Okay, hear me out. We just finished a 9-month migration from Mixpanel to a new CDP, and our CFO's favorite slide isn't about the new features—it's the one showing we're saving $12k/month by **not** migrating our 7 years of historical event data.
The new platform's storage and compute pricing for backfilling 4.2 billion events was astronomical. We ran the numbers every which way: rehydration costs, transformation jobs, and the ongoing query fees for analytics on that archive. The total 3-year TCO for moving the data was over $450k.
Here’s what we did instead:
* **Kept the old Mixpanel instance on a "legacy archive" plan.** It's now read-only for compliance and occasional historical trend analysis. Costs us ~$300/month.
* **Set a clear data boundary in our documentation.** Anything after the cutover date (Q2 2024) lives in the new CDP. Anything before lives in the archive.
* **Built a simple internal portal** that routes query requests based on the date range. For the handful of times we need a year-over-year view spanning the cutover, we run two queries and stitch the results.
The key was realizing that the "complete migration" dogma doesn't always align with the economics. The active data in our new CDP is faster and cheaper to work with, and we avoided a massive upfront spend.
Has anyone else taken a hybrid approach like this? I'd love to hear what cost trade-offs you analyzed.
— Jackson
Right-size everything
You're right about the cost. Most migration projects ignore the data gravity problem.
Just make sure that legacy archive plan has proper access controls and audit logging. An old, read-only system still holds PII and becomes a compliance blind spot if it's just "set and forget."
We had to do a similar split with a SIEM migration. The operational headache was training analysts to query two systems.
Trust but verify, then don't trust.