- Service Limits
Split maintains service limits for each account to help guarantee the availability of the Split service, as well as to minimize billing risks to customers. Split service limits require that you request limit increases manually. In most cases, you aren’t charged if your service limits are increased. You’re only charged if the limit you are requesting an increase to is a billing limit. You can review and request increases in the admin settings.
- Metric Capping
The value of a particular metric can now be capped. This enables teams to minimize the impact of outliers in their data and remove the impact of bots, allowing for more meaningful experiment results, faster.
Split’s new workspaces release makes it easier for product & engineering teams to manage the experiments and feature flags across different product lines and applications. Split Admins can create workspaces to have separate areas within Split for each product line or application. Each user can then navigate between workspaces to manage their splits, segments, and metrics separately in each workspace. This release also includes a new version (v2) of our Admin API for programmatically working with splits, segments, traffic types, environments, attributes, and customer identities as we have added in workspace context to the endpoints.
- Metric Definition Builder Enhancements
Split’s new metric definition builder makes it easier for users to create metrics that are tailored to the needs of their organization.
- Dynamic minimum detectable effect
Split has now launched dynamic minimum detectable effect where statistically significant results can be obtained based on the dynamic relationship between the observed effect size and observed sample size of experiment metrics. For example, Split will indicate the sample size required for the given effect size to be statistically significant. Alternatively, we will also indicate the effect size needed for the results to be statistically significant with a given number of samples.
- Review periods
Customers can now set how long an experiment should run before results can be reviewed. Split will produce a warning on the metrics impact dashboard if an experiment has not met the minimum period.
- Sample ratio check
The Split platform will now perform a sample ratio check with each calculation update to monitor for a significant deviation between the targeted and current sample ratios.