and once.
Carry source, consent and suppression records intact
Original source and the campaign parameters behind it move as written and are never set by the import. Consent moves as a record, not a yes or no: the channel, the source, the date and the wording, as the list building and consent guide describes. That guide notes that exports from an old platform are usually fine if the consent records travel with them; without them, the people are unknowns. Load Suppression ListA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. entries before contacts, so that nobody suppressed in the old system is created as reachable in the new one, and keep them as suppressions rather than deleting them, as the suppression list entry explains. Click identifiers and the email or phone captured at the form move unchanged, because the outcome uploads match on them.
Load a test sample and reconcile it
Choose a sample that covers every object, every stage, every source, suppressed and consented contacts, records owned by former staff and records with long histories. Load it into the new system, then compare it with the old system record by record and field by field, and run the agreed reports on both. Fix the map and load again until they agree. Delete each test load before the next one so leftovers never mix with the real data.
Rebuild workflows paused and integrations disconnected
Rebuild only the workflows the inventory says are still needed, against the new fields and stages, and keep every one paused. Leave forms and integrations pointing at the old system for now. A load into a system with live workflows enrols records as if they had just arrived, which is the enrolment problem the workflows hub warns about at activation; decide for each workflow whether existing records should ever enter it.
Cut over with a freeze and a final load of changes
Announce a freeze on the old system, take the full load, then load the changes made since the last export. Point forms, lead form integrations and the automation platform at the new system, send a test lead from every form and source, and walk it through the lead routing QA checklist. Switch workflows on one at a time, watching enrolment counts. The old system becomes read-only from this point.
Reconnect ad platform and analytics integrations
The CRM integration guide describes the outcome uploads: Google documents uploading CRM outcomes keyed to the click identifier stored on the lead, or to the email or phone captured at the form, and Meta documents CRM outcomes as server events through the Conversions API. Reconnect each offline conversion uploadUploading outcomes recorded outside the website, such as closed deals, to an ad platform. from the new system with the same conversion names, check that the first uploads match, and confirm that the consent state is still checked before anything is sent. Record the switch in the measurement change log so a dip in reported conversions is read as what it is.
Switch the old CRM off after a parallel check
Run the agreed reports in both systems over the same periods until they agree, and investigate every difference. Then take the final archive the retention decision calls for, remove users and integration credentials from the old system, and close it. Update the inventory and the field map so they describe the new system rather than the project.
A catering supplier moves its CRM (illustrative)
Vantrelle Catering Supplies moves its CRM
Vantrelle Catering Supplies, an invented supplier of commercial kitchen equipment, moves from a CRM it has used for eleven years. The inventory finds 18,600 contacts, 4,300 companies, 2,900 deals, 214 contact fields and 47 active workflows. With the data protection owner, the team leaves behind 3,100 contacts with no activity and no open deal in the retention period, archives closed deals older than the retention rule allows in the CRM, and keeps all 1,240 suppression entries. The field audit cuts the contact fields that move to 61, and 19 of the workflows are rebuilt. The first test load of 400 records fails the reconciliation twice. Original source on every record reads "import", because the new system sets the field on creation; the team carries it in a field of its own and rebuilds the source report on that. Email consent has been mapped to the SMS consent field, which would have made 2,700 people textable who had only agreed to email. Both are fixed before cutover. On the first day after cutover the offline conversion upload is reconnected with the same conversion names, and the first batch of 23 qualified leads matches. The old system stays read-only until two months of reports agree.
The order matters more than any tool. Moving the email programme at the same time adds sending reputation and authentication to the list of risks, and the tutorial on migrating to a new email platform covers that side; bringing any new tool into the stack, including the new CRM itself, follows the approval, access and training steps in the marketing tool onboarding SOP. The naming conventions guide is worth applying to the new field and picklist names before they are created rather than after, and the CRM hub shows where the records described here sit in the rest of the section.
Common mistakes
- Judging the load on record counts. Every record can arrive and the history still be gone. Reconcile reports, not totals.
- Letting the import write original source and created dates. Every lead then comes from "import" on migration day. Carry the originals in fields the import cannot overwrite.
- Mapping consent as a single yes or no. Consent is per channel, with a source, a date and the wording seen. Map it field by field and check a sample by hand.
- Loading contacts before suppressions. People who opted out become reachable in the gap. Suppressions go first and stay suppressions.
- Cleaning after the move. Merges made without the old history are guesses, and duplicates start creating new history the moment they land.
- Importing into live workflows. The load enrols everyone. Keep workflows paused until the data is in and reconciled.
- Dropping click identifiers as clutter. The ad platform uploads stop matching and bidding learns from nothing.
- Keeping the old system running indefinitely. A second copy of personal data with no owner is a risk, not a backup. Archive what the rules require and close it.
