Split vs. LaunchDarkly

LaunchDarkly and Split both offer feature flagging capabilities for accelerating development and mitigating release risk. Only Split provides the data privacy, feature-level monitoring, and experimentation best practices enterprises need to securely and effectively deliver impactful products. Compared to LaunchDarkly, Split’s features are unmatched! If you’re looking for a Launchdarkly alternative, check out this side-by-side comparison table and judge for yourself.

FEATURE MANAGEMENT & CONTROL
Feature Flagging
Feature Flag Essentials
Decouple deploy from release, so you can deploy at any time and release when ready.
Advanced Feature Flagging
Use sophisticated audience targeting and percentage-based rollouts to focus exposure and limit the blast radius of changes.
Scheduling
“Set it and forget it” scheduled changes: flag changes happen exactly when planned, without requiring late night logins for submitter or approver.
Approval Flows
Perform optional peer-review approvals or mandatory approvals including designated approvers to support separation of concerns.

No support in LaunchDarkly for a list of designated approvers or groups (submitter can ask anyone to approve their request). Must implement custom roles to remove approval powers.
Secure Client SDK Architecture
Client SDK makes decisions locally, never sends PII to vendor cloud for evaluation.

Local decision engine in all SDKs. PII never leaves your app.

Client SDKs always send data outside your app for evaluation.
Streaming Architecture
Push flag changes out within milliseconds to ensure instant control of feature exposure.

Built upon ably.com, an industry-leading commercially-supported solution with a global footprint.

Solution built in-house.
Multi-Environment Flag Control
Maintain separate targeting rules for test, staging, production or any other environment in your SDLC.
Trigger Flag State Change From External System
Allow an external system to change state of flag if issue is detected.

Can change flag to any state (not just on or off) via Admin API or integrations such as Jenkins.

Can only turn flag on or off via flag trigger URLs. Can make other state changes via API.
Broad SDK Support
Native support for most major programming languages + support for alternate usage pattern where no SDK exists.
FEATURE MANAGEMENT & CONTROL
Remote “Dynamic” Configurations
No-Code Dynamic Configuration
Enable non-technical users to enter dynamic configuration values as name:value pairs, instead of requiring them to write code.

Form-based name:value pair entry or compose as JSON

Must compose as JSON
Feature Measurement & Learning
Monitoring
Calculate all metrics for all rollouts
Avoid blind spots and reduce the risk of local optimizations that hurt organizational goals.

Detect unexpected consequences and maintain organizational SLOs by calculating all organizationally defined metrics against all percentage-based rollouts, all the time, with no manual effort.

Must start an experiment explicitly. Allows only one primary and a small number of secondary metrics, chosen by the flag owner, not the organization.
Feature-Level Metrics Alerts
End incidents faster with flag-specific causal analysis that determines which flag is causing issues.

Enforce organizational alerting thresholds, sending alerts to the flag owner and metrics stakeholders when thresholds are exceeded by a specific flag. All metrics are evaluated for all flags automatically, with no per-flag work.

APM integrations do not solve this problem. APMs do not perform flag-level causal analysis. Achieving any level of flag-specific awareness on the APM side requires ongoing manual work, costly in both time and talent.
Progressive Delivery Support
Address both the control and observability objectives of progressive delivery.

Feature control and observability are built-in.

Control only. Does not have on-board feature-aware monitoring or alerting.
FEATURE MEASUREMENT & LEARNING
Experimentation
Rigorous And Proven Statistical Approach
Is there a rigorously developed statistical engine behind the experimentation features?

Fully-implemented frequentist engine with sample ratio mismatch (SRM), multiple comparison correction (MCC), and review period notifications.

Incomplete Bayesian model. No ability to declare a prior (essential to Bayesian statistics).
Add Metrics Without Disruption
Add a metric anytime without being forced to start a new experiment.
Ingest events directly from Amazon S3
Bulk import millions of events per minute when needed for metrics calculation. Can be in near real-time or after the fact. Vastly more efficient than inserting via individual API calls.
Support For Trigger Testing
Trigger tests begin when a user reaches a specific part of your page or app. Ensures the experiment focuses on users who actually could be influenced by the feature.

All SDKs support just-in-time evaluations with in-memory decision-making, allowing local evaluations to be done only when user reaches feature in question.

Client SDKs have no decision engine. Must cache decisions at app launch or make costly round-trips to attempt trigger testing, skewing results.
Ready-To-Share Results In Multiple Formats
Experiment results can be downloaded as ready-to-present PDFs, spreadsheet-friendly CSVs, and developer-friendly JSON.

Requires API developer to process results for sharing outside the application.
Traffic Allocation

Ensures that users don’t move from one treatment/variation to another during ramp-up or ramp-down of the experiment.

Default “variation reassignment” re-allocates users when the ramp or number of treatments is changed.
Team and Organizational Features
Manage Permissions At Group/Team Level
Streamline administration of large groups, eliminating the need to set and maintain each user’s permissions by hand.

Users are assigned to one or more groups. Access to workspaces, permissions in environments, and object-level permission overrides can then be managed using groups.

Users are assigned to one or more teams. May require the creation and maintenance of complex custom roles on LaunchDarkly to achieve the desired control.
SCIM Support
Map group membership in your IdP to groups in the application, retaining control of provisioning and permissions control within your existing practices.

Map IdP groups to groups in Split. Grant those groups permissions as explained above. User provisioning and group membership are maintained directly in your IdP.

Requires addition and maintenance of LaunchDarkly-specific custom attributes within your IdP or use of Okta Group Push.
Pre-Assign Users To Groups/Teams At Invite Time
If not using SCIM, Users added via invite arrive with entitlements and permissions pre-established before the first login.
Integrate with ServiceNow to leverage existing approval processes.
Kanban-Style Flag Dashboard
View the status of all flags, grouped by lifecycle stage, in one place, with rich at-a-glance visualizations.

Rollout board visualizes flags by status, provides at-a-glance age and activity level across multiple environments for each flag. Advanced filters can be applied. Customized boards can be shared.

Flags list supports search and filtering but no visualization by status. More suitable for finding flags and acting on them than gaining an at-a-glance situational awareness.
Feature Management & Control
Feature FlaggingFeature Flag Essentials
Decouple deploy from release., so you can deploy at any time and release when ready.
Advanced Feature Flagging
Use sophisticated audience targeting and percentage-based rollouts to focus exposure and limit the blast radius of changes.
Scheduling
“Set it and forget it” scheduled changes: flag changes happen exactly when planned, without requiring late night logins for submitter or approver.
Approval Flows
Perform optional peer-review approvals or mandatory approvals including named approvers to support separation of concerns.

No support in LaunchDarkly for named approvers (anyone other than submitter can approve their request).
Secure Client SDK Architecture
Client SDK makes decisions locally, never sends PII to vendor cloud for evaluation.

Local decision engine in all SDKs. PII never leaves your app.

Client SDKs always send data outside your app for evaluation
Streaming Architecture
Push flag changes out within milliseconds to ensure instant control of feature exposure.

Built upon ably.com, an industry leading commercially-supported solution with a global footprint.

DIY solution built in-house.
Multi-Environment Flag Control
Maintain separate targeting rules for test, staging, production or any other environment in your SDLC.
Trigger Flag State Change From External System
Allow an external system to change state of flag if issue is detected.

Can change flag to any state (not just on or off) via Admin API or integrations such as Jenkins.

Can only turn flag on or off via flag trigger URLs. Can make other state changes via API.
Broad SDK Support
Native support for most major programming languages + support for alternate usage pattern where no SDK exists.
Remote “Dynamic” ConfigurationsNo-Code Dynamic Configuration
Enable non-technical users to enter dynamic configuration values as name:value pairs, instead of requiring them to write code.

Feature Measurement & Learning
MonitoringCalculate all metrics for all rollouts
Detect unexpected consequences and maintain organizational SLOs by calculating all metrics, not just one primary and 2-3 secondary metrics.
Feature-Level Metrics Alerts
Set organizational alerting thresholds and send alerts to flag owner and metrics stakeholders when thresholds are exceeded by a feature flag rollout.

Don’t be fooled by APM alert integrations. Achieving feature-level awareness on the APM side in order to send alerts is duplicate work, error-prone, and costly in time and talent.
Progressive Delivery Support
Address both the control and observability objectives of progressive delivery.

No observability support. Cannot perform feature-aware monitoring or alerting.
ExperimentationRigorous And Proven Statistical Approach
Is there a rigorously developed statistical engine behind the experimentation features?

Fully-implemented frequentist engine with sample ratio mismatch (SRM), multiple comparison correction (MCC), and review period notifications.

Incomplete Bayesian model. No ability to declare a prior (essential to Bayesian statistics).
Add Metrics Without Disruption
Add a metric anytime without being forced to start a new experiment.
Ingest events directly from Amazon S3
Bulk import millions of events per minute when needed for metrics calculation. Can be in near real-time or after the fact. Vastly more efficient than inserting via individual API calls.
Support For Trigger Testing
Trigger tests begin when a user reaches a specific part of your page or app. Ensures the experiment focuses on users who actually could be influenced by the feature.

All SDKs support just-in-time evaluations with in-memory decision-making, allowing local evaluations to be done only when user reaches feature in question.

Client SDKs have no decision engine. Must cache decisions at app launch or make costly round-trips to attempt trigger testing, skewing results.
Ready To Share Results In Multiple Formats

Experiment results can be downloaded as ready-to-present PDFs, spreadsheet-friendly CSVs, and developer-friendly JSON.

Requires API developer to retrieve and format results for sharing outside the application.
Traffic Allocation

Ensures that users don’t move from one treatment/variation to another during ramp-up or ramp-down of the experiment.

Default “variation reassignment” re-allocates users when the ramp or number of treatments is changed.
TEAM AND ORGANIZATIONAL FEATURES
Manage Permissions At Group/Team Level
Streamline administration of large groups, eliminating the need to set and maintain each user’s permissions by hand.

Users are assigned to one or more groups. Access to workspaces, permissions in environments, and object-level permission overrides can then be managed using groups.

Users are assigned to one or more teams. May require the creation and maintenance of complex custom roles on LaunchDarkly to achieve the desired control.
SCIM Support
Map group membership in your IdP to groups in the application, retaining control of provisioning and permissions control within your existing practices.

Map IdP groups to groups in Split. Grant those groups permissions as explained above. User provisioning and group membership are maintained directly in your IdP.

Requires addition and maintenance of LaunchDarkly-specific custom attributes within your IdP or use of Okta Group Push.
Pre-Assign Users To Groups/Teams At Invite Time
If not using SCIM, Users added via invite arrive with entitlements and permissions pre-established before the first login.
Integrate with ServiceNow to leverage existing approval processes.
Kanban-Style Flag Dashboard
View the status of all flags, grouped by lifecycle stage, in one place with rich at-a-glance visualizations.

Rollout board visualizes flags by status, provides at-a-glance age and activity level across multiple environments for each flag. Advanced filters can be applied. Customized boards can be shared.

Flags list supports search and filtering but no visualization by status. More suitable for finding flags and acting on them than gaining an at-a-glance situational awareness.

Customer Testimonials

Release with Confidence

“Split.io has been helping our team to ship code with more confidence, especially for features that are not yet production-ready. Feature flags are now an integral part of how we operate” – Yuka Moribe, Software Enginner, Vareto

Easy to Learn

“The Split LMS is really well done. It strikes the right balance of fun and getting people to use the tool. I had a really great moment today when discussing the split integration and setting up an A/B test and one teammate asked “have you implemented tracking yet.” — John Fly, VP of Engineering, Legacy

Maximize Impact

“Split provides our team with rich data on every feature allowing us to make accurate changes in our application. I can also easily share the metric data as PDF with my team so everyone knows why we have decided to move forward with a feature” – Miller Dugalech, Director of Digital Product Management, Quility Insurance

Private Data Stays inYour App

Unlike Launchdarkly, Split is built from the ground up with privacy in mind. When targeting on private attributes, such as email or demographics, our SDKs do the heavy lifting. Sensitive user data stays within your app or server and no private data (PII) is sent to Split. No user identifiable data is retained by default – as sharing sensitive user data to our cloud service is not required to target features.
Learn more about privacy

Split is a leader in Feature Management and Continuous Delivery

Essential Scheduling

Stop pushing back your releases because of Launchdarkly. Essential scheduling provides the capability to launch a feature on a certain date and time. This allows you to make changes to a feature flag and get approvals ahead of the release date which makes release planning and collaboration easier and more flexible while also increasing the likelihood to release on a target date.

Release New Features Up to 50x Faster!

Reimagine the software development process.
Get started today.