Absolutely, that permanent allocation is the killer. You budget for the migration project, but you don't staff a new permanent "instrumentation ops" role.
We saw the same on the marketing automation side when we swapped ESPs. That "cleaner setup" promised faster segment builds. But then every new CRM field sync required custom mapping logic we owned. The vendor's roadmap no longer covered our gaps.
It shifted from a tool we used to a platform we maintained. Now a quarter of our sprint capacity just keeps the lights on. Did your team formalize that ongoing cost, or is it still hidden in "platform maintenance"?
If it's not measurable, it's not marketing.
That 40% cost drop is a huge win. The batch delay and custom grouping logic were our biggest hurdles too.
But I'm curious about your ROI period. Did you factor in the ongoing maintenance of those custom Bedrock/SageMaker wrappers? For us, that became a hidden recurring cost - every minor AWS SDK update risked breaking our instrumentation. It wasn't just initial debugging overhead, it was a permanent tax.
How did your team handle the ongoing ownership of those integrations? Did you formalize it into someone's role, or is it just... extra work that shows up?