Short answer: migrate when the tool itself is the problem (price at scale, a missing feature, a product that was shut down like June.so). Stack when the problem is that nobody turns the reports into product changes: keep the collector, add something that reads it. Either way, the checklist below keeps you from losing history or counting events twice.
Three reasons to migrate, and one that isn't
Teams switch product analytics for three legitimate reasons. The bill grew with volume and a competitor's free tier now covers the same usage (Mixpanel and PostHog both give 1 million events a month free; Amplitude 2 million since its July 2026 move to event-based pricing). A feature is missing, usually session replay, feature flags or in-product guides. Or the vendor disappeared: June.so, popular with B2B SaaS teams, wound down in August 2025 and pointed customers to Amplitude, and Heap is now part of Contentsquare.
The reason that doesn't hold is "we don't get insights from it". A new charting tool with the same events and the same reader produces the same silence. If the pain is deciding what to build next, the fix is a layer that reads the data and proposes the change, not a new place to store events. That is what Holdy does on top of PostHog or Mixpanel, with no migration.
Migrate or stack?
| Your situation | Move | Why |
|---|---|---|
| The bill is the problem | Migrate to PostHog, Mixpanel or Amplitude at your real volume | All three have free tiers; compare at your event count, not at list price |
| You need replay or flags | Migrate to PostHog, or add it alongside | Both ship on its free tier |
| You need in-product guides | Add Userpilot, Appcues or Pendo | Analytics tools don't build guides |
| Your vendor shut down | Export now, then migrate | June.so gave 30 days to export; don't count on more |
| The data is there, the decisions aren't | Stack: keep the collector, add Holdy | Holdy pulls the history from PostHog or Mixpanel and opens the opportunity on its own |
The 7-step migration checklist
- Export the history first. Raw events with timestamps, user ids and properties, not just the saved reports. If the vendor is shutting down, do this before anything else.
- Write the tracking plan down. List the five to ten events that mark progress in your product (created the first item, invited a teammate, finished the core flow), the properties each carries, and which one identifies the customer account. Most migrations fail here: the old tool had events nobody remembers the meaning of.
- Keep the account id stable. In B2B SaaS the unit that pays is the company. Make sure every event in the new tool carries the same account identifier as the old one, or you lose the ability to say which customers activated.
- Run both for one cycle. Send events to old and new in parallel for at least one activation window (7 to 30 days) and compare counts on the same day. A gap above a few percent means a missing event or a double-counted one.
- Rebuild only the reports people open. Check which saved reports were viewed in the last 90 days. The rest were already dead.
- Cut over on a quiet day, and keep the old tool read-only until the first monthly review passes on the new one.
- Decide who reads it, weekly, by name. If no one can be named, that is the signal to stack a reader on top rather than migrate again in a year.
How stacking works with Holdy
Holdy connects to PostHog or Mixpanel, pulls the event history, and starts reading the same day. From then on it watches the behavior of every customer account, composes events into named behaviors (activated, adopted the core feature, went quiet), and when something changes it opens an opportunity with the evidence, the affected accounts and a written plan. The change ships as an in-product hint through Holdy's SDK, a message, or a draft pull request on GitHub, with a holdout group, and Holdy measures the result in behavior and in the MRR that converted or churned.
Once Holdy's SDK is installed, new events come from the SDK and nothing is counted twice; the connector is only the way to bring the past in. The analytics tool stays for the team's own questions, replay and flags. Holdy starts on a Free plan with no card, up to 20 active customer accounts, 5 questions and 3 AI-generated improvements a month; paid plans are Launch at US$29 (BRL 149) a month up to 100 customer companies, Growth at US$99 (BRL 490) up to 1,000, never per seat or event.
What you lose and what you don't
In a true migration you lose saved reports, dashboards and any cohort definitions, and you keep the raw events only if you exported them. Retention curves restart unless the new tool imports history with original timestamps. When stacking, you lose nothing: the collector keeps everything and the reader works from it.
Thinking about switching because nobody reads the data?
Holdy finds what blocks activation and conversion, ships the improvement ready for review, and measures the impact on retention and revenue against a holdout group. Start on the Free plan, no card.