Feature flags without buying a platform
Flag services are good products solving a problem most teams could solve with a database table and thirty lines of code.
Feature flags are genuinely valuable: they decouple deploy from release, make staged rollout possible, and give you an instant off switch that does not need a revert and a redeploy.
Whether you need a vendor for that depends on scale, and the honest answer for most teams is no.
What you actually need
- A table: key, enabled, optional percentage, optional allow-list
- A cached read, refreshed every few seconds
- A helper that resolves a flag for the current user
- Somewhere in the admin to toggle it
That is an afternoon. It covers kill switches, staged rollout and per-customer enablement, which is the overwhelming majority of real usage.
When a platform earns its money
- Non-engineers need to toggle flags many times a day
- You run real experiments and need statistics, not just on and off
- Audit trails are a compliance requirement
- Flags must resolve identically across many services and languages
Those are legitimate. Notice they are organisational as much as technical - which is the tell that the product is really about workflow, not flag evaluation.
The part everyone skips
Flags are debt with a timer. A flag that has been at 100% for six months is dead code with extra branches, and two interacting stale flags produce states nobody has tested.
Put an owner and an expiry on every flag at creation. Review them monthly. Deleting a flag should be part of finishing the feature, not a separate task nobody schedules.