Move from LaunchDarkly to Sendora.
Move flags + segments + experiments from LaunchDarkly to Sendora Experiments.
Step-by-step
- 1
1. Export flags + segments + variations
LaunchDarkly Export API gives JSON of every flag + segment + targeting rule. Audit which flags are still active vs. dead — kill 30% before migrating.
Half a day - 2
2. Map segments → Sendora Audiences
LaunchDarkly Segments port to Sendora Customers audiences. Same nested AND/OR semantics. Real-time membership replaces LaunchDarkly's sync cadence.
Half a day - 3
3. Create flags in Sendora Experiments
Each LaunchDarkly flag becomes a Sendora flag with same key + same type (boolean / string / JSON / multivariate). Default + variation values ported.
30 min + bulk-import script - 4
4. Swap SDK calls
Replace `ldClient.variation("flag-key", user, default)` with `sendora.flags.evaluate("flag-key")`. Web 2.x + RN 0.18.7+ + iOS / Android equivalents.
1 day per surface - 5
5. Wire experiment readout
LaunchDarkly Experiments → Sendora Analytics-driven cohort analysis. Variant assignment becomes a first-class event. Statistical readout auto-cohorted by variant.
Half a day - 6
6. Migrate kill switches + audit
Critical kill switches port first. Audit log auto-captures every flip — actor + timestamp + previous value.
1 hour - 7
7. Side-by-side test + sunset
Run both SDKs in parallel for 1 week. Diff evaluation results to catch any miss. Cut over once parity confirmed.
1 week observation
Watch-outs
- LaunchDarkly Code References (auto-detect flag usage in source) doesn't port — review manually before sunset.
- Deep custom targeting rules (Big Segments, custom rules) may need rewriting.
- LaunchDarkly's edge eval latency is load-bearing for some real-time UX — verify Sendora latency before final cutover.
Experiments
Audience-targeted feature flags with percentage rollouts + kill-switch toggle + evaluation log. Honest about not being LaunchDarkly statistical engine.
LaunchDarkly is a flag store. Statsig is an experiment platform. Sendora's Experiments module is leaner: typed flags (`boolean / string / number / json`), rule-based rollouts (audienceId + percentage + value), evaluate + evaluate-all endpoints with the same shape as LaunchDarkly's SDK contract, instant toggle + audit log. Honest about scope: no first-class `multivariate` flag type (use `string` or `json`), no statistical experiment readout engine, no winner auto-promote to In-App / Push, no `experiment.assigned` event on the platform bus (evaluations write to `flag_evaluations` table, not `events`).
Read full Experiments pageStuck on the migration? We'll help.
Launch partners get white-glove migration help direct from the engineering team. Free plan covers real product use during the parallel-run period.