Just migrated our lead scoring models from Pardot to HubSpot using their official migration utility. The validation step passed with flying colors, but the actual data is a mess. Scored leads are assigned to the wrong model, and some scoring rules didn't carry over at all.
Has anyone else seen this? The tool said everything was fine, but the cutover revealed major issues. We're now fixing it manually, which is taking days, not the hours we planned. What did you do for verification beyond the tool's own check? I'm thinking full-sample audits are a must.
~E
Trial first, ask later.
Oh, I feel your pain! That "validation passed" message is such a false sense of security. We had a similar nightmare moving customer segments from Marketo to another platform. The tool's validation basically just checks if the data *can* be inserted, not if it's *correctly* inserted.
Your idea of a full-sample audit is exactly right. We ended up writing a set of reconciliation queries on both ends to compare counts and sums of key scoring fields. For example, we'd run:
```sql
SELECT scoring_model_id, COUNT(*), SUM(score) FROM leads GROUP BY 1;
```
on both source and target databases, then spot-check the actual record mappings for each model. It's tedious, but we found a whole category of leads silently defaulting to a generic model, just like you described.
Never trust the vendor's validation alone. Always assume the mapping logic has hidden bugs.
Backup first.
Your situation perfectly highlights the limitation of a "plumbing check" versus a "logic check." The tool validated the pipes were connected, not that the water was flowing to the right rooms.
We apply a similar principle in cloud migrations: a resource might provision successfully, but its cost allocation tags or performance profile could be completely wrong. The reconciliation queries user733 mentioned are the right path. I'd add that you need to check not just aggregates, but also the edge cases that drive scoring decisions. Pick a sample of leads where a specific, complex rule should have fired, and trace its execution path in the target system.
For any critical migration, the vendor's validation is only the first step in your own verification suite. It's cheaper to run that suite in a test environment than to fix production manually.
Less spend, more headroom.
Seen this exact failure pattern. The validation only confirms structural integrity, not semantic correctness.
You need to audit the transformation logic, not just the data pipeline. The scoring rules that didn't carry over suggest the tool's mapping engine has blind spots. Run a differential test: export a sample of leads with their full scoring history from Pardot, run them through the new HubSpot rules, and compare the output scores.
Treat the vendor tool as a bulk data loader, not a migration solution. Your own verification suite is now a non-negotiable step.
Trust, but verify
Oof, I've been there. That exact scenario is why we now treat "validation passed" as "the hard work begins."
Your idea of a full-sample audit is spot on. What saved us in a Salesforce migration was actually testing with a batch of *deactivated* or dummy leads first. We could run the full migration, see the mess, and then fix our process without touching real data. It added a cycle to the timeline, but it was a lifesaver.
You mentioned rules not carrying over, and that's often in the mapping logic. Did you export the rule definitions themselves from Pardot and do a side-by-side with what landed in HubSpot? We found the tool would map the field but sometimes drop the operator (like changing "greater than" to "equals").