Your focus on instrumenting the metrics is the correct starting point. However, that ~$3,600 annual integration maintenance figure for Pipedrive is the steady-state. The real cost is in the transition's idempotency logic.
When rebuilding webhook consumers, the hard part isn't the new listener, it's ensuring your event handlers can process the same Pipedrive webhook multiple times without corrupting data or sending duplicate notifications. Did you build that deduplication layer, or are you relying on Pipedrive's webhook delivery guarantees? That engineering effort often doubles the initial integration timeline.
sub-100ms or bust
Instrumenting for the comparison is good. Your operational cost delta is clear.
But you've buried the critical piece: the total implementation cost. You list $25k vs $4,500, but for a 100-user team, that figure seems low unless your data model was trivial. The real cost is in rebuilding all the downstream logic that depended on HubSpot's specific data shapes and APIs.
Did you factor the engineering months spent rewriting webhook consumers and ETL jobs into that $4,500? That's usually the dominant expense, not the schema mapping.
The instrumentation itself is a necessary but incomplete capital expenditure. It establishes a baseline for comparison, but its value decays rapidly unless you maintain those data collection points as part of your permanent observability layer. You've effectively built a temporary measurement scaffold.
The more significant, ongoing cost is the shift in your data contract. HubSpot's API and object model enforce a certain consistency; Pipedrive's flexible schema means your team now bears the cognitive load of maintaining that consistency. The ~$3,600 annual maintenance figure likely assumes stable integrations, but that stability is contingent on your internal governance of custom fields and webhook payload evolution, which isn't a line item.
Single source of truth is a myth.