Operations
Marketing Workflow Automation Migration Mistakes to Avoid
Marketing workflow automation migration mistakes rarely come from bad data entry. They come from state that never travels: consent, IDs, scores and touch history.
What to take away
- The costliest migration failure is consent and suppression state left in the old system, because a wrong send is a legal event.
- Field mapping without a stable source ID produces duplicates that return after every import.
- A rebuilt lead score looks correct on day one and fails two quarters later at pipeline review.
- The failures that hide longest are scoring and reporting, not sending.
- Port consent, suppression, record IDs and outcome history before you move anyone's daily workflow.
The costly one
Teams plan a platform switch as a contact move, and the contact table is the easy part. What gets left behind is the state around it: consent timestamps, channel-level opt-outs, suppression lists and the record of who asked not to be contacted again.
In a HubSpot to Marketo move, the contact export carries names and emails cleanly. Subscription status, global unsubscribe settings and any suppression lists need their own export, and Marketo keeps unsubscribe and marketing suspended flags as lead fields.
What Migrates vs What Stays
Data class
- Contacts
- Name, email, company
- Consent
- Current status
- Suppression
- Global unsubscribe
- Outcomes
- Open and click counts
Usually migrated
- Contacts
- Legacy record ID
- Consent
- Timestamp and legal basis
- Suppression
- Per-channel opt-outs
- Outcomes
- Opportunity and revenue history
Often left behind
- Contacts
- Consent
- Suppression
- Outcomes
That gap carries legal weight in the United States. CAN-SPAM opt-out duties follow a contact into whatever system sends the mail, and California residents can hold CCPA deletion rights against data you still keep. The FTC's compliance guide on commercial email sets out the opt-out obligations a sender carries. Flags that do not travel cannot be honored.
Prevention is unglamorous. Export consent and suppression as their own records, with source, date and legal basis, and load them before the first contact import. A typical mid-size migration spends one to three weeks of hands-on data work on consent, suppression and record IDs alone.
Usually migrated
- Contacts
- Name, email, company
- Consent
- Current status
- Suppression
- Global unsubscribe
- Outcomes
- Open and click counts
Often left behind
- Contacts
- Legacy record ID
- Consent
- Timestamp and legal basis
- Suppression
- Per-channel opt-outs
- Outcomes
- Opportunity and revenue history
The ones that look fine at first
Three mistakes pass every launch review because the platform accepts the data without complaint.
Three Mistakes That Pass Review
- Carry legacy ID in a permanent custom field
- Compare step counts, wait times, exit rules
- Dedupe on email alone loses activity history
- Rebuilt workflows drift from old definitions
Look fine at first
- Contact imports without a stable source ID. Dedupe runs on email alone, merged records lose their activity history, and the same duplicates return after each import. Carry the legacy ID in a permanent custom field, not just during cutover. Documented data migration pitfalls explain how identity keys survive a move.
- Workflow rebuilds done by eye. A thirty-step nurture rebuilt over a week drifts: enrollment criteria shift and wait times round to whole days. Compare step counts, wait times and exit rules against the old definitions before go-live.
The ones that only show up later
These failures clear launch because the data they depend on arrives late, or never.
- Lead scoring refitted without outcome history. Situation: the old score blended engagement with closed-won patterns from years of opportunity data, and the new platform starts with none of it. Consequence: the score looks reasonable for a quarter, then sales stops trusting it, and the damage surfaces at the next pipeline review. Prevention: port opportunity history and refit before switching the score on. Durable identity models matter here, because a score attaches to a person, not a row.
- Attribution windows reset to platform defaults. The first quarter on the new system reports a cliff in channel performance that never happened, and year-over-year comparison breaks. Export touch history and rebuild the same window model before cutover.
- Sending reputation treated as automatic. Soft bounces and spam placement climb for two to six weeks on an unwarmed domain. Warm each new domain gradually and keep the old one live through the transition.
None of these three appears in a launch checklist. All three appear in a board deck.
What to check before cutover
Each failure is state that lives outside the contact record: consent, identity keys, scores, touch history, sending reputation. A migration plan built around moving objects rather than state will miss every one of them. That is why marketing data governance deserves a review before cutover, not an audit after it.
A checklist that survives production looks like this:
- Export contacts, consent, suppression, opportunity history and touch history as separate files, with a column map for each.
- Load consent and suppression before the first contact import, then reconcile counts against the old platform.
- Freeze scheduled sends, form submissions and list imports 48 hours before the export.
- Confirm the legacy record ID is populated on every imported contact.
- Check each rebuilt workflow against its old definition for step count, wait times and exit rules.
- Warm the new sending domain for two to six weeks while the old domain stays live.
- Run a test import of 200 to 500 records and reconcile it field by field.
Pre-Cutover Migration Checklist
- Export consent and suppression with timestamps
- Keep legacy record ID in a permanent field
- Port outcome history before rebuilding scores
- Warm each new sending domain two weeks
- Test on a subset of 500 contacts
Example of a quiet failure
An illustrative case: a mid-size B2B team moves 40,000 contacts over a weekend; the launch looks clean, sends go out, forms work, dashboards load.
The lead score was rebuilt from engagement data alone, because opportunity history sat in a warehouse nobody connected to the project. Scores look plausible for two quarters. Then a pipeline review shows marketing-sourced opportunities down sharply, and the cause traces to a score that stopped matching buying behavior in month two.
