Why one platform
Most teams reach Kerf after running flags, tests and replays in separate tools. Here is how the options differ in practice.
Kerf and the alternatives
| Kerf | Single purpose flag tools | Standalone A/B tools | In house | |
|---|---|---|---|---|
| Feature flags | Built in | Core product | Limited | You build it |
| Experiments on flags | Same data model | Add on or manual | Separate SDK | You build it |
| Sequential testing and CUPED | Default | Rare | Varies | You build it |
| Guardrail metrics | Automatic halts | Manual | Reporting only | You build it |
| Session replay on variants | Built in | Not offered | Not offered | Separate vendor |
| Warehouse metrics | Native sync | Export only | Varies | Your pipeline |
| AI readouts | Built in | Not offered | Varies | Rarely |
| Time to first test | Under a day | Days | Days | Months |
| Ongoing maintenance | Kerf | Kerf plus glue code | Kerf plus glue code | Your team |
Where each option fits
Single purpose flag tools
Good when you only need on and off switches. They fall short the day you want to know whether a change helped.
Standalone A/B tools
Good for marketing pages. For product changes, a second SDK and a second set of user IDs mean mismatched numbers.
In house systems
Good if experimentation is your product. For most teams, statistics and maintenance quickly outgrow the original plan.