Alright, I live in dashboards and alerts, so when the sales team told us they were migrating from Pipedrive to Freshsales, my first thought wasn't about features—it was about data fidelity and tracking. How do we monitor a *process* migration in real-time? This is my six-month retrospective on the deal tracking side, the metrics that mattered, and the gaps we missed.
**The Good: Data Mapping & Initial Sync**
We used the Freshsales migration tool. The mapping was straightforward for standard fields (contact, company, deal value, stage). We validated the sync with a sample dashboard before the cutover. Key Prometheus metrics we tracked during the sync:
```yaml
# Simplified metric we exposed from our sync script
freshsales_migration_records_total{type="deals", status="synced"} 12345
freshsales_migration_records_total{type="deals", status="failed"} 7
freshsales_migration_validation_gap{type="deal_value"} 0
```
This gave us a real-time heartbeat. The 7 failures were due to custom fields in Pipedrive with unsupported characters.
**What Broke: The "Stage Duration" Blackout**
Our biggest oversight was historical *time-in-stage* data. Pipedrive tracked it internally, but the migration only moved the *current* stage, not the timestamp history. Overnight, we lost all ability to:
* Calculate average stage progression time
* Alert on deals stuck in a stage (a critical night shift alert for us)
* Report on historical trends pre-migration
We had to backfill from Pipedrive's API archives, which took two weeks of scripting. Wish we'd known to extract that dataset *before* signing off on the migration scope.
**Dashboard & Alerting Rebuild**
Our old Pipedrive Grafana dashboards became obsolete. In Freshsales, we leaned heavily on their API and built new boards focused on:
* Deal flow health (incoming vs. closing)
* Stage transition rates (post-migration baseline)
* Value leakage detection (deals moving backwards in stage)
The lesson? Treat a CRM migration like a major service deployment. Have validation dashboards ready for the new system *before* cutover, and define your key SLOs (like data completeness) upfront. The sales team loved the new UI, but we spent nights rebuilding the observability layer they never see—until an alert fires.
zzz
Sleep is for the weak